EFFICIENT DECODING USING LARGE AND SMALL GENERATIVE ARTIFICIAL INTELLIGENCE MODELS

§101 §103

CTNF 18/486,653 CTNF 98440 DETAILED ACTION Notice of Pre-AIA or AIA Status 07-03-aia AIA 15-10-aia The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA. Information Disclosure Statement The information disclosure statement(s) (IDS) submitted on 10/13/2023, 01/12/2024, and 12/18/2024 is/are in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement(s) is/are being considered by the examiner. 07-30-03-h AIA Claim Interpretation 07-30-03 AIA The following is a quotation of 35 U.S.C. 112(f): (f) Element in Claim for a Combination. – An element in a claim for a combination may be expressed as a means or step for performing a specified function without the recital of structure, material, or acts in support thereof, and such claim shall be construed to cover the corresponding structure, material, or acts described in the specification and equivalents thereof. The following is a quotation of pre-AIA 35 U.S.C. 112, sixth paragraph: An element in a claim for a combination may be expressed as a means or step for performing a specified function without the recital of structure, material, or acts in support thereof, and such claim shall be construed to cover the corresponding structure, material, or acts described in the specification and equivalents thereof. 07-30-05 The claims in this application are given their broadest reasonable interpretation using the plain meaning of the claim language in light of the specification as it would be understood by one of ordinary skill in the art. The broadest reasonable interpretation of a claim element (also commonly referred to as a claim limitation) is limited by the description in the specification when 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is invoked. As explained in MPEP § 2181, subsection I, claim limitations that meet the following three-prong test will be interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph: (A) the claim limitation uses the term “means” or “step” or a term used as a substitute for “means” that is a generic placeholder (also called a nonce term or a non-structural term having no specific structural meaning) for performing the claimed function; (B) the term “means” or “step” or the generic placeholder is modified by functional language, typically, but not always linked by the transition word “for” (e.g., “means for”) or another linking word or phrase, such as “configured to” or “so that”; and (C) the term “means” or “step” or the generic placeholder is not modified by sufficient structure, material, or acts for performing the claimed function. Use of the word “means” (or “step”) in a claim with functional language creates a rebuttable presumption that the claim limitation is to be treated in accordance with 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph. The presumption that the claim limitation is interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is rebutted when the claim limitation recites sufficient structure, material, or acts to entirely perform the recited function. Absence of the word “means” (or “step”) in a claim creates a rebuttable presumption that the claim limitation is not to be treated in accordance with 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph. The presumption that the claim limitation is not interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is rebutted when the claim limitation recites function without reciting sufficient structure, material or acts to entirely perform the recited function. Claim limitations in this application that use the word “means” (or “step”) are being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, except as otherwise indicated in an Office action. Conversely, claim limitations in this application that do not use the word “means” (or “step”) are not being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, except as otherwise indicated in an Office action. This application includes one or more claim limitations that do not use the word “means,” but are nonetheless being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, because the claim limitation(s) uses a generic placeholder that is coupled with functional language without reciting sufficient structure to perform the recited function and the generic placeholder is not preceded by a structural modifier. The following limitations are interpreted as invoking 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph: Claim 21 recites means for receiving, means for generating, means for projecting, and means for outputting. Claim 24 recites means for generating and means for autoregressively generating. Claim 25 recites means for generating and means for projecting. Claim 26 recites means for generating and means for concatenating. Claim 27 recites means for concatenating. Claim 28 recites means for generating and means for projecting. If applicant does not intend to have these limitations interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, applicant may: (1) amend the claim limitations to avoid them being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph (e.g., by reciting sufficient structure to perform the claimed function); or (2) present a sufficient showing that the claim limitations recite sufficient structure to perform the claimed function so as to avoid them being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph. Claim Rejections - 35 USC § 101 07-04-01 AIA 07-04 35 U.S.C. 101 reads as follows: Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title. Claims 1-30 are rejected under 35 U.S.C.101 because the claimed invention is directed to an abstract idea without significantly more. Step 1: Claims 11-20 are directed to a process. Claims 1-10 and 21-30 are directed to a machine or an article of manufacture. With respect to claim(s) 1, 11, 21, and 30: 2A Prong 1 : The claim(s) recite(s) an abstract idea. Specifically: generate/generating […] an embedding representation of the received input query in a first dimensionality ; ( Mental process – A person can mentally evaluate a query and generate (think of) an embedding representation of a certain dimension (i.e., size) – see MPEP § 2106.04(a)(2)(III)) project/projecting the embedding representation of the received input query into a projected representation of the received input query, wherein the projected representation comprises a representation in a second dimensionality ; ( Mental process – A person could mentally project (thinking of a another dimension) an embedding representation into a second dimensionality (an embedding of another dimension) – see MPEP § 2106.04(a)(2)(III)) generate/generating a response to the received input query using […] the projected representation ; ( Mental process – A person can evaluate a projected representation to generate (think of) a response to a query– see MPEP § 2106.04(a)(2)(III)) If claim limitations, under their broadest reasonable interpretation, cover performance of the limitations as a mental process, but for the recitation of generic computer components, then the claim limitations fall within the mathematical or mental process grouping of abstract ideas. Accordingly, the claim “recites” an abstract idea. 2A Prong 2 : The additional elements recited in the claim(s) do not integrate the abstract idea into a practical application, individually or in combination. Additional elements: (Claim 1) A processing system, comprising: at least one memory having executable instructions stored thereon; and one or more processors configured to execute the executable instructions in order to cause the processing system to : (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).) (Claim 11) A processor-implemented method, comprising : (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).) (Claim 21) A processing system, comprising : (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).) (Claim 30) A non-transitory computer-readable medium having executable instructions stored thereon which, when executed by one or more processors, cause the one or more processors to perform an operation comprising: (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).) (Claim 21) means for […] (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).) receive/receiving an input query for processing ; (Mere data gathering – Adding insignificant extra-solution activity of mere data gathering to the judicial exception – see § MPEP2106.05(g).) […] using a first generative artificial intelligence model […] (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).) (Claim 30) and wherein the second dimensionality is smaller than the first dimensionality; (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).) […] using a second generative artificial intelligence model […] (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).) output/outputting the generated response . (Adding insignificant extra-solution activity to the judicial exception – see § MPEP2106.05(g).) Since the claim as a whole, looking at the additional elements individually and in combination, does not contain any other additional elements that are indicative of integration into a practical application, the claim is directed to an abstract idea. 2B : The claim(s) do(es) not include additional elements that are sufficient to amount to significantly more than the judicial exception. Additional elements: (Claim 1) A processing system, comprising: at least one memory having executable instructions stored thereon; and one or more processors configured to execute the executable instructions in order to cause the processing system to : (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).) (Claim 11) A processor-implemented method, comprising : (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).) (Claim 21) A processing system, comprising : (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).) (Claim 30) A non-transitory computer-readable medium having executable instructions stored thereon which, when executed by one or more processors, cause the one or more processors to perform an operation comprising: (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).) (Claim 21) means for […] (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).) receive/receiving an input query for processing ; (Simply appending well-understood, routine, conventional activities previously known to the industry, specified at a high level of generality, to the judicial exception (WURC)- see MPEP § 2106.05(d)(ll)(i) - Receiving or transmitting data over a network, e.g., using the Internet to gather data, Symantec, 838 F.3d at 1321, 120 USPQ2d at 1362 (utilizing an intermediary computer to forward information).) […] using a first generative artificial intelligence model […] (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).) (Claim 30) and wherein the second dimensionality is smaller than the first dimensionality; (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).) […] using a second generative artificial intelligence model […] (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).) output/outputting the generated response . (Simply appending well-understood, routine, conventional activities previously known to the industry, specified at a high level of generality, to the judicial exception (WURC)- see MPEP § 2106.05(d)(ll)(i) - Receiving or transmitting data over a network, e.g., using the Internet to gather data, Symantec, 838 F.3d at 1321, 120 USPQ2d at 1362 (utilizing an intermediary computer to forward information).) Considering the additional elements individually and in combination, and the claim as a whole, the additional elements do not provide significantly more than the abstract idea. Therefore, the claim is not patent eligible . With respect to claim(s) 2, 12, and 22: 2A Prong 2 : The additional elements recited in the claim(s) do not integrate the abstract idea into a practical application, individually or in combination. Additional elements: wherein the first generative artificial intelligence model comprises a model including a larger number of parameters than a number of parameters included in the second generative artificial intelligence model . (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).) 2B : The claim(s) do(es) not include additional elements that are sufficient to amount to significantly more than the judicial exception. Additional elements: wherein the first generative artificial intelligence model comprises a model including a larger number of parameters than a number of parameters included in the second generative artificial intelligence model . (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).) Since the claim does not recite additional elements that either integrate the judicial exception into a practical application, nor provide significantly more than the judicial exception, the claim is not patent eligible. Therefore, the claim is not patent eligible . With respect to claim(s) 3, 13, and 23: 2A Prong 2 : The additional elements recited in the claim(s) do not integrate the abstract idea into a practical application, individually or in combination. Additional elements: wherein the first generative artificial intelligence model and the second generative artificial intelligence model comprise models trained together on a same target task such that the first generative artificial intelligence model and the second generative artificial intelligence model are trained on a same number of tokens . (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).) 2B : The claim(s) do(es) not include additional elements that are sufficient to amount to significantly more than the judicial exception. Additional elements: wherein the first generative artificial intelligence model and the second generative artificial intelligence model comprise models trained together on a same target task such that the first generative artificial intelligence model and the second generative artificial intelligence model are trained on a same number of tokens . (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).) Since the claim does not recite additional elements that either integrate the judicial exception into a practical application, nor provide significantly more than the judicial exception, the claim is not patent eligible. Therefore, the claim is not patent eligible . With respect to claim(s) 4, 14, and 24: 2A Prong 1 : The claim(s) recite(s) an abstract idea. Specifically: generate/generating the response using […] the projected representation […] to autoregressively generate a first set of tokens including a threshold number of tokens . ( Mental process – A person can evaluate a projected representation and generate a response autoregressively (i.e., one word at a time by considering prior generated words) including a threshold number of tokens (until a desired word limit) – see MPEP § 2106.04(a)(2)(III)) 2A Prong 2 : The additional elements recited in the claim(s) do not integrate the abstract idea into a practical application, individually or in combination. Additional elements: […] using the second generative artificial intelligence model […] (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).) […] the one or more processors are configured to cause the processing system to […] (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).) 2B : The claim(s) do(es) not include additional elements that are sufficient to amount to significantly more than the judicial exception. Additional elements: […] using the second generative artificial intelligence model […] (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).) […] the one or more processors are configured to cause the processing system to […] (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).) Since the claim does not recite additional elements that either integrate the judicial exception into a practical application, nor provide significantly more than the judicial exception, the claim is not patent eligible. Therefore, the claim is not patent eligible . With respect to claim(s) 5, 15, and 25: 2A Prong 1 : The claim(s) recite(s) an abstract idea. Specifically: generate/generating […] an updated embedding representation, the updated embedding representation comprising an embedding of the received input query and the generated first set of tokens in the first dimensionality; ( Mental process – A person can evaluate a query and generated tokens to generate (think of) an updated embedding representation – see MPEP § 2106.04(a)(2)(III)) project/projecting the updated embedding representation into a projected updated embedding representation of the received input query and the generated first set of tokens in the second dimensionality; ( Mental process – A person can evaluate a query and generated tokens to project via mind an updated embedding (thinking of the embedding) – see MPEP § 2106.04(a)(2)(III)) generate/generating, using […] the projected updated embedding representation, a second set of tokens including the threshold number of tokens. ( Mental process – A person can mentally generate second set of tokens (i.e., more words) including the threshold number of tokens (i.e., up to a limit of words) by using the projected updated embedding representation – see MPEP § 2106.04(a)(2)(III)) 2A Prong 2 : The additional elements recited in the claim(s) do not integrate the abstract idea into a practical application, individually or in combination. Additional elements: (Claim 5) wherein the one or more processors are further configured to cause the processing system to : (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).) […] using the first generative artificial intelligence model, […] (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).) […] using the second generative artificial intelligence model […] (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).) 2B : The claim(s) do(es) not include additional elements that are sufficient to amount to significantly more than the judicial exception. Additional elements: (Claim 5) wherein the one or more processors are further configured to cause the processing system to : (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).) […] using the first generative artificial intelligence model, […] (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).) […] using the second generative artificial intelligence model […] (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).) Since the claim does not recite additional elements that either integrate the judicial exception into a practical application, nor provide significantly more than the judicial exception, the claim is not patent eligible. Therefore, the claim is not patent eligible . With respect to claim(s) 6, 16, and 26: 2A Prong 1 : The claim(s) recite(s) an abstract idea. Specifically: wherein to generate/generating the response using […] the projected representation ( Mental process – A person can evaluate a projected representation to generate (think of) a response – see MPEP § 2106.04(a)(2)(III)) generate/generating one or more first tokens based on the projected representation ; ( Mental process – A person can generate one or more tokens (i.e., words) based on mentally evaluating a projected representation – see MPEP § 2106.04(a)(2)(III)) concatenate/concatenating the projected representation and information related to the generated one or more first tokens ; ( Mental process – A person can mentally concatenate (i.e., combine) a projected representation and information related to generated tokens (i.e., words) – see MPEP § 2106.04(a)(2)(III)) generate/generating a second token based on the concatenated projected representation and the information related to the generated one or more first tokens . ( Mental process – A person can generate a second token (i.e., second word) based on the concatenated projected representation and information related to generated tokens – see MPEP § 2106.04(a)(2)(III)) 2A Prong 2 : The additional elements recited in the claim(s) do not integrate the abstract idea into a practical application, individually or in combination. Additional elements: […] using the second generative artificial intelligence model […] (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).) (Claim 6) […] the one or more processors are configured to cause the processing system to : (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).) 2B : The claim(s) do(es) not include additional elements that are sufficient to amount to significantly more than the judicial exception. Additional elements: […] using the second generative artificial intelligence model […] (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).) (Claim 6) […] the one or more processors are configured to cause the processing system to : (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).) Since the claim does not recite additional elements that either integrate the judicial exception into a practical application, nor provide significantly more than the judicial exception, the claim is not patent eligible. Therefore, the claim is not patent eligible . With respect to claim(s) 7, 17, and 27: 2A Prong 1 : The claim(s) recite(s) an abstract idea. Specifically: wherein to concatenate the projected representation and the information related to the generated one or more first tokens ( Mental process – A person can mentally concatenate (i.e., combine) a projected representation and information related to generated tokens – see MPEP § 2106.04(a)(2)(III)) concatenate the projected representation and embedding representations of the one or more first tokens. ( Mental process – A person can mentally concatenate (i.e., combine) the projected representation and embedding representations of the one or more first tokens (i.e., words) – see MPEP § 2106.04(a)(2)(III)) 2A Prong 2 : The additional elements recited in the claim(s) do not integrate the abstract idea into a practical application, individually or in combination. Additional elements: (Claim 7) […] the one or more processors are configured to cause the processing system to […] (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).) 2B : The claim(s) do(es) not include additional elements that are sufficient to amount to significantly more than the judicial exception. Additional elements: (Claim 7) […] the one or more processors are configured to cause the processing system to […] (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).) Since the claim does not recite additional elements that either integrate the judicial exception into a practical application, nor provide significantly more than the judicial exception, the claim is not patent eligible. Therefore, the claim is not patent eligible . With respect to claim(s) 8, 18, and 28: 2A Prong 1 : The claim(s) recite(s) an abstract idea. Specifically: wherein to generate/generating the response using […] the projected representation , ( Mental process – A person can generate (think of) a response using a projected representation – see MPEP § 2106.04(a)(2)(III)) generate/generating one or more first tokens based on the projected representation; ( Mental process – A person can generate (think of) tokens (i.e., words) using a projected representation – see MPEP § 2106.04(a)(2)(III)) project/projecting a combination of the embedding representation and the one or more first tokens into a projected representation of the received input query and the one or more first tokens in the second dimensionality; ( Mental process – A person can mentally project a combination of an embedding representation and tokens into a projected representation of a received query and one or more tokens – see MPEP § 2106.04(a)(2)(III)) generate/generating a second token based on the projected representation of the received input query and the one or more first tokens. ( Mental process – A person can mentally generate tokens (i.e., words) based on a projected representation of a received input query and tokens – see MPEP § 2106.04(a)(2)(III)) 2A Prong 2 : The additional elements recited in the claim(s) do not integrate the abstract idea into a practical application, individually or in combination. Additional elements: […] using the second generative artificial intelligence model […] (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).) (Claim 8) […] the one or more processors are configured to cause the processing system to : (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).) 2B : The claim(s) do(es) not include additional elements that are sufficient to amount to significantly more than the judicial exception. Additional elements: […] using the second generative artificial intelligence model […] (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).) (Claim 8) […] the one or more processors are configured to cause the processing system to : (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).) Since the claim does not recite additional elements that either integrate the judicial exception into a practical application, nor provide significantly more than the judicial exception, the claim is not patent eligible. Therefore, the claim is not patent eligible . With respect to claim(s) 9, 19, and 29: 2A Prong 1 : The claim(s) recite(s) an abstract idea. Specifically: generate the embedding representation in the first dimensionality, ( Mental process – A person can mentally generate an embedding representation in a first dimensionality (i.e., of a certain dimension) – see MPEP § 2106.04(a)(2)(III)) generate a response based on an input in the second dimensionality. ( Mental process – A person can mentally generate a response based on an input in the second dimensionality (i.e., of a certain dimension) – see MPEP § 2106.04(a)(2)(III)) 2A Prong 2 : The additional elements recited in the claim(s) do not integrate the abstract idea into a practical application, individually or in combination. Additional elements: the first generative artificial intelligence model comprises a large language model trained to (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).) the second generative artificial intelligence model comprises a small language model trained to (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).) 2B : The claim(s) do(es) not include additional elements that are sufficient to amount to significantly more than the judicial exception. Additional elements: the first generative artificial intelligence model comprises a large language model trained to (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).) the second generative artificial intelligence model comprises a small language model trained to (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).) Since the claim does not recite additional elements that either integrate the judicial exception into a practical application, nor provide significantly more than the judicial exception, the claim is not patent eligible. Therefore, the claim is not patent eligible . With respect to claim(s) 10 and 20: 2A Prong 2 : The additional elements recited in the claim(s) do not integrate the abstract idea into a practical application, individually or in combination. Additional elements: wherein the second dimensionality is smaller than the first dimensionality . (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).) 2B : The claim(s) do(es) not include additional elements that are sufficient to amount to significantly more than the judicial exception. Additional elements: wherein the second dimensionality is smaller than the first dimensionality . (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).) Since the claim does not recite additional elements that either integrate the judicial exception into a practical application, nor provide significantly more than the judicial exception, the claim is not patent eligible. Therefore, the claim is not patent eligible . Claim Rejections - 35 USC § 103 07-20-aia AIA The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action: A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made. 07-21-aia AIA Claim s 1, 10-11, 20-21, and 30 are rejected under 35 U.S.C. 103 as being unpatentable over LI ("BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models") in view of CHEN (US 20200034436 A1), hereafter LI and CHEN 436’ respectively . Regarding Claim 1: LI teaches: A processing system, comprising: at least one memory having executable instructions stored thereon; and one or more processors configured to execute the executable instructions in order to cause the processing system to : (LI [page 1] teaches the source code at https://github.com/salesforce/LAVIS/tree/main/projects/blip2. A person of ordinary skill would recognize that the source code requires hardware such as a computer processor and a memory in order to run.) receive an input query for processing ; (LI [page 1, Figure 1] teaches the Q-Former receiving queries and text for processing. LI [page 7, section 4.3. Visual Question Answering] teaches: "the question tokens are given as input to the Q-Former".) generate, using a first generative artificial intelligence model, an embedding representation of the received input query in a first dimensionality ; (LI [page 3, section 3.1. Model Architecture] teaches: "We initialize Q-Former with the pre-trained weights of BERTbase [...]" LI [page 3, section 3.2. Image-Text Matching] teaches: "The output query embeddings Z thus capture multimodal information." LI [page 3, section 3.1. Model Architecture] teaches: "We use Z to denote the output query representation. The size of Z (32 × 768) is much smaller than the size of frozen image features (e.g. 257 × 1024 for ViT-L/14)." Examiner's note: Under BRI, generate, a first generative artificial intelligence model, an embedding representation can be interpreted as the Q-former, which generates query representations (i.e., an embedding representation ) of images and text inputs from a user (i.e., of the received input query ). Additionally, in a first dimensionality can be interpreted as the query embeddings dimensions when they are outputted by the Q-former.) project the embedding representation of the received input query into a projected representation of the received input query, wherein the projected representation comprises a representation in a […] dimensionality ; (LI [page 4, section 3.3. Bootstrap Vision-to-Language Generative Learning from a Frozen LLM] teaches: "we use a fully-connected (FC) layer to linearly project the output query embeddings Z (i.e., project the embedding representation ) into the same dimension (i.e., into a projected representation , wherein the projected representation comprises a representation in a […] dimensionality ) as the text embedding of the LLM." LI [page 1, Figure 1] teaches the Q-Former receiving queries and text for processing. LI [page 7, section 4.3. Visual Question Answering] teaches: "the question tokens (i.e., of the received input query ) are given as input to the Q-Former". Examiner’s note: Under BRI, dimensionality can be interpreted as the dimension into which the query embeddings Z are being projected.) generate a response to the received input query using a second generative artificial intelligence model and the projected representation ; (LI [page 13, Figure 7] teaches: "the LLM (i.e., using a second generative artificial intelligence model ) receives Q-Former’s output (i.e., and the projected representation ) and the question as input, then predicts answers (i.e., generate a response to the received input query ).") output the generated response . (LI [page 13, Figure 7] teaches: "the LLM receives Q-Former’s output and the question as input, then predicts answers." LI [page 5, Figure 4] teaches the presenting the generated response (i.e., output the generated response ) to the user’s query.) However, LI is not relied upon for teaching, but CHEN 436’ teaches: wherein the projected representation comprises a representation in a second dimensionality (CHEN 436’ [0092] teaches: "Additional projection aims to reduce the dimensionality of the encoder output representations to match the decoder stack dimension." CHEN 436’ [0128] teaches: "After concatenation, the combined representation is projected down to the decoder dimension (i.e., a representation in a second dimensionality ) with a layer-normalized affine transformation." CHEN 436’ [0117] teaches: "As illustrated in FIG. 5, a multi-column encoder merges the outputs of several independent encoders into a single combined representation.”) Accordingly, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention, having the teachings of LI and CHEN 436’ before them, to include CHEN 436’s dimensionality reduction of encoder output representations in LI’s text generation method. One would have been motivated to make such a combination in order to reduce the dimensionality of the encoder output representations to match the decoder stack dimension (CHEN 436’ [0092]). Regarding Claim 10: LI in view of CHEN 436’ teaches the elements of claim 1 as outlined above. CHEN 436’ further teaches: wherein the second dimensionality is smaller than the first dimensionality . (CHEN 436’ [0128] teaches: "After concatenation, the combined representation is projected down to the decoder dimension with a layer-normalized affine transformation." CHEN 436’ [0117] teaches: "As illustrated in FIG. 5, a multi-column encoder merges the outputs of several independent encoders into a single combined representation." CHEN 436’ [0092] teaches: "Additional projection aims to reduce the dimensionality of the encoder output representations to match the decoder stack dimension (i.e., the second dimensionality is smaller than the first dimensionality ).") Regarding Claim 11: The claim recites similar limitations as corresponding claim 1 and is rejected for similar reasons as claim 1 using similar teachings and rationale. Regarding Claim 20: LI in view of CHEN 436’ teaches the elements of claim 11 as outlined above. Additionally, the claim recites similar limitations as corresponding claim 10 and is rejected for similar reasons as claim 10 using similar teachings and rationale. Regarding Claim 21: The claim recites similar limitations as corresponding claim 1 and is rejected for similar reasons as claim 1 using similar teachings and rationale. Regarding Claim 30: The claim recites similar limitations as corresponding claims 1 and 10 and is rejected for similar reasons as claims 1 and 10 using similar teachings and rationale . 07-21-aia AIA Claim s 2, 12, and 22 are rejected under 35 U.S.C. 103 as being unpatentable over LI in view of CHEN 436’ as applied respectively above to claims 1, 11, and 21, and further in view of KIM ("Speculative Decoding with Big Little Decoder"), hereafter KIM . Regarding Claim 2: LI in view of CHEN 436’ teaches the elements of claim 1 as outlined above. LI further teaches: wherein the first generative artificial intelligence model comprises a model including a […] number of parameters than a number of parameters included in the second generative artificial intelligence model . (LI [page 13, Figure 7] teaches: "the LLM receives Q-Former’s output and the question as input, then predicts answers." LI [page 3, section 3.1. Model Architecture] teaches: "In total, Q-Former contains 188M parameters (i.e., a model including a […] number of parameters than a number of parameters)." LI [page 6, table 2. shows the different number of parameters for the LLMs, such as 2.7B, 6.7B, etc. It should be noted that the LLM (i.e., the second generative artificial intelligence model ) in LI have more parameters than the Q-Former (i.e., the first generative artificial intelligence model ).) LI in view of CHEN 436’ is not relied upon for teaching, but KIM teaches: […] a model including a larger number of parameters than a number of parameters included in the […] generative artificial intelligence model . (KIM [page 1, Abstract] teaches: "The small model (i.e., generative artificial intelligence model ) runs autoregressively to generate text with a low inference cost, and the large model is only invoked occasionally to refine the small model’s inaccurate predictions in a non-autoregressive manner." KIM [page 2, section 1 Introduction] teaches: "We propose two policies for BiLD: the fallback policy that allows the small model to hand control over to the large model if it is not confident enough (Section 3.3), and the rollback policy that allows the large model to review and correct the small model’s inaccurate predictions (Section 3.4)." KIM [page 15, Table 2] teaches that the T5-small (e.g., small language model) has a total of 25M parameters and that the T5-large has a total of 402M parameters.) Accordingly, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention, having the teachings of LI, CHEN 436’, and KIM before them to modify LI’s LLM using KIM's T5-small model in LI and CHEN 436's text generation method to reduce autoregressive inference costs. As a result of this modification, LI's Q-Former (188M parameters) would result in a larger number of parameters than the T-5 small language model (25M parameters). Therefore, the modification for reducing autoregressive inference costs causes the first generative artificial intelligence model to have more parameters than the second generative artificial intelligence model, and thus teaching the first generative artificial intelligence model comprises a model including a larger number of parameters than a number of parameters included in the second artificial intelligence model . One would have been motivated to make such a combination in order to start inference with a small model which runs autoregressively for the majority of the time to generate text with a low inference cost (KIM [page 10, section 5 Conclusion]). Regarding Claim 12: LI in view of CHEN 436’ teaches the elements of claim 11 as outlined above. Additionally, the claim recites similar limitations as corresponding claim 2 and is rejected for similar reasons as claim 2 using similar teachings and rationale. Regarding Claim 22: LI in view of CHEN 436’ teaches the elements of claim 21 as outlined above. Additionally, the claim recites similar limitations as corresponding claim 2 and is rejected for similar reasons as claim 2 using similar teachings and rationale . 07-21-aia AIA Claim s 3, 13, and 23 are rejected under 35 U.S.C. 103 as being unpatentable over LI in view of CHEN 436’ as applied respectively above to claims 1, 11, and 21, and further in view of BIDERMAN ("Pythia: A Suite for Analyzing Large Language Models Across Training and Scaling"), hereafter BIDERMAN . Regarding Claim 3: LI in view of CHEN 436’ teaches the elements of claim 1 as outlined above. LI in view of CHEN 436’ is not relied upon for teaching, but BIDERMAN teaches: wherein the first generative artificial intelligence model and the second generative artificial intelligence model comprise models trained together on a same target task such that the first generative artificial intelligence model and the second generative artificial intelligence model are trained on a same number of tokens . (BIDERMAN [page 1, section 1. Introduction] teaches: "The Pythia suite is the only publicly released suite of LLMs that satisfies three key properties: 1. Models span several orders of magnitude of model scale. 2. All models were trained on the same data in the same order." BIDERMAN [page 17, section B. Correction and Updates] teaches: "All model sizes are now trained with uniform batch size of 2M tokens (i.e., model trained on a same number of tokens )." Examiner’s note: Under BRI, trained together on a same target task can be interpreted as all models having being trained on the same data in the same order.) Accordingly, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention, having the teachings of LI, CHEN 436', and BIDERMAN before them, to include BIDERMAN’s (model training using same data and same batch size of tokens in LI and CHEN 436’s text generation method. One would have been motivated to make such a combination in order to maintain consistency across all model training runs (BIDERMAN [page 4, section 2.4. Training]). Regarding Claim 13: LI in view of CHEN 436’ teaches the elements of claim 11 as outlined above. Additionally, the claim recites similar limitations as corresponding claim 3 and is rejected for similar reasons as claim 3 using similar teachings and rationale. Regarding Claim 23: LI in view of CHEN 436’ teaches the elements of claim 21 as outlined above. Additionally, the claim recites similar limitations as corresponding claim 3 and is rejected for similar reasons as claim 3 using similar teachings and rationale . 07-21-aia AIA Claim s 4-8, 14-18, and 24-28 are rejected under 35 U.S.C. 103 as being unpatentable over LI in view of CHEN 436’ as applied respectively above to claims 1, 11, and 21, and further in view of MENG (US 20250094233 A1), hereafter MENG . Regarding Claim 4: LI in view of CHEN 436’ teaches the elements of claim 1 as outlined above. LI further teaches: wherein to generate the response using the second generative artificial intelligence model and the projected representation […] (LI [page 4, section 3.3. Bootstrap Vision-to-Language Generative Learning from a Frozen LLM] teaches: "we use a fully-connected (FC) layer to linearly project the output query embeddings Z into the same dimension as the text embedding of the LLM." LI [page 13, Figure 7] teaches: "the LLM receives Q-Former’s output and the question as input, then predicts answers.") LI in view of CHEN 436’ is not relied upon for teaching, but MENG teaches: the one or more processors are configured to cause the processing system to autoregressively generate a first set of tokens including a threshold number of tokens . (MENG [0068] teaches: "One or more applications 512 (e.g., the request segmentation engine of FIG. 2, the trained sequential model 206 of FIG. 2) are loaded in the memory device(s) 504 and executed on the operating system 510 by the processor unit(s) 502 (i.e., one or more processors are configured to cause the processing system )." MENG [0021] teaches: "In one implementation, the disclosed techniques are applied to a language model that has been trained to receive an input sequence and generate an output sequence in an iterative fashion, such as by iteratively predicting a first word following the sequence, then a second word following the sequence, a third word, etc., where each “next word” generated is selected by the model as mathematically most likely to follow the input sequence and/or the previously-generated word(s) in the output sequence (i.e., to autoregressively generate a first set of tokens )." MENG [0025] teaches: "When performing a single NLP task, the decoder portion is iteratively queried with an input sequence that grows in length by one token each iteration." MENG [0026] teaches: "In this architecture, an encoder processes the input sequence to produce a continuous representation, or embedding, of the input before the embedding is provided to a decoder that iteratively generates the output sequence (e.g., one word at a time) (i.e., to autoregressively generate )." MENG [0028] teaches: "In the example shown, the processing request 110 specifies the output size parameter 128 (e.g., “max_token”) (i.e., threshold number of tokens ) which is set to 50 tokens. In an implementation where each token corresponds to one word of model output, this 50 token cap serves to inform the trained sequential model 104 that it is to reserve a quantity of memory sufficient for processing of up to 50 words and use this memory to generate the model output, which has a maximum length of 50 words." MENG [0015] teaches: “Since it is difficult to intelligently predict the exact data size of model output that will be needed to answer a given LLM input (text query), it is common practice for the specified output size parameter to serve as an upper bound on the permissible length of the model output. In a typical scenario, the model generates outputs until either the outputs reach a predefined size (i.e., including a threshold number of tokens ) or until a stop logic criterion is satisfied, where both the predefined size and the stop logic criterion are determined based on input parameters specified by the requesting application.) Accordingly, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention, having the teachings of LI, CHEN 436', and MENG before them, to include MENG’s LLM autoregressive generation and max token output size in LI and CHENG 436’s text generation method. One would have been motivated to make such a combination because since it is difficult to intelligently predict the exact data size of model output that will be needed to answer a given LLM input (text query), it is common practice for the specified output size parameter to serve as an upper bound on the permissible length of the model output (MENG [0015]). Regarding Claim 5: LI in view of CHEN 436’ and MENG teaches the elements of claim 4 as outlined above. LI further teaches: wherein the one or more processors are further configured to cause the processing system to: generate, using the first generative artificial intelligence model, an updated embedding representation, […] (LI [page 1] teaches the source code at https://github.com/salesforce/LAVIS/tree/main/projects/blip2. A person of ordinary skill would recognize that the source code requires hardware such as a computer processor and a memory in order to run. LI [page 3, section 3.2. Image-Text Matching] teaches: "The output query embeddings Z thus capture multimodal information." LI [page 3, section 3.1. Model Architecture] teaches: "We use Z to denote the output query representation. The size of Z (32 × 768) is much smaller than the size of frozen image features (e.g. 257 × 1024 for ViT-L/14).” LI [page 6, section 4.1. Instructed Zero-shot Image-to-Text Generation] teaches: "BLIP-2 effectively enables a LLM to understand images while preserving its capability in following text prompts, which allows us to control image-to-text generation with instructions. We simply append the text prompt after the visual prompt as input to the LLM." LI [page 5, Figure 4] shows various examples such as the user submitting "Which city is this?", "Singapore", "Why do you think so?", "The city has a statue of a merlion." Examiner's note: LI [page 5, Figure 4] demonstrates that the LLM preserves previous user queries in the conversation to answer a question like "Why do you think so?" corresponding the a previous response regarding the city. Under BRI, an updated embedding representation can be interpreted as the Q-former generating embeddings based on the preserved capabilities from prior text prompts for the LLM decoder.) project the updated embedding representation into a projected updated embedding representation of the received input query and […] tokens […] (LI [page 4, section 3.3. Bootstrap Vision-to-Language Generative Learning from a Frozen LLM] teaches: "we use a fully-connected (FC) layer to linearly project the output query embeddings Z into the same dimension as the text embedding of the LLM." LI [page 6, section 4.1. Instructed Zero-shot Image-to-Text Generation] teaches: "BLIP-2 effectively enables a LLM to understand images while preserving its capability in following text prompts, which allows us to control image-to-text generation with instructions. We simply append the text prompt after the visual prompt as input to the LLM." LI [page 5, Figure 4] shows various examples such as the user submitting "Which city is this?", "Singapore", "Why do you think so?", "The city has a statue of a merlion." Examiner's note: LI [page 5, Figure 4] demonstrates that the LLM preserves previous user queries in the conversation to answer a question like "Why do you think so?" corresponding the a previous response regarding the city. Under BRI, a projected updated embedding representation of the received input query can be interpreted as the fully connected layer projecting the representation of the generated embeddings by the Q-former (i.e., project the updated embedding representation ) based on the preserved capabilities from prior text prompts for generating a second response by LLM decoder, such as "The city has a statue of a merlion." In LI’s Figure 4. Additionally, into a projected updated embedding representation of […] tokens can be interpreted as the tokens used to generate an embedding representation for generating a second response by the LLM (e.g. "The city has a statue of a merlion.") based on the preserved capabilities from prior text prompts.) generate, using the second generative artificial intelligence model and the projected updated embedding representation, a second set of tokens […] (LI [page 13, Figure 7] teaches: "the LLM (i.e., second generative artificial intelligence model ) receives Q-Former’s output and the question as input, then predicts answers." LI [page 6, section 4.1. Instructed Zero-shot Image-to-Text Generation] teaches: "BLIP-2 effectively enables a LLM to understand images while preserving its capability in following text prompts, which allows us to control image-to-text generation with instructions. We simply append the text prompt after the visual prompt as input to the LLM." LI [page 5, Figure 4] shows various examples such as the user submitting "Which city is this?", "Singapore", "Why do you think so?", "The city has a statue of a merlion." Examiner's note: LI [page 5, Figure 4] demonstrates that the LLM preserves previous user queries in the conversation to answer a question like "Why do you think so?" corresponding the a previous response regarding the city. Under BRI, a second set of tokens can be interpreted as the tokens used to generated the second LLM response (e.g. "The city has a statue of a merlion.") in the examples of LI's Figure 4. Additionally, the projected updated embedding representation can be interpreted as the fully connected layer generating the projected representation of the Q-former generated embeddings based on the preserved capabilities from prior text prompts for the LLM decoder to generate the second response “The city has a statue of a merlion”.) MENG further teaches: generate, […] an updated embedding representation, the updated embedding representation comprising an embedding of the received input query and the generated first set of tokens in the first dimensionality ; (MENG [0051] teaches: "The request segmentation engine 210 generates a first partial processing request, Request_1, that is identical to the initial request 216 except for the value of the output size parameter (max_token), which has been reduced from the 1900 tokens specified in the initial request to a select chunk size of 500 tokens. In addition, the first partial processing request includes the initial text string (“Generate a three day Paris Itinerary”), referenced as “Request” in both FIG. 2 and the following description. The trained sequential model 206 executes the first partial processing request, Request_1, and outputs a first response, R1, that has a length of 500 tokens." MENG [0052] teaches: " Following receipt of R1, the request segmentation engine 210 generates and transmits a second partial processing request, Request_2 to the trained sequential model 206. The input text string specified in Request_2 includes the Request (“Generate a three day Paris Itinerary”) and R1 (e.g., the first 500 tokens of output are appended to the end of the request), with R1 being appended to the end of the Request." MENG [0053] teaches: "Following receipt of R2, the request segmentation engine 210 generates and transmits a third partial processing request, Request_3. The input text string specified in Request_3 includes the Request (“Generate a three day Paris Itinerary”), R1, and R2, with R1 being appended to the end of the Request and R2 being appended to the end of R1.” MENG [0025] teaches: "When performing a single NLP task, the decoder portion is iteratively queried with an input sequence that grows in length by one token each iteration." MENG [0026] teaches: "In this architecture, an encoder processes the input sequence to produce a continuous representation, or embedding, of the input before the embedding is provided to a decoder that iteratively generates the output sequence (e.g., one word at a time)." ." MENG [0015] teaches: “Since it is difficult to intelligently predict the exact data size of model output that will be needed to answer a given LLM input (text query), it is common practice for the specified output size parameter to serve as an upper bound on the permissible length of the model output. In a typical scenario, the model generates outputs until either the outputs reach a predefined size or until a stop logic criterion is satisfied, where both the predefined size and the stop logic criterion are determined based on input parameters specified by the requesting application.” Examiner's note: MENG [0026] and [0015] discloses various architectures readily understood by those of skill in the art, such as encoder-decoder architectures, in which an encoder receives a processing request to generate a token per iteration until a predefined size or until a stop logic criterion is satisfied. Additionally, MENG [0026] teaches that the encoder can produce embeddings before these embeddings are provided to the decoder for iterative text generation. Under BRI, updated embedding representation can be interpreted as the embedding that MENG’s encoder produces at each iteration after the first. At iteration 2, the encoder embeds the input query together with R1 and sends the embedding to the decoder; at iteration 3, the encoder embeds the input with R1 and R2; and so on, as supported by MENG [0026] for embedding prior to the decoder receiving the inputs, and the iterations in MENG [0050-0053], [FIG.1], and [FIG. 2]. Furthermore, MENG [FIG. 1] illustrates the iterative generation process by generating two tokens at a time, where request 2 includes output tokens “Mary” “had”, and request 3 in the next iteration includes “Mary” “had” “a” “little”, in order to produce the answer. Under BRI, the updated embedding representation comprising an embedding of the received input query and the generated first set of tokens in the first dimensionality can be interpreted as the embeddings received iteratively by the trained sequential model, that based on MENG [0026] can be embedded during each iteration prior to the decoder receiving them. Additionally, the first dimensionality can be interpreted as the dimension in which the tokens and input request are embedded.) […] updated embedding representation of the received input query and the generated first set of tokens in the […] dimensionality ; (Under BRI, updated embedding representation comprising an embedding of the received input query and the generated first set of tokens in the […] dimensionality can be interpreted as the embeddings received iteratively by the trained sequential model, that based on MENG [0026] can be embedded during each iteration prior to the decoder receiving them.) generate, using the […] generative artificial intelligence model and the […] updated embedding representation, a second set of tokens including the threshold number of tokens . (MENG [0025] teaches: "When performing a single NLP task, the decoder portion is iteratively queried with an input sequence that grows in length by one token each iteration." MENG [0026] teaches: "In this architecture, an encoder processes the input sequence to produce a continuous representation, or embedding, of the input before the embedding is provided to a decoder that iteratively generates the output sequence (e.g., one word at a time)" MENG [0028] teaches: "In the example shown, the processing request 110 specifies the output size parameter 128 (e.g., “max_token”) which is set to 50 tokens. In an implementation where each token corresponds to one word of model output, this 50 token cap serves to inform the trained sequential model 104 that it is to reserve a quantity of memory sufficient for processing of up to 50 words and use this memory to generate the model output, which has a maximum length of 50 words." Examiner’s note: Under BRI, generate, using […] the generative artificial intelligence model and the […] updated embedding representation, a second set of tokens including the threshold number of tokens can be interpreted as the embeddings received iteratively by the trained sequential model, that based on MENG [0026], can be embedded during each iteration prior to the decoder receiving them for generating answers. Additionally, a second set of tokens can be interpreted as a second input request by the user, such as the “Generate a three day paris itinerary” question as shown in MENG [FIG. 2], which will be generated until a predefined size set by the max_token parameter, as disclosed in MENG [0015].) CHEN 436’ further teaches: […] embedding representation […] in the second dimensionality […] (CHEN 436’ [0092] teaches: "Additional projection aims to reduce the dimensionality of the encoder output representations to match the decoder stack dimension." CHEN 436’ [0128] teaches: "After concatenation, the combined representation is projected down to the decoder dimension (i.e., embedding representation […] in the second dimensionality ) with a layer-normalized affine transformation." CHEN 436’ [0117] teaches: "As illustrated in FIG. 5, a multi-column encoder merges the outputs of several independent encoders into a single combined representation.”) Regarding Claim 6: LI in view of CHEN 436’ teaches the elements of claim 1 as outlined above. LI further teaches: wherein to generate the response using the second generative artificial intelligence model and the projected representation, the one or more processors are configured to cause the processing system to: generate one or more first tokens based on the projected representation ; (LI [page 1] teaches the source code at https://github.com/salesforce/LAVIS/tree/main/projects/blip2. A person of ordinary skill would recognize that the source code requires hardware such as a computer processor and a memory in order to run. LI [page 4, section 3.3. Bootstrap Vision-to-Language Generative Learning from a Frozen LLM] teaches: "The projected query embeddings (i.e., the projected representation ) are then prepended to the input text embeddings. [...] For decoder-based LLMs, we pre-train with the language modeling loss, where the frozen LLM (i.e., the second generative artificial intelligence model ) is tasked to generate the text (i.e., wherein to generate the response using the second generative artificial intelligence model / generate one or more first tokens ) conditioned on the visual representation from Q-Former (i.e., based on the projected representatio n)." LI [page 13, Figure 7] teaches the fully connected layer generating a projected query representation Z and prepending the query representation to the input text embedding for generating the answer by the LLM.) concatenate the projected representation and information […] (LI [page 4, section 3.3. Bootstrap Vision-to-Language Generative Learning from a Frozen LLM] teaches: "The projected query embeddings (i.e., the projected representation ) are then prepended (i.e., concatenate ) to the input text embeddings (i.e., and information )." LI [page 13, Figure 7] teaches the fully connected layer generating a projected query representation Z and prepending the query representation to the input text embedding for generating the answer by the LLM.) generate a second token based on the concatenated projected representation […] (LI [page 4, section 3.3. Bootstrap Vision-to-Language Generative Learning from a Frozen LLM] teaches: "The projected query embeddings are then prepended to the input text embeddings." LI [page 5, Figure 4] shows various examples such as the user submitting "Which city is this?", "Singapore", "Why do you think so?", "The city has a statue of a merlion." Examiner's note: LI [page 5, Figure 4] demonstrates that the LLM preserves previous user queries in the conversation to answer a question like "Why do you think so?" corresponding the a previous response regarding the city. Under BRI, generate a second token can be interpreted as the second token (i.e., word) generated by the LLM when responding to the input (e.g. "The city has a statue of a merlion."), as shown in the examples of LI's Figure 4, which is based on the prepended projected embeddings to the input embeddings (i.e., based on the concatenated projected representation ).) LI in view of CHEN 436’ is not relied upon for teaching, but MENG teaches: concatenate […] information related to the generated one or more first tokens ; (MENG [0025] teaches: "When performing a single NLP task, the decoder portion is iteratively queried with an input sequence that grows in length by one token each iteration. On the first iteration, the decoder input sequence includes the initial input to the model (e.g., the user input). The decoder generates a first word of an output sequence. Then, on the second decoder iteration, the decoder input includes the initial input to the model followed by the first word of the output sequence. The decoder outputs a second word of the output sequence. Then, on the third iteration, the decoder input includes the initial input to the model followed by the first two words of the outputs sequence. The decoder outputs a third word of the output sequence. This continues until the output sequence is completed (e.g., the output sequence reaches a pre-defined cap set on the permissible output size or the trained sequential model 104 otherwise determines that an applicable stop logic criterion is satisfied)." MENG [FIG.1] and [FIG. 2] show the trained sequential model receiving combined (i.e., concatenated ) inputs that include the generated tokens such as “Mary” “had” in FIG. 1, for example. Furthermore, information related to the generated one or more first tokens can be interpreted as the tokens generated by the trained sequential model.) generate a second token based on […] the information related to the generated one or more first tokens . (MENG [FIG.1] and [FIG. 2] show the trained sequential model receiving combined (i.e., concatenated ) inputs that include the generated tokens such as “Mary” “had” in FIG. 1, for example. Additionally, MENG [FIG. 1] teaches a second decoder iteration, where “a” “little” are the generated outputs. Under BRI, generate a second token based on […] the information related to the generated one or more first tokens can be interpreted as the tokens generated in the second decoder iteration using the previous input request and the tokens generated in the previous decoder generation (e.g., “Mary” “had”), as described in MENG [0025] and shown in MENG [FIG. 1].) Accordingly, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention, having the teachings of LI, CHEN 436', and MENG before them, to include MENG’s LLM autoregressive generation based on a max token output size in LI and CHENG 436’s text generation method. One would have been motivated to make such a combination because since it is difficult to intelligently predict the exact data size of model output that will be needed to answer a given LLM input (text query), it is common practice for the specified output size parameter to serve as an upper bound on the permissible length of the model output (MENG [0015]). Regarding Claim 7: LI in view of CHEN 436’ and MENG teaches the elements of claim 6 as outlined above. LI further teaches: wherein to concatenate the projected representation and the information […], the one or more processors are configured to cause the processing system to […] (LI [page 4, section 3.3. Bootstrap Vision-to-Language Generative Learning from a Frozen LLM] teaches: "The projected query embeddings are then prepended (i.e., concatenate ) to the input text embeddings (i.e., and information )." LI [page 13, Figure 7] teaches the fully connected layer generating a projected query representation Z and prepending the query representation to the input text embedding for generating the answer by the LLM. LI [page 1] teaches the source code at https://github.com/salesforce/LAVIS/tree/main/projects/blip2. A person of ordinary skill would recognize that the source code requires hardware such as a computer processor and a memory in order to run.) […] concatenate the projected representation and embedding representations [...] . (LI [page 4, section 3.3. Bootstrap Vision-to-Language Generative Learning from a Frozen LLM] teaches: "The projected query embeddings are then prepended (i.e., concatenate ) to the input text embeddings (i.e., and embedding representations )." LI [page 13, Figure 7] teaches the fully connected layer generating a projected query representation Z and prepending the query representation to the input text embedding for generating the answer by the LLM.) MENG further teaches: concatenate […] the information related to the generated one or more first tokens […] (MENG [0025] teaches: "When performing a single NLP task, the decoder portion is iteratively queried with an input sequence that grows in length by one token each iteration. On the first iteration, the decoder input sequence includes the initial input to the model (e.g., the user input). The decoder generates a first word of an output sequence. Then, on the second decoder iteration, the decoder input includes the initial input to the model followed by the first word of the output sequence. The decoder outputs a second word of the output sequence. Then, on the third iteration, the decoder input includes the initial input to the model followed by the first two words of the outputs sequence. The decoder outputs a third word of the output sequence. This continues until the output sequence is completed (e.g., the output sequence reaches a pre-defined cap set on the permissible output size or the trained sequential model 104 otherwise determines that an applicable stop logic criterion is satisfied)." MENG [FIG.1] and [FIG. 2] show the trained sequential model receiving combined (i.e., concatenated ) inputs that include the generated tokens such as “Mary” “had” in FIG. 1, for example. Furthermore, information related to the generated one or more first tokens can be interpreted as the tokens generated by the trained sequential model.) […] embedding representations of the one or more first tokens . (MENG [0025] teaches: "When performing a single NLP task, the decoder portion is iteratively queried with an input sequence that grows in length by one token each iteration. On the first iteration, the decoder input sequence includes the initial input to the model (e.g., the user input). The decoder generates a first word of an output sequence. Then, on the second decoder iteration, the decoder input includes the initial input to the model followed by the first word of the output sequence. The decoder outputs a second word of the output sequence. Then, on the third iteration, the decoder input includes the initial input to the model followed by the first two words of the outputs sequence. The decoder outputs a third word of the output sequence. This continues until the output sequence is completed (e.g., the output sequence reaches a pre-defined cap set on the permissible output size or the trained sequential model 104 otherwise determines that an applicable stop logic criterion is satisfied)." MENG [0026] teaches: "In this architecture, an encoder processes the input sequence to produce a continuous representation, or embedding, of the input before the embedding is provided to a decoder that iteratively generates the output sequence (e.g., one word at a time).” MENG [FIG.1] and [FIG. 2] show the trained sequential model receiving combined (i.e., concatenated ) inputs that include the generated tokens such as “Mary” “had” in FIG. 1, for example. Furthermore, embedding representations of the one or more first tokens can be interpreted as the tokens generated by the trained sequential model that are embedded for subsequent decoder iterations to generate the next tokens, as supported by MENG [0026].) Regarding Claim 8: LI in view of CHEN 436’ teaches the elements of claim 1 as outlined above. LI further teaches: wherein to generate the response using the second generative artificial intelligence model and the projected representation, the one or more processors are configured to cause the processing system to : (LI [page 1] teaches the source code at https://github.com/salesforce/LAVIS/tree/main/projects/blip2. A person of ordinary skill would recognize that the source code requires hardware such as a computer processor and a memory in order to run. LI [page 4, section 3.3. Bootstrap Vision-to-Language Generative Learning from a Frozen LLM] teaches: "The projected query embeddings are then prepended to the input text embeddings. [...] For decoder-based LLMs, we pre-train with the language modeling loss, where the frozen LLM is tasked to generate the text conditioned on the visual representation from Q-Former." LI [page 13, Figure 7] teaches the fully connected layer generating a projected query representation Z and prepending the query representation to the input text embedding for generating the answer by the LLM.) generate one or more first tokens based on the projected representation ; (LI [page 4, section 3.3. Bootstrap Vision-to-Language Generative Learning from a Frozen LLM] teaches: "The projected query embeddings are then prepended to the input text embeddings." LI [page 5, Figure 4] shows various examples such as the user submitting "Which city is this?", "Singapore", "Why do you think so?", "The city has a statue of a merlion." Examiner's note: LI [page 5, Figure 4] demonstrates that the LLM preserves previous user queries in the conversation to answer a question like "Why do you think so?" corresponding the a previous response regarding the city. Under BRI, one or more first tokens can be interpreted as the first token (i.e., word) generated by the LLM when responding to the input (e.g. “The” in the response "The city has a statue of a merlion."), as shown in the examples of LI's Figure 4.) […] the embedding representation […] the received input query […] (LI [page 1, Figure 1] teaches the Q-Former receiving queries and text for processing. LI [page 7, section 4.3. Visual Question Answering] teaches: "the question tokens are given as input to the Q-Former". LI [page 3, section 3.2. Image-Text Matching] teaches: "The output query embeddings Z thus capture multimodal information." LI [page 3, section 3.1. Model Architecture] teaches: "We use Z to denote the output query representation. The size of Z (32 × 768) is much smaller than the size of frozen image features (e.g. 257 × 1024 for ViT-L/14)." Examiner’s note: Under BRI, the embedding representation and the received input query can be interpreted as the output query representation Z, which are the outputs of the Q-Former based on the received queries and text.) generate a second token based on the projected representation of the received input query […] (LI [page 4, section 3.3. Bootstrap Vision-to-Language Generative Learning from a Frozen LLM] teaches: "The projected query embeddings are then prepended to the input text embeddings." LI [page 5, Figure 4] shows various examples such as the user submitting "Which city is this?", "Singapore", "Why do you think so?", "The city has a statue of a merlion." Examiner's note: LI [page 5, Figure 4] demonstrates that the LLM preserves previous user queries in the conversation to answer a question like "Why do you think so?" corresponding the a previous response regarding the city. Under BRI, a second token can be interpreted as the second token (i.e., word) generated by the LLM when responding to the input (e.g. “city” in the response "The city has a statue of a merlion."), as shown in the examples of LI's Figure 4.) CHEN 436’ further teaches: project a combination of the embedding representation […] and the one or more […] into a projected representation […] in the second dimensionality ; (CHEN 436’ [0092] teaches: "Additional projection aims to reduce the dimensionality of the encoder output representations to match the decoder stack dimension." CHEN 436’ [0128] teaches: "After concatenation, the combined representation is projected down to the decoder dimension (i.e., project a combination of the embedding representation […] into a representation […] in the second dimensionality ) with a layer-normalized affine transformation." CHEN 436’ [0117] teaches: "As illustrated in FIG. 5, a multi-column encoder merges the outputs (i.e., and the one or more ) of several independent encoders into a single combined representation.” CHEN 436’ [0111] teaches: "Decoding is inherently auto-regressive, and keeping track of the state information should therefore be intuitively beneficial for conditional generation.") LI in view of CHEN 436’ is not relied upon for teaching, but MENG teaches: […] the one or more first tokens […] (MENG [0025] teaches: "When performing a single NLP task, the decoder portion is iteratively queried with an input sequence that grows in length by one token each iteration. On the first iteration, the decoder input sequence includes the initial input to the model (e.g., the user input). The decoder generates a first word (i.e., the one or more first tokens ) of an output sequence." MENG [0026] teaches: "In this architecture, an encoder processes the input sequence to produce a continuous representation, or embedding, of the input before the embedding is provided to a decoder that iteratively generates the output sequence (e.g., one word at a time).") generate a second token based on the […] representation of received input query and the one or more first tokens. (MENG [0025] teaches: "When performing a single NLP task, the decoder portion is iteratively queried with an input sequence that grows in length by one token each iteration. On the first iteration, the decoder input sequence includes the initial input to the model (e.g., the user input). The decoder generates a first word of an output sequence. Then, on the second decoder iteration, the decoder input includes the initial input to the model followed by the first word of the output sequence. The decoder outputs a second word of the output sequence. Then, on the third iteration, the decoder input includes the initial input to the model followed by the first two words of the outputs sequence. The decoder outputs a third word of the output sequence. This continues until the output sequence is completed (e.g., the output sequence reaches a pre-defined cap set on the permissible output size or the trained sequential model 104 otherwise determines that an applicable stop logic criterion is satisfied)." MENG [0026] teaches: "In this architecture, an encoder processes the input sequence to produce a continuous representation, or embedding, of the input before the embedding is provided to a decoder that iteratively generates the output sequence (e.g., one word at a time)." MENG [FIG.1] and [FIG. 2] show the trained sequential model receiving combined inputs that include the generated tokens such as “Mary” “had” in FIG. 1, for example. Additionally, MENG [FIG. 1] teaches a second decoder iteration, where “a” “little” are the generated outputs. Furthermore, MENG [0026] teaches that input sequence containing the input query and the generated tokens from the previous decoder iteration can be embedded prior to being provided to the decoder for a subsequent decoder iteration until the max_token parameter is satisfied. Under BRI, generate a second token based on […] representation of the received input query and the one or more first tokens can be interpreted as the input request by the user combined with the tokens generated in the second decoder iteration using the input request (e.g., “Mary” “had”), as described in MENG [0025] and shown in MENG [FIG. 1].) Accordingly, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention, having the teachings of LI, CHEN 436', and MENG before them, to include MENG’s including the initial input of the user for LLM autoregressive generation based on a max token output size in LI and CHENG 436’s text generation method. MENG [0026] teaches that the input sequence containing the initial input query and the generated words from the previous decoder iteration can be embedded prior to being provided to the decoder for a subsequent decoder iteration. Additionally, since both MENG and CHEN 436’ disclose autoregressive generation, a person would have been motivated to project MENG’s embeddings of the initial user input and the generated words (i.e., tokens) by using CHEN 436’s combined embedding projection to reduce dimensionality to match the decoder stack dimension, as disclosed respectively in MENG [0026] and CHEN 436’ [0092]. One would have been motivated to make such a combination because since it is difficult to intelligently predict the exact data size of model output that will be needed to answer a given LLM input (text query), it is common practice for the specified output size parameter to serve as an upper bound on the permissible length of the model output (MENG [0015]). Regarding Claim 14: LI in view of CHEN 436’ teaches the elements of claim 11 as outlined above. Additionally, the claim recites similar limitations as corresponding claim 4 and is rejected for similar reasons as claim 4 using similar teachings and rationale. Regarding Claim 15: LI in view of CHEN 436’ and MENG teaches the elements of claim 14 as outlined above. Additionally, the claim recites similar limitations as corresponding claim 5 and is rejected for similar reasons as claim 5 using similar teachings and rationale. Regarding Claim 16: LI in view of CHEN 436’ teaches the elements of claim 11 as outlined above. Additionally, the claim recites similar limitations as corresponding claim 6 and is rejected for similar reasons as claim 6 using similar teachings and rationale. Regarding Claim 17: LI in view of CHEN 436’ and MENG teaches the elements of claim 16 as outlined above. Additionally, the claim recites similar limitations as corresponding claim 7 and is rejected for similar reasons as claim 7 using similar teachings and rationale. Regarding Claim 18: LI in view of CHEN 436’ teaches the elements of claim 11 as outlined above. Additionally, the claim recites similar limitations as corresponding claim 8 and is rejected for similar reasons as claim 8 using similar teachings and rationale. Regarding Claim 24: LI in view of CHEN 436’ teaches the elements of claim 21 as outlined above. Additionally, the claim recites similar limitations as corresponding claim 4 and is rejected for similar reasons as claim 4 using similar teachings and rationale. Regarding Claim 25: LI in view of CHEN 436’ and MENG teaches the elements of claim 24 as outlined above. Additionally, the claim recites similar limitations as corresponding claim 5 and is rejected for similar reasons as claim 5 using similar teachings and rationale. Regarding Claim 26: LI in view of CHEN 436’ teaches the elements of claim 21 as outlined above. Additionally, the claim recites similar limitations as corresponding claim 6 and is rejected for similar reasons as claim 6 using similar teachings and rationale. Regarding Claim 27: LI in view of CHEN 436’ and MENG teaches the elements of claim 26 as outlined above. Additionally, the claim recites similar limitations as corresponding claim 7 and is rejected for similar reasons as claim 7 using similar teachings and rationale. Regarding Claim 28: LI in view of CHEN 436’ teaches the elements of claim 21 as outlined above. Additionally, the claim recites similar limitations as corresponding claim 8 and is rejected for similar reasons as claim 8 using similar teachings and rationale . 07-21-aia AIA Claim s 9, 19, and 29 are rejected under 35 U.S.C. 103 as being unpatentable over LI in view of CHEN 436’ as applied respectively above to claims 1, 11, and 21, and further in view of NGUYEN (US 20250078484 A1) and CHEN (US 20240249077 A1), hereafter NGUYEN and CHEN 077’ respectively . Regarding Claim 9: LI in view of CHEN 436’ teaches the elements of claim 1 as outlined above. LI further teaches: the first generative artificial intelligence model […] (LI [page 3, section 3.1. Model Architecture] teaches: "We initialize Q-Former (i.e., first generative artificial intelligence model ) with the pre-trained weights of BERTbase [...].”) CHEN 436’ further teaches: […] input in the second dimensionality . (CHEN 436’ [0128] teaches: "After concatenation, the combined representation is projected down to the decoder dimension (i.e., input in the second dimensionality ) with a layer-normalized affine transformation." CHEN 436’ [0117] teaches: "As illustrated in FIG. 5, a multi-column encoder merges the outputs of several independent encoders into a single combined representation." CHEN 436’ [0092] teaches: "Additional projection aims to reduce the dimensionality of the encoder output representations to match the decoder stack dimension.") LI in view of CHEN 436’ is not relied upon for teaching: the […] generative artificial intelligence model comprises a large language model trained to generate the embedding representation in the first dimensionality, and the second generative artificial intelligence model comprises a small language model trained to generate a response based on an input in the […] dimensionality . However, NGUYEN teaches: the […] generative artificial intelligence model comprises a large language model trained to generate the embedding representation in the first dimensionality , (NGUYEN [0024] teaches: " In various implementations, the visual and audio embeddings may be assembled into a LLM prompt. The LLM prompt may then be processed, e.g., using an onboard LLM that may be scaled down relative to a remote/cloud-based full scale LLM, to generate a multimodal contextual embedding. For instance, the onboard LLM may include a subset of layers of the full scale LLM, such as the first n layers. This may enable the onboard LLM to generate the multimodal contextual embedding (i.e., a large language model trained to generate the embedding representation in the first dimensionality ) in a reasonable amount of time.”) Accordingly, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention, having the teachings of LI, CHEN 436’, and NGUYEN before them, to include NGUYEN’s LLM for generating multimodal embeddings in LI and CHEN 436’s text generation method. NGUYEN [FIG. 1] shows that the LLMs are able to receive multimodal inputs (e.g., audio, image, and video), and thus a person would be motivated to make such a combination and include more input types into the LLM in order to generate a semantically rich multimodal embedding that represents a user's context (NGUYEN [0003]). LI in view of CHEN 436’ and NGUYEN is not relied upon for teaching, but CHEN 077’ teaches: the second generative artificial intelligence model comprises a small language model trained to generate a response based on an input in the […] dimensionality . (CHEN 077’ [0005] teaches: "Therefore, there is a need for an in-context learning system using small-scale LMs." CHEN 077’ [0019] teaches: "generates, via the small LM (i.e., a small language model ), a task output based on an input concatenating the soft prompts as augmented context and the input query (i.e ., trained to generate a response based on an input in the […] dimensionality )." CHEN 077’ [0026] teaches: "For example, the augmented contexts may comprise a concatenation of the soft embeddings [...].") Accordingly, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention, having the teachings of LI, CHEN 436’, NGUYEN, and CHEN 077’ before them, to include CHEN 077’s small language model generating response using soft prompts (i.e., embedding representation of an input query) in LI, CHEN 436’, and NGUYEN’s text generation method. One would have been motivated to make such a combination because the use of smaller backbone LMs largely reduces computational complexity and thus improves system efficiency (CHEN 077’ [0022].) Regarding Claim 19: LI in view of CHEN 436’ teaches the elements of claim 11 as outlined above. Additionally, the claim recites similar limitations as corresponding claim 9 and is rejected for similar reasons as claim 9 using similar teachings and rationale. Regarding Claim 29: LI in view of CHEN 436’ teaches the elements of claim 21 as outlined above. Additionally, the claim recites similar limitations as corresponding claim 9 and is rejected for similar reasons as claim 9 using similar teachings and rationale . Conclusion 07-96 AIA The prior art made of record and not relied upon is considered pertinent to applicant's disclosure : LI (US 20240161520 A1) is the associated patent publication of LI ("BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models") Any inquiry concerning this communication or earlier communications from the examiner should be directed to Alvaro S Laham Bauzo whose telephone number is (571)272-5650. The examiner can normally be reached Mon-Fri 7:30 AM - 11:00 AM | 1:00 PM - 5:30 PM ET. Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice. If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Usmaan Saeed can be reached on (571) 272-4046. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300. Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000. /A.S.L./Examiner, Art Unit 2146 /USMAAN SAEED/Supervisory Patent Examiner, Art Unit 2146 Application/Control Number: 18/486,653 Page 2 Art Unit: 2146 Application/Control Number: 18/486,653 Page 3 Art Unit: 2146 Application/Control Number: 18/486,653 Page 4 Art Unit: 2146 Application/Control Number: 18/486,653 Page 5 Art Unit: 2146 Application/Control Number: 18/486,653 Page 6 Art Unit: 2146 Application/Control Number: 18/486,653 Page 7 Art Unit: 2146 Application/Control Number: 18/486,653 Page 8 Art Unit: 2146 Application/Control Number: 18/486,653 Page 9 Art Unit: 2146 Application/Control Number: 18/486,653 Page 10 Art Unit: 2146 Application/Control Number: 18/486,653 Page 11 Art Unit: 2146 Application/Control Number: 18/486,653 Page 12 Art Unit: 2146 Application/Control Number: 18/486,653 Page 13 Art Unit: 2146 Application/Control Number: 18/486,653 Page 14 Art Unit: 2146 Application/Control Number: 18/486,653 Page 15 Art Unit: 2146 Application/Control Number: 18/486,653 Page 16 Art Unit: 2146 Application/Control Number: 18/486,653 Page 17 Art Unit: 2146 Application/Control Number: 18/486,653 Page 18 Art Unit: 2146 Application/Control Number: 18/486,653 Page 19 Art Unit: 2146 Application/Control Number: 18/486,653 Page 20 Art Unit: 2146 Application/Control Number: 18/486,653 Page 21 Art Unit: 2146 Application/Control Number: 18/486,653 Page 22 Art Unit: 2146 Application/Control Number: 18/486,653 Page 23 Art Unit: 2146 Application/Control Number: 18/486,653 Page 24 Art Unit: 2146 Application/Control Number: 18/486,653 Page 25 Art Unit: 2146 Application/Control Number: 18/486,653 Page 26 Art Unit: 2146 Application/Control Number: 18/486,653 Page 27 Art Unit: 2146 Application/Control Number: 18/486,653 Page 28 Art Unit: 2146 Application/Control Number: 18/486,653 Page 29 Art Unit: 2146 Application/Control Number: 18/486,653 Page 30 Art Unit: 2146 Application/Control Number: 18/486,653 Page 31 Art Unit: 2146 Application/Control Number: 18/486,653 Page 32 Art Unit: 2146 Application/Control Number: 18/486,653 Page 33 Art Unit: 2146 Application/Control Number: 18/486,653 Page 34 Art Unit: 2146 Application/Control Number: 18/486,653 Page 35 Art Unit: 2146 Application/Control Number: 18/486,653 Page 36 Art Unit: 2146 Application/Control Number: 18/486,653 Page 37 Art Unit: 2146 Application/Control Number: 18/486,653 Page 38 Art Unit: 2146 Application/Control Number: 18/486,653 Page 39 Art Unit: 2146 Application/Control Number: 18/486,653 Page 40 Art Unit: 2146 Application/Control Number: 18/486,653 Page 41 Art Unit: 2146 Application/Control Number: 18/486,653 Page 42 Art Unit: 2146 Application/Control Number: 18/486,653 Page 43 Art Unit: 2146 Application/Control Number: 18/486,653 Page 44 Art Unit: 2146 Application/Control Number: 18/486,653 Page 45 Art Unit: 2146 Application/Control Number: 18/486,653 Page 46 Art Unit: 2146 Application/Control Number: 18/486,653 Page 47 Art Unit: 2146 Application/Control Number: 18/486,653 Page 48 Art Unit: 2146 Application/Control Number: 18/486,653 Page 49 Art Unit: 2146