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 .
DETAILED ACTION
Claims 1-20 are present in this application. Claims 1-20 are pending in this office
action.
This office action is NON-FINAL.
Drawings
The Drawings filed on 03/14/25 are acceptable for examination purposes.
Specification
The Specification filed on 03/14/25 is acceptable for examination purposes.
Information Disclosure Statement
The information disclosure statements (IDS) filed on 04/04/25 has been
considered by the Examiner and made of record in the application file.
Claim Rejections - 35 U.S.C. §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 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.
6. 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 of this
title, 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 1-20 are rejected under 35 U.S.C. 103 as being unpatentable
over SANTHANAM et al. (US 2025/0021761 A1) in view of Sussman et al. (US 2025/0077376 A1).
Regarding claim 1, SANTHANAM teaches a computer-implemented method for optimizing an augmentation of a prompt provided to a generative Artificial Intelligence, Al, model, (See SANTHANAM paragraph [0004], generating a response to a query (also referred to as a prompt) using generative artificial intelligence, See SANTHANAM paragraph [0039], augment a query for initial processing by the generative artificial intelligence model 124); wherein the prompt is an input sequence of input segments respectively comprising one or more input tokens, (See SANTHANAM paragraph [0005], a sequence of tokens corresponding to a candidate response to the input query; receiving, from a second generative artificial intelligence model, a response based on the generated sequence of tokens); and wherein the generative Al model is configured to generate, from the prompt, (See SANTHANAM paragraph [0004], generating a response to a query (also referred to as a prompt) using generative artificial intelligence), an output sequence of output segments respectively comprising one or more output tokens; the computer-implemented method comprising, (See SANTHANAM paragraph [0004], The output of each pass may be a probability distribution on a sequence of tokens (words or parts of words) from which the next token (word or part of word) may be selected):
obtaining at least one target output sequence for the prompt provided to the generative Al model, (See SANTHANAM Abstract, generated sequence of tokens are received from the second generative artificial intelligence model…output as a response to the received input query),
obtaining one or more augmented prompts, (See SANTHANAM paragraph [0040], receive requests to process input queries…to augment responses generated by the generative artificial intelligence model 124).
SANTHANAM does not explicitly disclose by adjusting one or more input segments with respect to at least one reference prompt, determining prompt importance scores for the respective output segments of the at least one target output sequence, wherein a prompt importance score of a respective output segment, is indicative for a change in probability of said output segment within the output sequence, generated by the generative Al model as a result of adjusting the reference prompt, and- optimizing the augmentation of the prompt based on the prompt importance scores of the respective output segments.
However Sussman teaches by adjusting one or more input segments with respect to at least one reference prompt, (See Sussman paragraph [0089], the system is adjusted until the difference is minimized. In backpropagation, the encoder is trained to extract the right information from the input sequence),
determining prompt importance scores for the respective output segments of the at least one target output sequence, (See Sussman paragraph [0069], The sentiment analysis LLM reviews the output and applies a sentiment score to the keywords or topics in the output); wherein a prompt importance score of a respective output segment, (See Sussman paragraph [0046], the apparatus calculates an F1 score for the identified categories in the LLM outputs), is indicative for a change in probability of said output segment within the output sequence, (See Sussman paragraph [0054], Once the change to the posterior probability of the keywords or topics in iterative outputs is less than a preset threshold, the iterative prompt to the AI to be tested ceases and a new prompt is sent to the AI to be tested 209), generated by the generative Al model as a result of adjusting the reference prompt, (See Sussman paragraph [0039], The apparatus then submits the prompts to the AI to be tested 105 and the resulting output is stored in the output cluster 106); and- optimizing the augmentation of the prompt based on the prompt importance scores of the respective output segments, (See Sussman paragraph [0105], When calculating self-attention scores for a given element, the dot products between the query vector of this element and the key vectors of all other input elements are calculated. To make the model mathematically more stable, these self-attention scores are divided by the root of the size of the vectors).
It would have been obvious to one with ordinary skill in the art before the
effective filing date of the claimed invention was made to modify by adjusting one or more input segments with respect to at least one reference prompt, determining prompt importance scores for the respective output segments of the at least one target output sequence, wherein a prompt importance score of a respective output segment, is indicative for a change in probability of said output segment within the output sequence, generated by the generative Al model as a result of adjusting the reference prompt, and- optimizing the augmentation of the prompt based on the prompt importance scores of the respective output segments of Sussman to accurately estimate the potential utility of an example on improving the performance, such that the model can be well trained with minimal queries.
Claim 15 recites the same limitations as claim 1 above. Therefore,
Claim 15 is rejected based on the same reasoning.
Regarding claim 2, SANTHANAM taught the computer-implemented method according to claim1, as described above. SANTHANAM further teaches wherein the at least one target output sequence, (See SANTHANAM paragraph [0073], the operations 500 proceed with outputting, to a second generative artificial intelligence model (e.g., target model) is a configuration instruction for configuring a network node or a controller, (See SANTHANAM paragraph [0064], the target model to verify a candidate response based on a scoring model, which may be implemented by a neural network or other machine learning model).
Claim 16 recites the same limitations as claim 2 above. Therefore,
Claim 13 is rejected based on the same reasoning.
Regarding claim 3, SANTHANAM taught the computer-implemented method according to claim1, as described above. SANTHANAM further teaches wherein the at least one target output sequence, (See SANTHANAM paragraph [0073], the operations 500 proceed with outputting, to a second generative artificial intelligence model (e.g., target model), is a formatted query for interacting with a queryable system, (See SANTHANAM paragraph [0037], an interface through which input queries and/or candidate responses generated by the generative artificial intelligence model 114).
Claim 17 recites the same limitations as claim 3 above. Therefore,
Claim 17 is rejected based on the same reasoning.
Regarding claim 4, SANTHANAM taught the computer-implemented method according to claim1, as described above. SANTHANAM further teaches wherein determining the prompt importance scores comprises, (See SANTHANAM paragraph [0064], determine whether to use the target model to verify a candidate response based on a scoring model), for each augmented prompt, (See SANTHANAM paragraph [0004], generate a response to a query (prompt) formatted as a text query (prompt)):
providing the augmented prompt and the at least one target output sequence to the generative Al model; (See SANTHANAM Abstract, generated sequence of tokens are received from the second generative artificial intelligence model…output as a response to the received input query), and
SANTHANAM does not explicitly disclose obtaining the probabilities for the respective output segments of the target output sequence, by extracting a measure of predicted likelihood associated with the respective output segments from the generative Al model.
However Sussman teaches obtaining the probabilities for the respective output segments of the target output sequence, (See Sussman paragraph [0054], Once the change to the posterior probability of the keywords or topics in iterative outputs is less than a preset threshold, the iterative prompt to the AI to be tested ceases and a new prompt is sent to the AI to be tested 209), by extracting a measure of predicted likelihood associated with the respective output segments from the generative Al model, (See Sussman paragraph [0054], The Posterior predictive check involves comparing a model's predictions against new samples to assess if the current model is sufficiently capturing the distribution. In this example, the apparatus uses the output of a selected model to predict new samples and compare these predictions to actual samples of output from the AI to be tested. If the predictions match closely, it suggests the model has captured the distribution well).
It would have been obvious to one with ordinary skill in the art before the
effective filing date of the claimed invention was made to modify obtaining the probabilities for the respective output segments of the target output sequence, by extracting a measure of predicted likelihood associated with the respective output segments from the generative Al model of Sussman to accurately estimate the potential utility of an example on improving the performance, such that the model can be well trained with minimal queries.
Claim 18 recites the same limitations as claim 4 above. Therefore,
Claim 18 is rejected based on the same reasoning.
Regarding claim 5, SANTHANAM taught the computer-implemented method according to claim 4, as described above. SANTHANAM further teaches when providing the augmented prompt to the generative Al model, (See SANTHANAM Abstract, generated sequence of tokens are received from the second generative artificial intelligence model…output as a response to the received input query), relative to the probability of said output segment when providing the reference prompt to the generative Al model, (See SANTHANAM paragraph [0030], generated by the draft model need not exactly match the probability distribution for a response generated by the target model. Additionally, inferencing using generative artificial intelligence models).
SANTHANAM does not explicitly disclose wherein the prompt importance score of a respective output segment, is determined as the complement of a ratio of the probability of said output segment.
However Sussman teaches wherein the prompt importance score of a respective output segment, (See Sussman paragraph [0105], When calculating self-attention scores for a given element, the dot products between the query vector of this element and the key vectors of all other input elements are calculated. To make the model mathematically more stable, these self-attention scores are divided by the root of the size of the vectors), is determined as the complement of a ratio of the probability of said output segment, (See Sussman paragraph [0054], Once the change to the posterior probability of the keywords or topics in iterative outputs is less than a preset threshold, the iterative prompt to the AI to be tested ceases and a new prompt is sent to the AI to be tested 209).
It would have been obvious to one with ordinary skill in the art before the
effective filing date of the claimed invention was made to modify wherein the prompt importance score of a respective output segment, is determined as the complement of a ratio of the probability of said output segment of Sussman to accurately estimate the potential utility of an example on improving the performance, such that the model can be well trained with minimal queries.
Claim 19 recites the same limitations as claim 5 above. Therefore,
Claim 19 is rejected based on the same reasoning.
Regarding claim 6, SANTHANAM taught the computer-implemented method according to claim 4, as described above. SANTHANAM further teaches when providing the reference prompt to the generative Al model, (See Conway paragraph [0009], the prompt construction facility, or the generative model included in the generative AI system), and the probability of said output segment when providing the augmented prompt to the generative Al model, (See SANTHANAM paragraph [0030], generated by the draft model need not exactly match the probability distribution for a response generated by the target model. Additionally, inferencing using generative artificial intelligence models).
SANTHANAM does not explicitly disclose wherein the prompt importance score of a respective output segment, is determined as an absolute difference between the probability of said output segment.
However Conway teaches wherein the prompt importance score of a respective output segment, (See Sussman paragraph [0046], The apparatus will continue to sample the output for topics until both metrics stabilize. In an example, the apparatus calculates an F1 score for the identified categories in the LLM outputs), is determined as an absolute difference between the probability of said output segment, (See Conway paragraph [0228], the output may be classified with a multidimensional vector. Each vector may be directed to a different aspect of responsiveness),
It would have been obvious to one with ordinary skill in the art before the
effective filing date of the claimed invention was made to modify wherein the prompt importance score of a respective output segment, is determined as an absolute difference between the probability of said output segment of Conway to accurately estimate the potential utility of an example on improving the performance, such that the model can be well trained with minimal queries.
Claim 20 recites the same limitations as claim 6 above. Therefore,
Claim 20 is rejected based on the same reasoning.
Regarding claim 7, SANTHANAM taught the computer-implemented method according to claim 4, as described above. SANTHANAM further teaches wherein the prompt importance score of a respective output segment, (See SANTHANAM paragraph [0064], The scoring model can generally account for the conditional token probabilities associated with each of the tokens included in the candidate response. If the conditional token probabilities), is determined as the relative probability of said output segment with respect to the highest probability of said output segment, (See SANTHANAM paragraph [0004], The output of each pass may be a probability distribution on a sequence of tokens (words or parts of words) from which the next token (word or part of word) may be selected, either by sampling or based on maximum likelihood).
Regarding claim 8, SANTHANAM taught the computer-implemented method according to claim 1, as described above.
SANTHANAM does not explicitly disclose adjusting one or more input segments with respect to a reference prompt, comprises omitting and/or reordering the one or more input segments of the reference prompt.
However Sussman teaches wherein adjusting one or more input segments with respect to a reference prompt, (See Sussman paragraph [0089], the system is adjusted until the difference is minimized. In backpropagation, the encoder is trained to extract the right information from the input sequence), comprises omitting and/or reordering the one or more input segments of the reference prompt, (See Sussman paragraph [0089], Input/output (I/O) training a neural network in which the initial system output is compared to the desired output and the system is adjusted until the difference is minimized. In backpropagation, the encoder is trained to extract the right information from the input sequence).
It would have been obvious to one with ordinary skill in the art before the
effective filing date of the claimed invention was made to modify adjusting one or more input segments with respect to a reference prompt, comprises omitting and/or reordering the one or more input segments of the reference prompt of Sussman to accurately estimate the potential utility of an example on improving the performance, such that the model can be well trained with minimal queries.
Regarding claim 9, SANTHANAM taught the computer-implemented method according to claim 1, as described above.
SANTHANAM does not explicitly disclose wherein adjusting one or more input segments with respect to a reference prompt, comprises sampling one or more input segments from a set of possible input segments; and adding or replacing the one or more input segments of the reference prompt with the one or more sampled input segments.
However Sussman teaches wherein adjusting one or more input segments with respect to a reference prompt, (See Sussman paragraph [0089], the system is adjusted until the difference is minimized. In backpropagation, the encoder is trained to extract the right information from the input sequence), comprises sampling one or more input segments from a set of possible input segments; (See Sussman paragraph [0038], The apparatus provides the response AI with input and output of the AI to be tested for each query to determine if the topics in the iterative output are similar enough to generate a sufficient sample size) and adding or replacing the one or more input segments of the reference prompt with the one or more sampled input segments, (See Sussman paragraph [0037], the user may interact with the apparatus by adding a prompt or topic or removing a prompt or similar for testing. The prompts may be a word, a group of words, or a sentence based on the concept. An AI can analyze the one or more concepts input by the user and suggest prompts 103).
It would have been obvious to one with ordinary skill in the art before the
effective filing date of the claimed invention was made to modify wherein adjusting one or more input segments with respect to a reference prompt, comprises sampling one or more input segments from a set of possible input segments; and adding or replacing the one or more input segments of the reference prompt with the one or more sampled input segments of Sussman to accurately estimate the potential utility of an example on improving the performance, such that the model can be well trained with minimal queries.
Regarding claim 10, SANTHANAM taught the computer-implemented method according to claim 1, as described above.
SANTHANAM does not explicitly disclose further comprising determining an effectiveness of input segments based on the prompt importance scores; wherein the effectiveness of an input segment is indicative for the number of input tokens that, are included within the input segment relative to the number of output tokens, affected by augmenting the input segment and the change in prompt importance score of these affected output tokens.
However Sussman teaches further comprising determining an effectiveness of input segments based on the prompt importance scores; (See Sussman paragraph [0105], the model mathematically more stable, these self-attention scores are divided by the root of the size of the vectors. This has the effect of reducing the importance of the scalar thus emphasizing the importance of the direction of the vector), wherein the effectiveness of an input segment is indicative for the number of input tokens that, (See Sussman paragraph [0132], The class token vector is extracted from the output of the last Transformer block and is passed into a multilayer perceptron (MLP) head whose output is the final classification. The perceptron takes the normalized input and places the output in categories), are included within the input segment relative to the number of output tokens, (See Sussman paragraph [0132], The class token vector is extracted from the output of the last Transformer block and is passed into a multilayer perceptron (MLP) head whose output is the final classification), affected by augmenting the input segment and the change in prompt importance score of these affected output tokens, (See Sussman paragraph [0132], A special class token vector is added to the sequence of embedding vectors to include all representative information of all tokens through the multi-layer encoding procedure…The class token vector is extracted from the output of the last Transformer block and is passed into a multilayer perceptron (MLP) head whose output is the final classification).
It would have been obvious to one with ordinary skill in the art before the
effective filing date of the claimed invention was made to modify further comprising determining an effectiveness of input segments based on the prompt importance scores; wherein the effectiveness of an input segment is indicative for the number of input tokens that, are included within the input segment relative to the number of output tokens, affected by augmenting the input segment and the change in prompt importance score of these affected output tokens of Sussman to accurately estimate the potential utility of an example on improving the performance, such that the model can be well trained with minimal queries.
Regarding claim 11, SANTHANAM taught the computer-implemented method according to claim 10, as described above.
SANTHANAM does not explicitly disclose further further comprising determining whether to perform optimizing the augmentation of the prompt based on the effectiveness of the respective input segments, in the prompt provided to the generative Al model.
However Sussman teaches further comprising determining whether to perform optimizing the augmentation of the prompt based on the effectiveness of the respective input segments, (See Sussman paragraph [0132], The first layer of a ViT extracts a fixed number of patches from an input image (FIG. 14A). The patches are then projected to linear embeddings. A special class token vector is added to the sequence of embedding vectors to include all representative information of all tokens through the multi-layer encoding procedure), in the prompt provided to the generative Al model, (See Sussman paragraph [0054], the iterative prompt to the AI to be tested ceases and a new prompt is sent to the AI to be tested 209).
It would have been obvious to one with ordinary skill in the art before the
effective filing date of the claimed invention was made to modify further comprising determining whether to perform optimizing the augmentation of the prompt based on the effectiveness of the respective input segments, in the prompt provided to the generative Al model of Sussman to accurately estimate the potential utility of an example on improving the performance, such that the model can be well trained with minimal queries.
Regarding claim 12, SANTHANAM taught the computer-implemented method according to claim 1, as described above. SANTHANAM further teaches wherein optimizing the augmentation of the prompt comprises, (See SANTHANAM paragraph [0004], generating a response to a query (also referred to as a prompt) using generative artificial intelligence, See SANTHANAM paragraph [0039], augment a query for initial processing by the generative artificial intelligence model 124).
SANTHANAM does not explicitly disclose at least one of improving the selecting of input segments from a set of possible input segments, improving the formatting of the input segments, improving the order of input segments in the input sequence of the prompt; tuning a model for generating an input segment; and/or initiating a model for generating an input segment.
However Sussman teaches at least one of improving the selecting of input segments from a set of possible input segments, improving the formatting of the input segments, (See Sussman paragraph [0013], iteratively selects the most useful examples from the unlabeled dataset to query their labels from the oracle. After adding the newly labeled data into the training set, the model can be updated to achieve better performance. The key task in active learning is how to accurately estimate the potential utility of an example on improving the performance), improving the order of input segments in the input sequence of the prompt; tuning a model for generating an input segment; and/or initiating a model for generating an input segment, (See Sussman paragraph [0013], After adding the newly labeled data into the training set, the model can be updated to achieve better performance. The key task in active learning is how to accurately estimate the potential utility of an example on improving the performance, such that the model can be well trained with minimal queries).
It would have been obvious to one with ordinary skill in the art before the
effective filing date of the claimed invention was made to modify at least one of improving the selecting of input segments from a set of possible input segments, improving the formatting of the input segments, improving the order of input segments in the input sequence of the prompt; tuning a model for generating an input segment; and/or initiating a model for generating an input segment of Sussman to accurately estimate the potential utility of an example on improving the performance, such that the model can be well trained with minimal queries.
Regarding claim 13, SANTHANAM taught the computer-implemented method according to claim 1, as described above. SANTHANAM further teaches wherein the at least one target output sequence is the output sequence generated by the generative Al model, (See SANTHANAM paragraph [0073], the operations 500 proceed with outputting, to a second generative artificial intelligence model (e.g., target model).
SANTHANAM does not explicitly disclose when provided with the reference prompt, or the at least one target output sequence is a desired output sequence. However Sussman teaches when provided with the reference prompt, or the at least one target output sequence is a desired output sequence, (See Sussman paragraph [0089], the system is adjusted until the difference is minimized. In backpropagation, the encoder is trained to extract the right information from the input sequence).
It would have been obvious to one with ordinary skill in the art before the
effective filing date of the claimed invention was made to modify when provided with the reference prompt, or the at least one target output sequence is a desired output sequence of Sussman to accurately estimate the potential utility of an example on improving the performance, such that the model can be well trained with minimal queries.
Regarding claim 14, SANTHANAM taught the computer-implemented method according to claim 1, as described above. SANTHANAM further teaches wherein the reference prompt is a user provided prompt, an empty prompt, and/or a complete prompt, (See SANTHANAM paragraph [0102], generally configured to operate on complete models. Such NPUs may thus be configured to input a new piece of data and rapidly process this new piece through an already trained model to generate a model output (e.g., an inference), comprising an ordered sequence of all input segments in a set of possible input segments wherefrom a prompt can be constructed, (See SANTHANAM paragraph [0004], The output of each pass may be a probability distribution on a sequence of tokens (words or parts of words) from which the next token (word or part of word) may be selected…to generate each token in a response to a query (prompt).
Conclusions/Points of Contacts
The prior art made of record and not relied upon is considered pertinent to
applicant’s disclosure. See form PTO-892.
WAGNER et al. (US 2021/0365633 A1), a system is used to perform a number of different token packing techniques. For example, the system can receive a set of input data that includes a sequence (e.g., a set of sentences) of tokens (e.g., words). The system may group the sequence of tokens into several groups of tokens (e.g., groups of words that form a sentence).
QIN (US 2024/0346256 A1) computer program products are disclosed for using retrieval augmented artificial intelligence to generate a response to a query. A first feature vector is generated based at least on the query. The first feature vector is compared to a plurality of second feature vector.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to MULUEMEBET GURMU whose telephone number is (571)270-7095. The examiner can normally be reached M-F 9am - 5pm.
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, Tony Mahmoudi can be reached at 5712724078. 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.
/MULUEMEBET GURMU/Primary Examiner, Art Unit 2163
Prosecution Insights