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 .
Response to Arguments
Applicant’s arguments with respect to the rejection(s) of claim(s) 1, 3-5, 7-9 and 11-12 under 35 U.S.C. 102(a)(2) have been fully considered and are persuasive. Therefore, the rejection has been withdrawn. However, upon further consideration, a new ground(s) of rejection is made in view of Doggett et al.
Applicant argues that Worthington does not disclose the newly added limitation of “retrieving one or more entries of an output of the at least one LLM agent from an LLM agent memory based on a cosine similarity between a current internal state of the at least one LLM agent and the LLM agent memory to include in the next planning step”, as recited in independent claims 1, 5, and 9. While this is true, Doggett et al. disclose a method/system/medium that retrieves entries of an output of an LLM agent from an LLM agent memory based on a cosine similarity between a current internal state of the LLM agent and the LLM agent memory. Additionally, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to use a cosine similarity between a current internal state of the LLM agent and the LLM agent memory for the reasons provided in the rejections below.
Applicant’s amendments necessitated the new grounds of rejection. Accordingly, this action is FINAL.
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(d):
(d) REFERENCE IN DEPENDENT FORMS.—Subject to subsection (e), a claim in dependent form shall contain a reference to a claim previously set forth and then specify a further limitation of the subject matter claimed. A claim in dependent form shall be construed to incorporate by reference all the limitations of the claim to which it refers.
The following is a quotation of pre-AIA 35 U.S.C. 112, fourth paragraph:
Subject to the following paragraph [i.e., the fifth paragraph of pre-AIA 35 U.S.C. 112], a claim in dependent form shall contain a reference to a claim previously set forth and then specify a further limitation of the subject matter claimed. A claim in dependent form shall be construed to incorporate by reference all the limitations of the claim to which it refers.
Claims 13-15 are rejected under 35 U.S.C. 112(d) or pre-AIA 35 U.S.C. 112, 4th paragraph, as being of improper dependent form for failing to further limit the subject matter of the claim upon which it depends, or for failing to include all the limitations of the claim upon which it depends.
Claims 13-15 depend on independent claims 1, 5, and 9, respectively. Each of claims 1, 5 and 9 recite “retrieving one or more entries of an output of the at least one LLM agent from an LLM agent memory based on a cosine similarity between a current internal state of the at least one LLM agent and the LLM agent memory to include in the next planning step”. Thus, “the similarity” as recited in claims 13-15 is already established as being “a cosine similarity between the current internal state of the at least one LLM agent and the LLM agent memory”. Claims 13-15 therefore fail to further limit claims 1, 5, and 9, respectively.
Applicant may cancel the claim(s), amend the claim(s) to place the claim(s) in proper dependent form, rewrite the claim(s) in independent form, or present a sufficient showing that the dependent claim(s) complies with the statutory requirements.
Claim Rejections - 35 USC § 103
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.
Claim(s) 1, 3-5, 7-9 and 11-15 is/are rejected under 35 U.S.C. 103 as being unpatentable over Worthington (U.S. Patent Application Pub. No. 2025/0148558), in view of Doggett et al. (U.S. Patent Application Pub. No. 2024/0394965, hereinafter “Doggett”).
In regard to claim 1, Worthington discloses a method for determining a prediction of a population level response to presented information (Fig. 1, 100), comprising:
(a) conditioning at least one large language model (LLM) agent on a plurality of population or group features using in-weight training or in-context tokens (a plurality of juror-agents are defined using demographics data, social science survey data, social characteristics, political preferences, etc. through prompting of an LLM, paragraph [0015]);
(b) recording an initial memory state of the at least one LLM agent (each juror-agent is initialized with a memory 106, paragraph [0015]);
(c) retrieving one or more entries of an output of the at least one LLM agent from an
LLM agent memory to include in the next planning step (juror memories of current and past opinions are utilized, paragraph [0019]);
(d) planning a response of the at least one LLM agent to an environment for the presented information (rules governing deliberations between the synthetic juror-agents are established, paragraph [0019]);
(e) transmitting one or more conditioned intra-agent communications to a plurality of additional LLM agents (deliberations in the form of iterative interaction and communication among the synthetic juror agents in a synthetic deliberative discussion environment, paragraph [0019]);
(f) receiving the one or more conditioned intra-agent communications from the plurality of additional LLM agents;
(g) recording an updated memory state of the LLM agent based on the one or more sent and received conditioned intra-agent communications (juror-agent memories are updating according to the rules in response to the deliberations, paragraphs [0019-0020]); and
(h) generating the prediction based on the updated memory state (a decision prediction is generated based on the deliberation and repeated iterative updating of the synthetic jury model state and individual juror-agent states, paragraphs [0020-0021]).
Worthington does not expressly disclose the retrieving one or more entries of an output of the at least one LLM agent from an LLM agent memory is based on a cosine similarity between a current internal state of the at least one LLM agent and the LLM agent memory.
Doggett discloses a method for retrieving one or more entries of an output of at least one LLM agent from an LLM agent memory, wherein the retrieving is based on a cosine similarity between a current internal state of the at least one LLM agent and the LLM agent memory (a query creation module a query using an LLM by prompting the LLM with a chat history representing the most recent messages generated by a virtual character, paragraphs [0033-0037]; the query is then compared to virtual character memories using a cosine similarity to determine matching memories to generate a response, paragraphs [0052-0053] and [0063]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to retrieve one or more entries of an output of the at least one LLM agent from an LLM agent memory one or more entries of an output of the at least one LLM agent from an LLM agent memory, because it would allow the responses to be grounded to the agent’s memories, thereby avoiding hallucinations and/or other inconsistencies in output from the LLMs, as taught by Doggett (paragraph [0008]).
In regard to claim 3, Worthington discloses the plurality of additional LLM agents are defined by the environment for the information (the rules of deliberation define the interactions allowed between the juror-agents, paragraph [0019]).
In regard to claim 4, Worthington discloses iterating procedures (a)-(h) for one or more additional time points (the simulations are repeated for multiple iterations, paragraphs [0029-0030]).
In regard to claim 5, Worthington discloses a system for determining a prediction of a population level response to presented information (Fig. 3, 300), comprising:
at least one processor (processing unit 302) configured to:
(a) condition at least one large language model (LLM) agent on a plurality of population or group features using in-weight training or in-context tokens (a plurality of juror-agents are defined using demographics data, social science survey data, social characteristics, political preferences, etc. through prompting of an LLM, paragraph [0015]);
(b) record an initial memory state of the at least one LLM agent (each juror-agent is initialized with a memory 106, paragraph [0015]);
(c) retrieve one or more entries of an output of the at least one LLM agent from an
LLM agent memory to include in the next planning step (juror memories of current and past opinions are utilized, paragraph [0019]);
(d) plan a response of the at least one LLM agent to an environment for the presented information (rules governing deliberations between the synthetic juror-agents are established, paragraph [0019]);
(e) transmit one or more conditioned intra-agent communications to a plurality of additional LLM agents (deliberations in the form of iterative interaction and communication among the synthetic juror agents in a synthetic deliberative discussion environment, paragraph [0019]);
(f) receive the one or more conditioned intra-agent communications from the plurality of additional LLM agents;
(g) record an updated memory state of the LLM agent based on the one or more sent and received conditioned intra-agent communications (juror-agent memories are updating according to the rules in response to the deliberations, paragraphs [0019-0020]); and
(h) generate the prediction based on the updated memory state (a decision prediction is generated based on the deliberation and repeated iterative updating of the synthetic jury model state and individual juror-agent states, paragraphs [0020-0021]).
Worthington does not expressly disclose the retrieving one or more entries of an output of the at least one LLM agent from an LLM agent memory is based on a cosine similarity between a current internal state of the at least one LLM agent and the LLM agent memory.
Doggett discloses a method for retrieving one or more entries of an output of at least one LLM agent from an LLM agent memory, wherein the retrieving is based on a cosine similarity between a current internal state of the at least one LLM agent and the LLM agent memory (a query creation module a query using an LLM by prompting the LLM with a chat history representing the most recent messages generated by a virtual character, paragraphs [0033-0037]; the query is then compared to virtual character memories using a cosine similarity to determine matching memories to generate a response, paragraphs [0052-0053] and [0063]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to retrieve one or more entries of an output of the at least one LLM agent from an LLM agent memory one or more entries of an output of the at least one LLM agent from an LLM agent memory, because it would allow the responses to be grounded to the agent’s memories, thereby avoiding hallucinations and/or other inconsistencies in output from the LLMs, as taught by Doggett (paragraph [0008]).
In regard to claim 7, Worthington discloses the plurality of additional LLM agents are defined by the environment for the information (the rules of deliberation define the interactions allowed between the juror-agents, paragraph [0019]).
In regard to claim 8, Worthington discloses iterating procedures (a)-(h) for one or more additional time points (the simulations are repeated for multiple iterations, paragraphs [0029-0030]).
In regard to claim 9, Worthington discloses a computer readable medium which includes software thereon for determining a prediction of a population level response to presented information (Fig. 3, system memory 304 and/or removable storage 308 and/or non-removable storage 310), wherein, when at least one computer processor executes the software (processing unit 302), the computer processor is configured to perform the procedures, comprising:
(a) conditioning at least one large language model (LLM) agent on a plurality of population or group features using in-weight training or in-context tokens (a plurality of juror-agents are defined using demographics data, social science survey data, social characteristics, political preferences, etc. through prompting of an LLM, paragraph [0015]);
(b) recording an initial memory state of the at least one LLM agent (each juror-agent is initialized with a memory 106, paragraph [0015]);
(c) retrieving one or more entries of an output of the at least one LLM agent from an
LLM agent memory to include in the next planning step (juror memories of current and past opinions are utilized, paragraph [0019]);
(d) planning a response of the at least one LLM agent to an environment for the presented information (rules governing deliberations between the synthetic juror-agents are established, paragraph [0019]);
(e) transmitting one or more conditioned intra-agent communications to a plurality of additional LLM agents (deliberations in the form of iterative interaction and communication among the synthetic juror agents in a synthetic deliberative discussion environment, paragraph [0019]);
(f) receiving the one or more conditioned intra-agent communications from the plurality of additional LLM agents;
(g) recording an updated memory state of the LLM agent based on the one or more sent and received conditioned intra-agent communications (juror-agent memories are updating according to the rules in response to the deliberations, paragraphs [0019-0020]); and
(h) generating the prediction based on the updated memory state (a decision prediction is generated based on the deliberation and repeated iterative updating of the synthetic jury model state and individual juror-agent states, paragraphs [0020-0021]).
Worthington does not expressly disclose the retrieving one or more entries of an output of the at least one LLM agent from an LLM agent memory is based on a cosine similarity between a current internal state of the at least one LLM agent and the LLM agent memory.
Doggett discloses a method for retrieving one or more entries of an output of at least one LLM agent from an LLM agent memory, wherein the retrieving is based on a cosine similarity between a current internal state of the at least one LLM agent and the LLM agent memory (a query creation module a query using an LLM by prompting the LLM with a chat history representing the most recent messages generated by a virtual character, paragraphs [0033-0037]; the query is then compared to virtual character memories using a cosine similarity to determine matching memories to generate a response, paragraphs [0052-0053] and [0063]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to retrieve one or more entries of an output of the at least one LLM agent from an LLM agent memory one or more entries of an output of the at least one LLM agent from an LLM agent memory, because it would allow the responses to be grounded to the agent’s memories, thereby avoiding hallucinations and/or other inconsistencies in output from the LLMs, as taught by Doggett (paragraph [0008]).
In regard to claim 11, Worthington discloses the plurality of additional LLM agents are defined by the environment for the information (the rules of deliberation define the interactions allowed between the juror-agents, paragraph [0019]).
In regard to claim 12, Worthington discloses iterating procedures (a)-(h) for one or more additional time points (the simulations are repeated for multiple iterations, paragraphs [0029-0030]).
In regard to claims 13-15, Worthington does not expressly disclose the retrieving one or more entries of an output of the at least one LLM agent from an LLM agent memory is based on a cosine similarity between a current internal state of the at least one LLM agent and the LLM agent memory.
Doggett discloses a method for retrieving one or more entries of an output of at least one LLM agent from an LLM agent memory, wherein the retrieving is based on a cosine similarity between a current internal state of the at least one LLM agent and the LLM agent memory (a query creation module a query using an LLM by prompting the LLM with a chat history representing the most recent messages generated by a virtual character, paragraphs [0033-0037]; the query is then compared to virtual character memories using a cosine similarity to determine matching memories to generate a response, paragraphs [0052-0053] and [0063]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to retrieve one or more entries of an output of the at least one LLM agent from an LLM agent memory one or more entries of an output of the at least one LLM agent from an LLM agent memory, because it would allow the responses to be grounded to the agent’s memories, thereby avoiding hallucinations and/or other inconsistencies in output from the LLMs, as taught by Doggett (paragraph [0008]).
Claim(s) 2, 6, and 10 is/are rejected under 35 U.S.C. 103 as being unpatentable over Worthington, in view of Doggett, and further in view of Simmons et al. (Large Language Models as Subpopulation Representative Models: A Review, hereinafter “Simmons”).
In regard to claims 2, 6 and 10, Worthington discloses the synthetic jury may be applied to non-legal questions such as the evaluation of political candidates (paragraph [0016]) and further discloses the juror-agents may output a vote (paragraphs [0018-0019]). However, Worthington and Doggett do not expressly disclose the prediction comprises an election outcome.
Simmons discloses a comparable method of using LLMs to estimate subpopulation representative models (SRMs), and discloses using SRMs to predict an election outcome (section 4.1).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to apply the voting juror-agents of Worthington to predict the outcome of an election as taught by Simmons by simply tallying the votes output by the juror-agents and this would predictably allow one to predict an election outcome by determining a candidate that received the most votes.
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to BRIAN LOUIS ALBERTALLI whose telephone number is (571)272-7616. The examiner can normally be reached M-F 8AM-3PM, 4PM-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, Bhavesh Mehta can be reached at 571-272-7453. 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.
BLA 4/20/26
/BRIAN L ALBERTALLI/ Primary Examiner, Art Unit 2656
Prosecution Insights