DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Claims 1 - 20 are pending and claims 1, 15 and 20 are independent claims.
Information Disclosure Statement
The information disclosure statements (IDS) submitted on 03/02/2025 and 03/02/2025 is in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the examiner.
Claim Rejections - 35 USC § 101
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 1-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more. The independent claims 1, 15 and 20 recite ”receiving …processing …receiving… transforming …generating …appending … passing” as drafted cover an abstract idea of data analysis/retrieval and mental steps. More specifically, the “receiving the input information; processing the input information using the machine-trained model, the machine-trained model including a series of processing blocks, a particular processing block of the series of processing blocks performing operations of: receiving a set of input tokens; transforming the set of input tokens into a set of output tokens; generating a control decision based on the output tokens, the control decision having at least two selectable states, including a first state and a second state; for the first state, appending at least one of the output tokens to the input tokens to define a new set of input tokens, and repeating the transforming and generating; and for the second state, passing at least some of the set of output tokens to a next processing block for processing” which requires just data analysis / retrieval step and mental process. For instance, one can receive input information and process the input information through paper and pen/pencil. Then one can receive a set of input tokens and transform the set of input tokens into a set of output tokens. One can generate a control decision based on the output tokens, and the control decision can have two or more selectable states, including a first state and a second state; for the first state, a person may be appending at least one of the output tokens to the input tokens to define a new set of input tokens, and repeating the transforming and generating steps; and similarly, for the second state, a person can mentally pass at least some of the set of output tokens to a next processing block, or using a paper and pen/pencil. The claims do not provide a noble inventive concept, and the claim doesn’t integrate a judicial exception into a practical application by adding elements that are not well-understood, routine and conventional, rather the claims introduce a generalized machine-trained model which is just an additional element. The claimed invention is, therefore, directed to an abstract idea and a mental process without significantly more and thus, claims 1, 15 and 20 are rejected under 35 U.S.C. 101.
Similarly, the dependent claims 2-14 and 16-19 recite similar claim language as in claims 1, 15 and 20.
Claim 2 recites “performs a first number of total transforming operations using a first amount of resources for a first query having a first complexity, wherein the method performs a second number of total transforming operations using a second amount of resources for a second query having a second complexity, the first complexity level being less than the second complexity level, the first number of total transforming operations being less than the second number of total transforming operations, and the first amount of resources being less than the second amount of resources,” which requires just a mathematical step of transforming operations using an amount of resources for a query. One can apply a mathematical step or using pen and paper perform a first number of total transforming operations using a first amount of resources for a first query having a first complexity, or this might just be implemented in using organizational data and infrastructure to automatically create, optimize, and execute database or search queries which can be implemented using a generalized or any conventional computer or lookup table. Thus, claim 2 is directed to an abstract idea.
Claim 3 recites “the machine-trained model is a transformer language model, and wherein the plural processing blocks are plural transformer processing blocks,” which just introduces the additional element, since the “machine-trained model” or “transformer language model’ is just an additional element that does not provide integration of a judicial exception into a practical application nor the claim provides an inventive concept by adding elements that are not well-understood. Thus, claim 3 is directed to an abstract idea.
Claim 4 recites “the set of input tokens that is fed to the particular processing block includes first input tokens that originate from the input information and a second input token that serves as an input control token, wherein the set of output tokens produced by the transforming includes first output tokens that are transformed counterparts of the first input tokens and a second output token that is a transformed counterpart of the second input token, and wherein the second output token expresses information that is supplemental to information expressed by the first output tokens,” which requires just a mental step or a step that can be implemented using pen and paper of feeding input tokens to the particular processing block. For instance, the set of input tokens that is fed to the particular processing block represents input information having any type or combination of content types (Spec. 0032). For example, each of the set of input tokens represents a word or part of a word in a query, input sentence, document, or other text-based item. The transformation predicts a next output token in a sequence of tokens based on the modified set of input tokens. Each pass generates supplemental information regarding the set of input tokens, which, in turn, facilitates the interpretation of the set of input tokens in subsequent processing blocks. Such steps can be performed using pen and paper. Thus, claim 4 is directed to an abstract idea.
Claims 5 and 17 which recite “assessing complexity and generating a control decision are based on the second output token,” which also requires just a mental step of assessing complexity and generating a control decision that are based on the second output token. A person can mentally assess complexity and generate a control decision as well. Thus, claims 5 and 17 are directed to an abstract idea.
Claim 6 recites “for the first state, said at least one of the output tokens that is appended to the input set of tokens includes the second output token, whereupon the second output token assumes a role as a next input control token in a next iteration of the transforming and generating,” which requires just a mental step or a step performed using pen/paper of appending the output tokens to the input set of tokens where the output token can assume a role as a next input control token in a next iteration of the transforming and generating. For instance, the first tokens are transformed counterparts of linguistic input tokens produced by a tokenizer, in some cases, based on a submitted query. The second tokens are control tokens that serve two purposes. First, in some contexts, the control tokens convey information on the basis of whether to repeat a transformation operation. Second, in other contexts, the control tokens express supplemental information. Such a step of appending the output tokens to the input set of tokens for next iteration of transforming and generating can be done using pen and paper. Thus, claim 6 is directed to an abstract idea.
Claim 7 recites “for the second state, said at least some of the set of output tokens that are passed to the next processing block include one or more instances of the second output token that have been generated in one or more respective iterations of the transforming and generating,” which requires just a mental step or a step performed using pen/paper of passing tokens that to the next processing block that include one or more instances of the second output token that have been generated in one or more respective iterations of the transforming and generating. Thus, claim 7 is directed to an abstract idea.
Claim 8 recites “the one or more instances of the second output token excludes an instance of the second output token on which a control decision to cease iterating the transforming and generating is based,” which requires just a mental step of excluding the output token , for instance, on which a control decision may stop iterating the transformation and generation process. Thus, claim 8 is directed to an abstract idea.
Claim 9 recites “the second output token embodies information that is inexpressible in linguistic tokens,” which requires just a mental step or a step performed using pen/paper. For instance, representing the second output token information as an information that is inexpressible in linguistic tokens but can be expressed using pen and pencil. Thus, claim 9 is directed to an abstract idea.
Claim 10 recites “the second output token expresses information that lies outside a vector space of concepts that are directly expressible using linguistic tokens,” which requires just a mental step or a step performed using pen/paper. for instance, expressing token information that lies outside a specified vector space of concepts when those concepts or terms or words are totally different, for instance, instead of being synonyms if those terms or words are antonyms, that are directly expressible using linguistic tokens and they may lie in different vector spaces. Thus, claim 10 is directed to an abstract idea.
Claim 11 recites “the input control token that is provided to the particular processing block has a value that is produced in a machine-training process,” which requires just a mental step or a step performed using pen/paper. For instance, an input control token that is provided to the particular processing block can have a value produced by a pen and paper. Thus, claim 11 is directed to an abstract idea.
Claims 12 and 19 recite “the machine-trained model includes parameters that are trained using a loss function that rewards agreement between model-generated responses and ground-truth responses, and penalizes repetition of transformation operations,” which requires just a mathematical procedure performed using pen/paper … The loss function or the objective function can be used for calculating the rewards.. The “claims do not provide a noble inventive concept, and the claim doesn’t integrate a judicial exception into a practical application by adding elements that are not well-understood, routine and conventional, rather the claims introduce a generalized machine-trained model which is just an additional element. Thus, claims 12 and 19 are directed to an abstract idea.
Claim 13 recites “the generating a control decision is also based on an assessed resource capability of a computing device that executes the machine-trained model,” which requires just a mental step or a step performed using pen/paper for generating or making a control decision. This control decision can be based on the assessed resource capability of a computing device. The “claims do not provide a noble inventive concept, and the claim doesn’t integrate a judicial exception into a practical application by adding elements that are not well-understood, routine and conventional, rather the claim introduces a generalized “machine-trained model” which is just an additional element. Thus, claim 13 is directed to an abstract idea.
Claim 14 recites “predicting, in a last processing block in the series of processing blocks, a linguistic token, and wherein the method further includes appending the predicted linguistic token to a prior sequence of linguistic tokens that were processed by the machine-trained model, together with any control tokens generated in a prior pass through the machine-trained model,” which requires just a step that can be performed using pen/paper. For instance, using a pen and paper one can predict a linguistic token by appending the predicted linguistic token to a prior sequence of linguistic tokens. . The “machine-trained model” is just an additional element. Thus, claim 14 is directed to an abstract idea.
Claim 16 recites “the set of input tokens that is fed to the particular processing block includes first input tokens that originate from the input information and a second input token that serves as an input control token, wherein the set of output tokens produced by the transforming includes first output tokens that are transformed counterparts of the first input tokens and a second output token that is a transformed counterpart of the second input token, and wherein the second output expresses information that lies outside a vector space of concepts that are directly expressible using linguistic tokens,” which requires just a mental step or a step that can be performed using pen/paper. For instance, the input tokens can use a specific algorithm and provide output semantically similar in space k as the first input tokens originating from the input information and this can be done on a pen and paper. Thus, claim 7 is directed to an abstract idea.
Claim 18 recites “for the first state, said at least one of the output tokens that is appended to the input set of tokens includes the second output token, whereupon the second output token assumes a role as a next input control token in a next iteration of the transforming and generating, and wherein, for the second state, said at least some of the set of output tokens that are passed to the next processing block include one or more instances of the second output token that have been generated in one or more respective iterations of the transforming and generating,” which requires just a mental step or a step that can be performed using pen/paper for appending the input tokens and associated steps in the claims with a pen and paper. . Thus, claim 7 is directed to an abstract idea.
Thus, claims 1-20 as drafted cover a mental process and abstract idea of data gathering/retrieval and analysis/processing steps, and they are mental processes directed to an abstract idea of implementing mathematical formulae for data processing and data analysis using a conventional/generic (general-purpose) computer as well and thus, all the claims are directed to an abstract idea.
This judicial exception is not integrated into a practical application. In particular, claims 1 and 12 recite additional element of “processor” and “memory” as per the independent claims. Accordingly, these additional elements do not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea. The claim is directed to an abstract idea.
Thus, taken alone, the additional elements do not amount to significantly more than the above-identified judicial exception (the abstract idea). Looking at the limitations as an ordered combination adds nothing that is not already present when looking at the elements taken individually. There is no indication that the combination of elements improves the functioning of a computer or improves any other technology. Their collective functions merely provide conventional general purpose computer implementation (Spec., para 0134). Claims 1-20, are therefore not drawn to patent eligible subject matter as they are directed to an abstract idea without significantly more. Thus, the claimed invention is directed to an abstract idea and a mental process without significantly more and thus, claims 1-20 are rejected under 35 U.S.C. 101.
The claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to the integration of the abstract idea into a practical application, the additional element of using a computer is noted as a general computer as noted. Mere instructions to apply an exception using a generic computer component cannot provide an inventive concept (Spec., para 0134). Further, the additional limitation in the claims noted above are directed towards insignificant solution activity. The claims are not patent eligible.
Dependent claims 2-14 and 16-19 are also directed toward an abstract idea and do not include additional elements that are sufficient to amount to significantly more than the judicial exception because the additional elements when considered both individually and as an ordered combination do not amount to significantly more than the abstract idea. Therefore, claims 1-20 do not contain patent eligible subject matter that has been identified by the courts.
Claim Rejections - 35 USC § 102
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
Claims 1, 3, 12-13, 19 and 20 are rejected under 35 U.S.C. 102(a)(1) as being anticipated by Memisevic et al Pat App No US 20240126987 A1 (Memisevic).
Regarding Claim 1, Memisevic discloses a method for processing input information using a machine-trained model (Memisevic, para 0006, processor-implemented method includes receiving an input comprising a previous language stream. The method also includes generating an output language stream by a pre-trained language mode; [“input comprising a previous language stream” as “input information”; “pre-trained language mode” as “machine-trained model”]), comprising:
receiving the input information (Memisevic, para 0030, receive an input comprising a previous language stream);
processing the input information using the machine-trained model (Memisevic, para 0006, processor-implemented method includes receiving an input comprising a previous language stream. The method also includes generating an output language stream by a pre-trained language mode; [“input comprising a previous language stream” as “input information”; “pre-trained language mode” as “machine-trained model”]), the machine-trained model including a series of processing blocks (Memisevic, para 0039, The example transformer architecture 200 may include alternating layers of attention blocks 206 and MLP blocks 204), a particular processing block of the series of processing blocks performing operations of:
receiving a set of input tokens (Memisevic, para 0040, The example transformer architecture 200 may receive an input such as sequential data 210);
transforming the set of input tokens into a set of output tokens (Memisevic, para 0040, the sequential data 210 may include (but is not limited to) a sequence of characters or textual data (e.g., a sentence) or audio data, for instance. The transformer architecture 200 may divide the input 210 into portions or tokens (e.g., words));
generating a control decision based on the output tokens, the control decision having at least two selectable states, including a first state and a second state (Memisevic, para 0047-0049, To enable the pre-training of the language model for performing decision making tasks, a tokenizer and vocabulary may be determined to express actions, states, and returns as “language.” … the decision-making task may be solely absorbed within the language modeling objective rather than solving the decision-making task using a specialized architecture. Therefore, unlike conventional approaches, the ANN model architecture disclosed may not have any task-specific heads, any task-specific tokens, or word embeddings);
for the first state, appending at least one of the output tokens to the input tokens to define a new set of input tokens (Memisevic, para 0050-0053, The environment may continuously monitor and parse the ANN model output to detect an occurrence of well-formed actions within the ANN model output stream. For example, a well-formed action may be detected if a regular expression matches an action output. When a well-formed action is detected, the environment (or an agent within the environment) may perform an operation, such as a state transition, and return a result, such as a new state and/or reward. The result may be written back into the output stream generated by the ANN model. The process may be repeated with the updated input; [“the ANN model output stream… the updated input” as “appending … output tokens to the input tokens to define a new set of input tokens”]… When a well-formed action is detected, the environment may generate a state and a reward, which may be encoded and appended to the ANN model output stream according to the position of the pointer. That is, the pointer may be moved to the position immediately after the matched pattern; ALTERNATIVE, Memisevic, para 0063, At block 310, the processor may append the result to the output language stream to obtain an updated output language stream. For instance, as described, the state and reward generated by the environment may be encoded and appended to the ANN model output stream according to the position of the pointer), and repeating the transforming and generating (Memisevic, para 0065, A processor-implemented method comprising: receiving an input comprising a previous language stream; generating an output language stream by a pre-trained language model, based on the input; detecting a well-formed action based on patterns in the output language stream; performing an operation, by an environment, in response to detecting the well-formed action, the operation returning a result; appending the result to the output language stream to obtain an updated output language stream; and repeating the generating, with the updated output language stream as the input); and
for the second state, passing at least some of the set of output tokens to a next processing block for processing (Memisevic, para 0046-0047, At inference time, an agent may be prompted with a start state s and desired return r and may generate an action a (followed by further states, returns, and actions). In an example, the agent may be a robot operating in an environment comprising a labyrinth. In the example, the objective may be to find a shortest path to navigate to a target location (e.g., a center or an exit) of the labyrinth. With each action (e.g., a move) taken by the robot through the labyrinth, the state (e.g., location of the robot) may be observed and a reward may be determined (e.g., +1 if closer to the target, or −1 if farther from the target, or each move may result in a reward of −1). To enable the pre-training of the language model for performing decision making tasks, a tokenizer and vocabulary may be determined to express actions, states, and returns as “language”; [i.e., “start state” as “first state”; “further states” as “second state” …]; Memisevic, para 0029-0032, The SOC 100 may also include additional processing blocks tailored to specific functions, such as a GPU 104, a DSP 106, a connectivity block 110, which may include fifth generation (5G) connectivity, fourth generation long term evolution (4G LTE) connectivity, Wi-Fi connectivity, USB connectivity, Bluetooth connectivity, and the like, and a multimedia processor 112 that may, for example, detect and recognize gestures…The general-purpose processor 102 may also include code to append the result to the output language stream to obtain an updated output language stream…with the updated output language stream as the input…Deep learning architectures may perform an object recognition task by learning to represent inputs at successively higher levels of abstraction in each layer, thereby building up a useful feature representation of the input data…A deep learning architecture may learn a hierarchy of features…The second layer, taking the output of the first layer as input, may learn to recognize combinations of features, such as simple shapes for visual data or combinations of sounds for auditory data. For instance, higher layers may learn to represent complex shapes in visual data or words in auditory data; [The “deep learning architecture” using data (including words/tokens) from previous layer/block outputs to get another output in the next layer]).
Regarding Claim 3, Memisevic discloses the method of claim 1, wherein the machine-trained model is a transformer language model (Memisevic, para 0043, aspects of the present disclosure are directed to an artificial neural network (ANN) model that combines language modeling and decision making within a single model. The ANN model may comprise (but is not limited to) a decision transformer model that treats the model as a pure language model, for example.), and wherein the plural processing blocks are plural transformer processing blocks (Memisevic, para 0039 - 0043, FIG. 2 is a block diagram illustrating an example transformer architecture 200. The example transformer architecture 200 may include alternating layers of attention blocks 206 and MLP blocks 204. A layer norm block 208 (e.g., 208a, 208b) may be applied before every block of the transformer architecture 200. The attention blocks 206 and MLP blocks 204 may each include a residual connection… The ANN model may be pre-trained and/or regularized for natural language processing (NLP) tasks, while fine-tuning the ANN model on the task-specific reinforcement learning (RL) objective).
Regarding Claim 12, Memisevic discloses the method of claim 1, wherein the machine-trained model includes parameters that are trained using a loss function that rewards agreement between model-generated responses and ground-truth responses, and penalizes repetition of transformation operations (Memisevic, para 0020, Reinforcement learning is a machine learning technique that enables an agent to learn in an interactive environment using feedback from the agent actions. A reward signal may encourage or discourage certain actions or behavior. One goal in reinforcement learning is to determine actions that increase, or possibly maximize, a total cumulative reward; Memisevic, para 0056, In some aspects, the training data for the ANN model may also be pre-processed such that initial rewards are replaced by higher reward values. The pre-processing may train the ANN model to interpret the initial reward information as a reward value to approximate, rather than as a reward value to perfectly replicate. For example, if an ideal reward is a ten on a scale of one to ten, the pre-processing may convert a reward value of five to a reward value that falls between five and ten. The new value may be deterministically calculated, for example. The pre-processing may enable configuring the ANN model at inference time without knowing the achievable reward ahead of time by starting the language generation with a high reward value.; [“new (reward) values… deterministically calculated” as “model-generated responses”; “ideal values” or “initial reward information as a reward value to approximate” as “ground-truth responses”]; para 0043, The ANN model may be pre-trained and/or regularized for natural language processing (NLP) tasks, while fine-tuning the ANN model on the task-specific reinforcement learning (RL) objective; [i.e., “reinforcement learning (RL) objective” (i.e., objective function) as “trained using a loss function”, OR “objective function” as “loss function”]).
Regarding Claim 13, Memisevic discloses the method of claim 1, wherein the generating a control decision is also based on an assessed resource capability of a computing device that executes the machine-trained model (Memisevic, para 0070, query optimizer embodiments may employ a learning-based approach to train cost models (e.g., resource consumption models as described above) from past workloads and feed them back to optimize future queries, thereby providing an integrated workload-driven approach to query optimization. Embodiments described herein are capable of harnessing massive workloads visible in modern cloud data services to continuously learn models that accurately capture the query runtime behavior).
Regarding Claim 19, Memisevic in view of Jackson discloses the computing system of claim 14, wherein the machine-trained model includes parameters that are trained using a loss function that rewards agreement between model-generated responses and ground-truth responses, and penalizes repetition of transformation operations (Memisevic, para 0020, Reinforcement learning is a machine learning technique that enables an agent to learn in an interactive environment using feedback from the agent actions. A reward signal may encourage or discourage certain actions or behavior. One goal in reinforcement learning is to determine actions that increase, or possibly maximize, a total cumulative reward; Memisevic, para 0056, In some aspects, the training data for the ANN model may also be pre-processed such that initial rewards are replaced by higher reward values. The pre-processing may train the ANN model to interpret the initial reward information as a reward value to approximate, rather than as a reward value to perfectly replicate. For example, if an ideal reward is a ten on a scale of one to ten, the pre-processing may convert a reward value of five to a reward value that falls between five and ten. The new value may be deterministically calculated, for example. The pre-processing may enable configuring the ANN model at inference time without knowing the achievable reward ahead of time by starting the language generation with a high reward value.; [“new (reward) values… deterministically calculated” as “model-generated responses”; “ideal values” or “initial reward information as a reward value to approximate” as “ground-truth responses”]).
Regarding Claim 20, Memisevic discloses a computer-readable storage medium for storing computer-readable instructions, a processing system executing the computer-readable instructions to perform operations (Memisevic, para 0009, computer-readable medium with program code recorded thereon is disclosed. The program code is executed by a processor and includes program code to receive an input), the operations comprising:
receiving input information (Memisevic, para 0030, receive an input comprising a previous language stream);
processing the input information using a machine-trained model including a series of processing blocks, a particular processing block of the series of processing blocks performing operations of (Memisevic, para 0039, The example transformer architecture 200 may include alternating layers of attention blocks 206 and MLP blocks 204):
receiving a set of input tokens, the set of input tokens including first input tokens that originate from the input information and a second input token that serves as an input control token(Memisevic, para 0040, The example transformer architecture 200 may receive an input such as sequential data 210);
transforming the set of input tokens into a set of output tokens, wherein the set of output tokens produced by the transforming includes first output tokens that are transformed counterparts of the first input tokens and a second output token that is a transformed counterpart of the second input token, the second output token conveying supplemental information to the first output tokens (Memisevic, para 0040, the sequential data 210 may include (but is not limited to) a sequence of characters or textual data (e.g., a sentence) or audio data, for instance. The transformer architecture 200 may divide the input 210 into portions or tokens (e.g., words));
generating a control decision based on the second output token, the control decision having at least two selectable states, including a first state and a second state (Memisevic, para 0047-0049, To enable the pre-training of the language model for performing decision making tasks, a tokenizer and vocabulary may be determined to express actions, states, and returns as “language.” … the decision-making task may be solely absorbed within the language modeling objective rather than solving the decision-making task using a specialized architecture. Therefore, unlike conventional approaches, the ANN model architecture disclosed may not have any task-specific heads, any task-specific tokens, or word embeddings);
for the first state, appending the second output token to the input tokens to define a new set of input tokens, and repeating the transforming and generating (Memisevic, para 0050-0053, The environment may continuously monitor and parse the ANN model output to detect an occurrence of well-formed actions within the ANN model output stream. For example, a well-formed action may be detected if a regular expression matches an action output. When a well-formed action is detected, the environment (or an agent within the environment) may perform an operation, such as a state transition, and return a result, such as a new state and/or reward. The result may be written back into the output stream generated by the ANN model. The process may be repeated with the updated input; [“the ANN model output stream… the updated input” as “appending … output tokens to the input tokens to define a new set of input tokens”]… When a well-formed action is detected, the environment may generate a state and a reward, which may be encoded and appended to the ANN model output stream according to the position of the pointer. That is, the pointer may be moved to the position immediately after the matched pattern; ALTERNATIVE, Memisevic, para 0063, At block 310, the processor may append the result to the output language stream to obtain an updated output language stream. For instance, as described, the state and reward generated by the environment may be encoded and appended to the ANN model output stream according to the position of the pointer), and repeating the transforming and generating (Memisevic, para 0065, A processor-implemented method comprising: receiving an input comprising a previous language stream; generating an output language stream by a pre-trained language model, based on the input; detecting a well-formed action based on patterns in the output language stream; performing an operation, by an environment, in response to detecting the well-formed action, the operation returning a result; appending the result to the output language stream to obtain an updated output language stream; and repeating the generating, with the updated output language stream as the input); and
for a second state, passing one or more instances of the second output token produced over one or more iterations of the transforming to a next processing block for processing (Memisevic, para 0046-0047, At inference time, an agent may be prompted with a start state s and desired return r and may generate an action a (followed by further states, returns, and actions). In an example, the agent may be a robot operating in an environment comprising a labyrinth. In the example, the objective may be to find a shortest path to navigate to a target location (e.g., a center or an exit) of the labyrinth. With each action (e.g., a move) taken by the robot through the labyrinth, the state (e.g., location of the robot) may be observed and a reward may be determined (e.g., +1 if closer to the target, or −1 if farther from the target, or each move may result in a reward of −1). To enable the pre-training of the language model for performing decision making tasks, a tokenizer and vocabulary may be determined to express actions, states, and returns as “language”; [i.e., “start state” as “first state”; “further states” as “second state” …]; Memisevic, para 0029-0032, The SOC 100 may also include additional processing blocks tailored to specific functions, such as a GPU 104, a DSP 106, a connectivity block 110, which may include fifth generation (5G) connectivity, fourth generation long term evolution (4G LTE) connectivity, Wi-Fi connectivity, USB connectivity, Bluetooth connectivity, and the like, and a multimedia processor 112 that may, for example, detect and recognize gestures…The general-purpose processor 102 may also include code to append the result to the output language stream to obtain an updated output language stream…with the updated output language stream as the input…Deep learning architectures may perform an object recognition task by learning to represent inputs at successively higher levels of abstraction in each layer, thereby building up a useful feature representation of the input data…A deep learning architecture may learn a hierarchy of features…The second layer, taking the output of the first layer as input, may learn to recognize combinations of features, such as simple shapes for visual data or combinations of sounds for auditory data. For instance, higher layers may learn to represent complex shapes in visual data or words in auditory data; [The “deep learning architecture” using data (including words/tokens) from previous layer/block outputs to get another output in the next layer]).
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
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.
Claims 4-7, 9-11 and 14-18 are rejected under 35 U.S.C. 103 as being unpatentable Memisevic in view of Jackson et al. Pat App No US 20260147998 A1 (Jackson).
Regarding Claim 4, Memisevic discloses the method of claim 1.
Memisevic does not specifically disclose wherein the set of input tokens that is fed to the particular processing block includes first input tokens that originate from the input information and a second input token that serves as an input control token, wherein the set of output tokens produced by the transforming includes first output tokens that are transformed counterparts of the first input tokens and a second output token that is a transformed counterpart of the second input token, and wherein the second output token expresses information that is supplemental to information expressed by the first output tokens.
However, Jackson, in the same field of endeavor, discloses:
wherein the set of input tokens that is fed to the particular processing block includes first input tokens that originate from the input information and a second input token that serves as an input control token (Jackson, para 0040-0046, input tokens, in accordance with one or more embodiments. In particular, FIG. 2 shows an example of using a component that takes an input token and generates n probable next tokens using a text generation model. For example, text 202 may include the word “hurricane” into user interface 200. This word may comprise multiple input tokens (e.g., the input token “hurr” has been detected)…The control circuitry may comprise any suitable processing, storage, and/or input/output circuitry. Each of these devices may also include a user input interface and/or user output interface (e.g., a display) for use in receiving and displaying data… It should be noted that in some embodiments, the devices may have neither user input interfaces nor displays, and may instead receive and display content using another device (e.g., a dedicated display device such as a computer screen, and/or a dedicated input device such as a remote control, mouse, voice input, etc.).; [i.e., For instance “voice input” as “input control token”]),
wherein the set of output tokens produced by the transforming includes first output tokens that are transformed counterparts of the first input tokens and a second output token that is a transformed counterpart of the second input token (Jackson, para 0029, The input, typically a token or sequence of tokens, is transformed into a numerical representation (embedding) that captures its features. This embedding is then passed through several layers of the classification model, where the system analyzes the input based on its learned patterns from training data. At the output layer of the model, the system produces a set of raw scores, or logits, for each potential classification. These scores represent the model's initial assessment of how likely the input belongs to each class, but they are not yet interpretable as probabilities. To convert these logits into a probability distribution, the system applies a softmax function. The softmax function normalizes the scores so that they sum to 1, turning them into probabilities. Each probability represents the likelihood (e.g., expressed as a percentage) that the input corresponds to a specific classification based on the model's analysis), and
wherein the second output token expresses information that is supplemental to information expressed by the first output tokens (Jackson, para 0094 - 0102, determining a final classification for the first input token based on the first aggregated probability; and generating for display, on a user interface, a first output based on the final classification… determining a supplemental token based on the conversation history; and processing the supplemental token with the first input token in the first classification engine).
Therefore, it would have been obvious for one having ordinary skill in the art before the effective filing date of the claimed invention to incorporate the method of Jackson in the method of Memisevic because this would enable chatbots to be integrated into websites, messaging platforms, or voice-enabled systems, and make them versatile tools for automating conversations and improving user experience, such as for example, with an input to the chatbot, if AI-driven, where more advanced models, like large language models (LLMs), generate dynamic and context-aware responses (Jackson, para 0001-0002).
Regarding Claim 5, Memisevic in view of Jackson discloses the method of claim 4.
Memisevic further teaches:
wherein the generating a control decision is based on the second output token (Memisevic, para 0058, decision making based on results obtained from a pre-trained language model);
Jackson further teaches (ALTERNATIVE to the above Memisevic citation):
herein the generating a control decision is based on the second output token (Jackson, para 0038-0039, the system improve decision-making and handle situations where multiple classifications represent similar or equivalent meanings. The system may determine a final classification for the first input token based on the first aggregated probability. The system may generate for display, on a user interface, a first output based on the final classification. For example, the output may be a response to text entered by a user).
Regarding Claim 6, Memisevic in view of Jackson discloses the method of claim 4.
Memisevic further teaches:
wherein, for the first state, said at least one of the output tokens that is appended to the input set of tokens includes the second output token, whereupon the second output token assumes a role as a next input control token in a next iteration of the transforming and generating (Memisevic, para 0065-0066, receiving an input comprising a previous language stream; generating an output language stream by a pre-trained language model, based on the input; detecting a well-formed action based on patterns in the output language stream; performing an operation, by an environment, in response to detecting the well-formed action, the operation returning a result; appending the result to the output language stream to obtain an updated output language stream; and repeating the generating, with the updated output language stream as the input, the detecting, the performing, and the appending until a termination condition is satisfied. Aspect 2: The processor-implemented method of Aspect 1, in which the operation is a state transition in the environment, and the result includes a new state and/or a reward).
Regarding Claim 7, Memisevic in view of Jackson discloses the method of claim 4.
Jackson further teaches:
wherein, for the second state, said at least some of the set of output tokens that are passed to the next processing block include one or more instances of the second output token that have been generated in one or more respective iterations of the transforming and generating (Jackson, para 0088-0089, the system to generate new responses based on the combined understanding of the first and second tokens, or to modify existing responses that were generated based solely on the first token. This approach enables the system to handle continuous input and adjust its predictions or responses in real-time as more tokens are received. Whether processing single tokens or multi-token sequences, the system iteratively updates the probability arrays and refines its classification and response generation, ensuring that each new token contributes meaningfully to the overall interpretation of the input. This method allows the system to adapt dynamically to the flow of conversation or text input, generating more accurate and coherent responses).
Regarding Claim 9, Memisevic in view of Jackson discloses the method of claim 4.
Jackson further teaches:
wherein the second output token embodies information that is inexpressible in linguistic tokens (Jackson, para 0072-0075, The first expression represents these raw probabilities as real numbers, but they are not yet in a form that is easy to compare or work with directly, especially when dealing with very small values. To improve the handling of these probabilities, the system applies a logarithm function to the first expression. This process converts each probability into its logarithmic form, creating what is referred to as the second expression for the respective probabilities of the classifications… the system processes the first input token through the first classification engine to generate the first probability array by determining a feature input for the first input token and inputting the feature input into a first artificial intelligence model, wherein the first artificial intelligence model is trained to generate vector representations for the first input token. For example, the feature input may be a numerical or vector representation that captures various characteristics of the token, such as its linguistic or contextual properties, including its position in a sentence, the surrounding words, and any semantic meaning derived from the text).
Regarding Claim 10, Memisevic in view of Jackson discloses the method of claim 4.
Jackson further teaches:
wherein the second output token expresses information that lies outside a vector space of concepts that are directly expressible using linguistic tokens (Jackson, para 0017-0027, The system may generate one or more tokens based on text input 102. For example, the system may receive text comprising a token. A token used by a large language model (LLM), for instance, may be a fundamental unit of text that the model processes. Tokens can represent different linguistic elements, such as words, subwords, or even individual characters, depending on the tokenization scheme used. A single word may be represented by one token if it is common, or by multiple subword tokens if it is rare or complex… The vectors represent tokens in a way that tokens with similar meanings or roles in a sentence are placed closer together in the vector space, while those with different meanings are positioned farther apart. The neural network may learn these embeddings through training on large datasets, where it captures the relationships between words and how they are used in various contexts. For example, words with similar meanings, such as “dog” and “canine,” are placed close to each other in this vector space, allowing the classification engine to detect synonyms; [i.e., “those (tokens) with different meanings are positioned farther apart” as “(tokens that) lie outside a (specified) vector space”]).
Regarding Claim 11, Memisevic in view of Jackson discloses the method of claim 4.
Memisevic further teaches:
wherein the input control token that is provided to the particular processing block has a value that is produced in a machine-training process (para 0056, the training data for the ANN model may also be pre-processed such that initial rewards are replaced by higher reward values; [i.e., ‘ “higher reward values”… produced by the training data for the ANN model’ as “value that is produced in a machine-training process”]).
Regarding Claim 14, Memisevic discloses the method of claim 1.
Memisevic does not specifically disclose wherein the method further includes predicting, in a last processing block in the series of processing blocks, a linguistic token, and wherein the method further includes appending the predicted linguistic token to a prior sequence of linguistic tokens that were processed by the machine-trained model, together with any control tokens generated in a prior pass through the machine-trained model.
However, Jackson, in the same field of endeavor, discloses:
wherein the method further includes predicting, in a last processing block in the series of processing blocks, a linguistic token (Jackson, para 0017, the system may receive text comprising a token. A token used by a large language model (LLM), for instance, may be a fundamental unit of text that the model processes. Tokens can represent different linguistic elements, such as words, subwords, or even individual characters, depending on the tokenization scheme used. A single word may be represented by one token if it is common, or by multiple subword tokens if it is rare or complex. This allows the model to handle a broad range of language inputs, including uncommon words and misspellings. Tokens may serve as an input to the language model… The model treats tokens as the building blocks of language, and by combining them, it can comprehend the relationships between words and phrases, predict the next word in a sequence, answer questions, or perform translations. In essence, tokens allow the model to break down complex text into pieces it can work with, enabling it to interpret and generate language effectively), and
wherein the method further includes appending the predicted linguistic token to a prior sequence of linguistic tokens that were processed by the machine-trained model, together with any control tokens generated in a prior pass through the machine-trained model (Jackson, para 0088-0089, the system to generate new responses based on the combined understanding of the first and second tokens, or to modify existing responses that were generated based solely on the first token. This approach enables the system to handle continuous input and adjust its predictions or responses in real-time as more tokens are received. Whether processing single tokens or multi-token sequences, the system iteratively updates the probability arrays and refines its classification and response generation, ensuring that each new token contributes meaningfully to the overall interpretation of the input. This method allows the system to adapt dynamically to the flow of conversation or text input, generating more accurate and coherent responses).
Therefore, it would have been obvious for one having ordinary skill in the art before the effective filing date of the claimed invention to incorporate the method of Jackson in the method of Memisevic because this would enable chatbots to be integrated into websites, messaging platforms, or voice-enabled systems, and make them versatile tools for automating conversations and improving user experience, such as for example, with an input to the chatbot, if AI-driven, where more advanced models, like large language models (LLMs), generate dynamic and context-aware responses (Jackson, para 0001-0002).
Regarding Claim 15, Memisevic discloses a computing system for processing input information in a machine-trained model (Memisevic, para 0040, The example transformer architecture 200 may receive an input such as sequential data 210. In some examples, the sequential data 210 may include (but is not limited to) a sequence of characters or textual data (e.g., a sentence) or audio data, for instance. The transformer architecture 200 may divide the input 210 into portions or tokens (e.g., words)), comprising:
a data store for storing computer-readable instructions (Memisevic, para 0009, computer-readable medium with program code recorded thereon and
a processing system for executing the computer-readable instructions in the instruction data store, to perform operations including (Memisevic, para 0028, Instructions executed at the CPU 102 may be loaded from a program memory associated with the CPU 102 or may be loaded from a memory block 118):
receiving the input information (Memisevic, para 0030, receive an input comprising a previous language stream);
processing the input information using the machine-trained model (Memisevic, para 0006, processor-implemented method includes receiving an input comprising a previous language stream. The method also includes generating an output language stream by a pre-trained language mode; [“input comprising a previous language stream” as “input information”; “pre-trained language mode” as “machine-trained model”]), the machine-trained model including a series of processing blocks, a particular processing block of the series of processing blocks performing operations of (Memisevic, para 0039, The example transformer architecture 200 may include alternating layers of attention blocks 206 and MLP blocks 204):
receiving a set of input tokens (Memisevic, para 0040, The example transformer architecture 200 may receive an input such as sequential data 210);
transforming the set of input tokens into a set of output tokens (Memisevic, para 0040, the sequential data 210 may include (but is not limited to) a sequence of characters or textual data (e.g., a sentence) or audio data, for instance. The transformer architecture 200 may divide the input 210 into portions or tokens (e.g., words));
generating a control decision based on the complexity that is assessed, the control decision having at least two selectable states, including a first state and a second state (Memisevic, para 0047-0049, To enable the pre-training of the language model for performing decision making tasks, a tokenizer and vocabulary may be determined to express actions, states, and returns as “language.” … the decision-making task may be solely absorbed within the language modeling objective rather than solving the decision-making task using a specialized architecture. Therefore, unlike conventional approaches, the ANN model architecture disclosed may not have any task-specific heads, any task-specific tokens, or word embeddings);
for the first state, appending at least one of the output tokens to the input tokens to define a new set of input tokens, and repeating the transforming and generating (Memisevic, para 0050-0053, The environment may continuously monitor and parse the ANN model output to detect an occurrence of well-formed actions within the ANN model output stream. For example, a well-formed action may be detected if a regular expression matches an action output. When a well-formed action is detected, the environment (or an agent within the environment) may perform an operation, such as a state transition, and return a result, such as a new state and/or reward. The result may be written back into the output stream generated by the ANN model. The process may be repeated with the updated input; [“the ANN model output stream… the updated input” as “appending … output tokens to the input tokens to define a new set of input tokens”]… When a well-formed action is detected, the environment may generate a state and a reward, which may be encoded and appended to the ANN model output stream according to the position of the pointer. That is, the pointer may be moved to the position immediately after the matched pattern; ALTERNATIVE, Memisevic, para 0063, At block 310, the processor may append the result to the output language stream to obtain an updated output language stream. For instance, as described, the state and reward generated by the environment may be encoded and appended to the ANN model output stream according to the position of the pointer), and repeating the transforming and generating (Memisevic, para 0065, A processor-implemented method comprising: receiving an input comprising a previous language stream; generating an output language stream by a pre-trained language model, based on the input; detecting a well-formed action based on patterns in the output language stream; performing an operation, by an environment, in response to detecting the well-formed action, the operation returning a result; appending the result to the output language stream to obtain an updated output language stream; and repeating the generating, with the updated output language stream as the input); and
for the second state, passing at least some of the set of output tokens to a next processing block for processing (Memisevic, para 0046-0047, At inference time, an agent may be prompted with a start state s and desired return r and may generate an action a (followed by further states, returns, and actions). In an example, the agent may be a robot operating in an environment comprising a labyrinth. In the example, the objective may be to find a shortest path to navigate to a target location (e.g., a center or an exit) of the labyrinth. With each action (e.g., a move) taken by the robot through the labyrinth, the state (e.g., location of the robot) may be observed and a reward may be determined (e.g., +1 if closer to the target, or −1 if farther from the target, or each move may result in a reward of −1). To enable the pre-training of the language model for performing decision making tasks, a tokenizer and vocabulary may be determined to express actions, states, and returns as “language”; [i.e., “start state” as “first state”; “further states” as “second state” …]; Memisevic, para 0029-0032, The SOC 100 may also include additional processing blocks tailored to specific functions, such as a GPU 104, a DSP 106, a connectivity block 110, which may include fifth generation (5G) connectivity, fourth generation long term evolution (4G LTE) connectivity, Wi-Fi connectivity, USB connectivity, Bluetooth connectivity, and the like, and a multimedia processor 112 that may, for example, detect and recognize gestures…The general-purpose processor 102 may also include code to append the result to the output language stream to obtain an updated output language stream…with the updated output language stream as the input…Deep learning architectures may perform an object recognition task by learning to represent inputs at successively higher levels of abstraction in each layer, thereby building up a useful feature representation of the input data…A deep learning architecture may learn a hierarchy of features…The second layer, taking the output of the first layer as input, may learn to recognize combinations of features, such as simple shapes for visual data or combinations of sounds for auditory data. For instance, higher layers may learn to represent complex shapes in visual data or words in auditory data; [The “deep learning architecture” using data (including words/tokens) from previous layer/block outputs to get another output in the next layer]).
Memisevic does not specifically disclose assessing complexity of information expressed in the set of input tokens.
However, Jackson, in the same field of endeavor, discloses assessing complexity of information expressed in the set of input tokens (Jackson, para 0069, The system evaluates the input's content, structure, and context, breaking it down into tokens and generating vector representations that capture its meaning);
Therefore, it would have been obvious for one having ordinary skill in the art before the effective filing date of the claimed invention to incorporate the method of Jackson in the method of Memisevic because this would enable chatbots to be integrated into websites, messaging platforms, or voice-enabled systems, and make them versatile tools for automating conversations and improving user experience, such as for example, with an input to the chatbot, if AI-driven, where more advanced models, like large language models (LLMs), generate dynamic and context-aware responses (Jackson, para 0001-0002).
Regarding Claim 16, Memisevic in view of Jackson disclose the computing system of claim 15.
Memisevic does not specifically disclose wherein the set of input tokens that is fed to the particular processing block includes first input tokens that originate from the input information and a second input token that serves as an input control token, wherein the set of output tokens produced by the transforming includes first output tokens that are transformed counterparts of the first input tokens and a second output token that is a transformed counterpart of the second input token, and wherein the second output expresses information that lies outside a vector space of concepts that are directly expressible using linguistic tokens.
Jackson further discloses:
wherein the set of input tokens that is fed to the particular processing block includes first input tokens that originate from the input information and a second input token that serves as an input control token (Jackson, para 0040-0046, input tokens, in accordance with one or more embodiments. In particular, FIG. 2 shows an example of using a component that takes an input token and generates n probable next tokens using a text generation model. For example, text 202 may include the word “hurricane” into user interface 200. This word may comprise multiple input tokens (e.g., the input token “hurr” has been detected)…The control circuitry may comprise any suitable processing, storage, and/or input/output circuitry. Each of these devices may also include a user input interface and/or user output interface (e.g., a display) for use in receiving and displaying data… It should be noted that in some embodiments, the devices may have neither user input interfaces nor displays, and may instead receive and display content using another device (e.g., a dedicated display device such as a computer screen, and/or a dedicated input device such as a remote control, mouse, voice input, etc.).; [i.e., For instance “voice input” as “input control token”]),
wherein the set of output tokens produced by the transforming includes first output tokens that are transformed counterparts of the first input tokens and a second output token that is a transformed counterpart of the second input token (Jackson, para 0029, The input, typically a token or sequence of tokens, is transformed into a numerical representation (embedding) that captures its features. This embedding is then passed through several layers of the classification model, where the system analyzes the input based on its learned patterns from training data. At the output layer of the model, the system produces a set of raw scores, or logits, for each potential classification. These scores represent the model's initial assessment of how likely the input belongs to each class, but they are not yet interpretable as probabilities. To convert these logits into a probability distribution, the system applies a softmax function. The softmax function normalizes the scores so that they sum to 1, turning them into probabilities. Each probability represents the likelihood (e.g., expressed as a percentage) that the input corresponds to a specific classification based on the model's analysis), and
wherein the second output expresses information that lies outside a vector space of concepts that are directly expressible using linguistic tokens (Jackson, para 0017-0027, The system may generate one or more tokens based on text input 102. For example, the system may receive text comprising a token. A token used by a large language model (LLM), for instance, may be a fundamental unit of text that the model processes. Tokens can represent different linguistic elements, such as words, subwords, or even individual characters, depending on the tokenization scheme used. A single word may be represented by one token if it is common, or by multiple subword tokens if it is rare or complex… The vectors represent tokens in a way that tokens with similar meanings or roles in a sentence are placed closer together in the vector space, while those with different meanings are positioned farther apart. The neural network may learn these embeddings through training on large datasets, where it captures the relationships between words and how they are used in various contexts. For example, words with similar meanings, such as “dog” and “canine,” are placed close to each other in this vector space, allowing the classification engine to detect synonyms; [i.e., “those (tokens) with different meanings are positioned farther apart” as “(tokens that) lie outside a (specified) vector space”]).
Regarding Claim 17, Memisevic in view of Jackson discloses the computing system of claim 16.
Memisevic further teaches:
wherein the assessing complexity and generating a control decision are based on the second output token (Memisevic, para 0058, decision making based on results obtained from a pre-trained language model).
Jackson further teaches (ALTERNATIVE to the above Memisevic citation):
wherein the generating a control decision is based on the second output token (Jackson, para 0038-0039, the system improve decision-making and handle situations where multiple classifications represent similar or equivalent meanings. The system may determine a final classification for the first input token based on the first aggregated probability. The system may generate for display, on a user interface, a first output based on the final classification. For example, the output may be a response to text entered by a user).
Regarding Claim 18, Memisevic in view of Jackson discloses the computing system of claim 16.
Memisevic further teaches:
wherein, for the first state, said at least one of the output tokens that is appended to the input set of tokens includes the second output token, whereupon the second output token assumes a role as a next input control token in a next iteration of the transforming and generating (Memisevic, para 0065-0066, receiving an input comprising a previous language stream; generating an output language stream by a pre-trained language model, based on the input; detecting a well-formed action based on patterns in the output language stream; performing an operation, by an environment, in response to detecting the well-formed action, the operation returning a result; appending the result to the output language stream to obtain an updated output language stream; and repeating the generating, with the updated output language stream as the input, the detecting, the performing, and the appending until a termination condition is satisfied. Aspect 2: The processor-implemented method of Aspect 1, in which the operation is a state transition in the environment, and the result includes a new state and/or a reward), and
Jackson further teaches:
wherein, for the second state, said at least some of the set of output tokens that are passed to the next processing block include one or more instances of the second output token that have been generated in one or more respective iterations of the transforming and generating (Jackson, para 0088-0089, the system to generate new responses based on the combined understanding of the first and second tokens, or to modify existing responses that were generated based solely on the first token. This approach enables the system to handle continuous input and adjust its predictions or responses in real-time as more tokens are received. Whether processing single tokens or multi-token sequences, the system iteratively updates the probability arrays and refines its classification and response generation, ensuring that each new token contributes meaningfully to the overall interpretation of the input. This method allows the system to adapt dynamically to the flow of conversation or text input, generating more accurate and coherent responses).
Claim 8 is rejected under 35 U.S.C. 103 as being unpatentable Memisevic in view of Jackson, and further in view of Hartmann et al. Pat App No. US 20260220137 A1 (Hartmann).
Regarding Claim 8, Memisevic in view of Jackson discloses the method of claim 7.
Memisevic in view of Jackson do not specifically disclose wherein the one or more instances of the second output token excludes an instance of the second output token on which a control decision to cease iterating the transforming and generating is based.
However, Hartmann, in the same field of endeavor, discloses wherein the one or more instances of the second output token excludes an instance of the second output token on which a control decision to cease iterating the transforming and generating is based (Hartmann, para 0057-0061, the output at each iteration or update cycle can be or can include the plurality (e.g., sequence) of tokens. Examples of a sequence processing model include, but are not limited to, a large language model (LLM) or a large multimodal model (LMM). For instance, the sequence processing model may generate tokens in a sequential manner. The sequence processing model may not provide a fixed-length block of tokens but rather generate each token as evaluation of the model progresses. The EOS token may signal the model to cease generation of tokens…fundamental tokens can be associated with factual information that is directly responsive to a particular query. For example, tokens corresponding to the words “George Washington” are highly useful, if not strictly necessary, to provide an accurate answer to the query “Who was the first president of the United States?” Therefore, some example embodiments can include identifying certain fundamental tokens and excluding those tokens from the penalization techniques described herein, thereby enabling the model to create diverse, yet factually grounded or otherwise accurate responses).
Therefore, it would have been obvious for one having ordinary skill in the art before the effective filing date of the claimed invention to incorporate the method of Hartmann in the method of Memisevic in view of Jackson because this would enable the model's output to have fewer tokens than a maximum length, for instance, and iterations of the sequence processing model (e.g., update cycles) can refer to multiple “passes” over the sequence of tokens, where tokens in the sequence from one update cycle and/or their respective rendering values are provided as input in a subsequent update cycle to improve the model's understanding about the entire sequence of tokens at the subsequent update cycle (Hartmann, para 0057).
Allowable Subject Matter
Claim 2 is objected to as being dependent upon rejected base claim but would be allowable if rewritten in independent form including all of the limitations of the base claim and any intervening claims, and also if all these claims overcome the 101 rejections. The reasons for allowance are that the prior art of record do not specifically teach the limitations as recited in the claims mentioned, esp. the prior arts do not teach “the first complexity level being less than the second complexity level, the first number of total transforming operations being less than the second number of total transforming operations, and the first amount of resources being less than the second amount of resources.”
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to MULUGETA T. DUGDA whose telephone number is (703)756-1106. The examiner can normally be reached Mon - Fri, 4:30am - 7:00pm.
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, Paras D. Shah can be reached at 571-270-1650. 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.
/MULUGETA TUJI DUGDA/Examiner, Art Unit 2653
/Paras D Shah/Supervisory Patent Examiner, Art Unit 2653
09/05/2026
Prosecution Insights