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 .
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 USC 101 because the claimed invention is directed to an abstract idea without significantly more.
Claim 1 recites (emphasis added)
1. A method for scaling reinforcement learning comprising:
receiving, by one or more processors, model-generated responses to a task and a prompt associated with providing respective reward scores for the model-generated responses;
processing, by the one or more processors, the model-generated responses and the prompt
using a generative model to generate reward data indicative of the reward scores;
training, by the one or more processors, one or more machine learning models via reinforcement learning based on the reward data;
and outputting, by the one or more processors, the one or more trained machine learning models.
Examiner finds that the emphasized (bolded) portions of claim 1 above recite an abstract idea—namely, mental processes. See MPEP 2106.04(a)(2)(III):
Accordingly, the ‘mental processes’ abstract idea grouping is defined as concepts performed in the human mind, and examples of mental processes include observations, evaluations, judgments, and opinions
When read as a whole, the recited limitations are directed to using mental steps to observe, evaluate, and make judgements about electronic data.
Taking each element individually, Examiner provides the following analysis:
Bolded Abstract Idea Claim Elements
Examiner analysis of bolded abstract idea elements considered individually
1. A method for scaling reinforcement learning comprising:
receiving, by one or more processors, model-generated responses to a task and a prompt associated with providing respective reward scores for the model-generated responses;
processing, by the one or more processors, the model-generated responses and the prompt
This element merely requires observation and evaluation of the responses and the prompt.
using a generative model to generate reward data indicative of the reward scores;
This element merely requires observation and evaluation of the responses and the prompt and an evaluation/judgment as to how to generate the reward data.
training, by the one or more processors, one or more machine learning models via reinforcement learning based on the reward data;
and outputting, by the one or more processors, the one or more trained machine learning models.
Turning to the additional elements and whether they integrate the exception and whether they recite an inventive concept, Examiner provides the following analysis:
Italicized Additional elements
Examiner analysis of italicized additional elements and whether they integrate the exception and whether they recite an inventive concept.
Relevant MPEP sections
1. A method for scaling reinforcement learning comprising:
This element recites mere instructions to apply the exception and/or generally links the abstract idea to the field of use of reinforcement learning and thus does not integrate and does not recite an inventive concept.
2106.05(f), (h)
receiving, by one or more processors, model-generated responses to a task and a prompt associated with providing respective reward scores for the model-generated responses;
This element insignificant extra solution activity in the form of mere data gathering.
This element recites a well-understood, routine, and conventional (WURC) computer function (i.e. receiving or transmitting data over a network) and thus does not recite an inventive concept.
2106.05(g)
2106.05(d)(II)
processing,
by the one or more processors,
This element recites mere instructions to apply the exception thus does not integrate and does not recite an inventive concept.
2106.05(f)
the model-generated responses and the prompt
using a generative model to
This element recites mere instructions to apply the exception and/or generally links the abstract idea to the field of use of generative AI and thus does not integrate and does not recite an inventive concept.
2106.05(f),(h)
generate reward data indicative of the reward scores;
training, by the one or more processors, one or more machine learning models via reinforcement learning based on the reward data;
This element generally links the abstract idea to the field of use of reinforcement learning and thus does to integrate the exception and does not recite an inventive concept.
2106.05(h)
and outputting, by the one or more processors, the one or more trained machine learning models.
This element generally links the abstract idea to the field of use of machine learning and thus does to integrate the exception and does not recite an inventive concept.
2106.05(h)
The additional elements above “[a]dd nothing … that is not already present when the steps are considered separately’”. MPEP 2106.05 (I)(B)(quoting Alice).
As such, when the claim elements are considered as a whole and individually, claim 1 recites an abstract idea without significantly more.
Claims 11 and 20 are also rejected for the reasons given above for claim 1. Additionally, “A system comprising: one or more processors; and one or more storage devices coupled to the one or more processors and storing instructions that, when executed by the one or more processors, cause the one or more processors to perform operations for scaling reinforcement learning, the operations comprising” and “A non-transitory computer readable medium for storing instructions that, when executed by one or more processors, cause the one or more processors to perform operations for scaling reinforcement learning, the operations comprising” recite mere instructions to apply the exception and thus do not integrate the exception and do not recite an inventive concept.
Dependent claims 3-10 and 12-19 are rejected under 35 USC 101 for the reasons indicated below.
Claim
Bold = abstract idea
Italics = additional elements
Analysis
MPEP
2. The method of claim 1, wherein the generative model is at least one of a large language model, large foundation model, or large graphical model.
This element generally links the abstract idea to the field of use of large language models and thus does not integrate the exception and does not recite an inventive concept.
2106.05(h)
3. The method of claim 1, wherein the prompt comprises instructions for the generative model to rate a quality of the respective responses.
This element merely requires human judgment as to how to craft the prompt.
2106.04(a)(2)(III)
4. The method of claim 3, wherein the instructions comprise rating the quality of the respective responses on a scale.
This element merely requires human judgment as to how to craft the prompt.
2106.04(a)(2)(III)
5. The method of claim 3, wherein the instructions further comprise one or more attributes for the generative model to consider in rating the quality of the respective responses.
This element merely requires human judgment as to how to craft the prompt.
2106.04(a)(2)(III)
6. The method of claim 5, wherein the instructions further comprise descriptions for the one or more attributes.
This element merely requires human judgment as to how to craft the prompt.
2106.04(a)(2)(III)
7. The method of claim 1, wherein processing the model-generated responses and the prompt further comprises calculating a probability weighted average of ratings to generate the reward scores.
This element recites a mathematical calculation.
2106.04(a)(2)(I)
8. The method of claim 7, wherein processing the model-generated responses and the prompt further comprises normalizing the probability weighted average of ratings.
This element recites a mathematical calculation.
2106.04(a)(2)(I)
9. The method of claim 1, wherein the one or more machine learning models are trained via reinforcement learning based on policy-gradient-based techniques.
This element generally links the abstract idea to the field of use of reinforcement learning and thus does not integrate the exception and does not recite an inventive concept.
2106.05(h)
10. The method of claim 1, wherein the task comprises at least one of summarization or dialogue generation.
This element insignificant extra solution activity in the form of mere data gathering.
This element recites a well-understood, routine, and conventional (WURC) computer function (i.e. receiving or transmitting data over a network) and thus does not recite an inventive concept.
2106.05(g)
2106.05(d)(II)
Claims 12-19 are rejected for the same reasons given above for claims 2-9.
The additional elements above “[a]dd nothing … that is not already present when the steps are considered separately’”. MPEP 2106.05 (I)(B)(quoting Alice).
As such, when the claim elements above are considered as a whole and individually, claims recite an abstract idea without significantly more.
Claim Rejections - 35 USC § 102
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.
Claim(s) 1-2, 9, 11-12, and 19-20 is/are rejected under 35 U.S.C. 102(a)(1) as anticipated by Kwon, REWARD DESIGN WITH LANGUAGE MODELS, 27 Feb 20231
Claim
Kwon, REWARD DESIGN WITH LANGUAGE MODELS, 27 Feb 2023
1. A method for scaling reinforcement learning comprising: receiving, by one or more processors, model-generated responses to a task and a prompt associated with providing respective reward scores for the model-generated responses;
p. 2; Fig. 1 step 5 (prompt o3 includes answers (responses) generated by LLM (e.g. “Agreement!”); Examiner finds o4 prompt (“Is Alice a versatile negotiator”); teaches the prompt “associated with providing respective reward scores for the model-generated responses”); See also Fig. 1 caption;
processing, by the one or more processors, the model-generated responses and the prompt
p. 2; Fig. 1 step 2 (“No”) is a result of the processing of the prompt o4 and the previous responses given in prompt o3); see also Fig. 1 caption;
using a generative model to generate reward data indicative of the reward scores;
p. 2; Fig. 1 step 3 (reward score is 0 or 1, for example); see also Fig. 1 caption;
p. 3 section 3 2nd paragraph
Our framework is depicted in Fig. 1. Before training, a user specifies ρ2 which can be N examples
describing their objective or a description of their objective using natural language. In Fig. 1 a user provides
an example of their objective: versatile negotiating behavior. During training, we construct a prompt ρ
by concatenating a description of the task, the user-specified examples/description, an episode’s outcome,
and a question asking if the outcome satisfies the objective. We (1) feed the prompt to the LLM, (2) take
its output, and (3) parse it into an integer using function g; we use a handcrafted, task-specific parser. We
use the integer as the reward signal. (4) The RL agent then updates its weights and rolls out an episode.
(5) We parse the episode outcome into a string using f and continue training; we also instantiate f as
handcrafted, task-specific parser. To evaluate our framework, we sample a trajectory (e.g., a negotiation)
from the agent and evaluate whether the trajectory is aligned with the user’s objective (e.g., whether Alice
demonstrated versatile negotiating behavior)
training, by the one or more processors, one or more machine learning models via reinforcement learning based on the reward data;
p.2 Fig. 1 step 4 (updating the agent teaches training the model via RL based on reward data);
p. 3
Our framework is depicted in Fig. 1. Before training, a user specifies ρ2 which can be N examples
describing their objective or a description of their objective using natural language. In Fig. 1 a user provides
an example of their objective: versatile negotiating behavior. During training, we construct a prompt ρ
by concatenating a description of the task, the user-specified examples/description, an episode’s outcome,
and a question asking if the outcome satisfies the objective. We (1) feed the prompt to the LLM, (2) take
its output, and (3) parse it into an integer using function g; we use a handcrafted, task-specific parser. We
use the integer as the reward signal. (4) The RL agent then updates its weights and rolls out an episode.
(5) We parse the episode outcome into a string using f and continue training; we also instantiate f as
handcrafted, task-specific parser. To evaluate our framework, we sample a trajectory (e.g., a negotiation)
from the agent and evaluate whether the trajectory is aligned with the user’s objective (e.g., whether Alice
demonstrated versatile negotiating behavior)
and outputting, by the one or more processors, the one or more trained machine learning models.
p. 2
Figure 1: Depiction of our framework on the DEALORNODEAL negotiation task. A user provides an example and
explanation of desired negotiating behavior (e.g., versatility) before training. During training, (1) we provide the LLM
with a task description, a user’s description of their objective, an outcome of an episode that is converted to a string,
and a question asking if the outcome episode satisfies the user objective. (2-3) We then parse the LLM’s response
back into a string and use that as the reward signal for the Alice the RL agent. (4) Alice updates their weights and
rolls out a new episode. (5)We parse the episode outcome int a string and continue training. During evaluation, we
sample a trajectory from Alice and evaluate whether it is aligned with the user’s objective.
p. 3
Our framework is depicted in Fig. 1. Before training, a user specifies ρ2 which can be N examples
describing their objective or a description of their objective using natural language. In Fig. 1 a user provides
an example of their objective: versatile negotiating behavior. During training, we construct a prompt ρ
by concatenating a description of the task, the user-specified examples/description, an episode’s outcome,
and a question asking if the outcome satisfies the objective.
We (1) feed the prompt to the LLM, (2) take
its output, and (3) parse it into an integer using function g; we use a handcrafted, task-specific parser. We
use the integer as the reward signal. (4) The RL agent then updates its weights and rolls out an episode.
(5) We parse the episode outcome into a string using f and continue training; we also instantiate f as
handcrafted, task-specific parser. To evaluate our framework, we sample a trajectory (e.g., a negotiation)
from the agent and evaluate whether the trajectory is aligned with the user’s objective (e.g., whether Alice
demonstrated versatile negotiating behavior)
(Examiner finds “updating weights” teaches an outputting of a trained model)
Claims 11 and 20 are rejected for the reasons given above for claim 1.
2. The method of claim 1, wherein the generative model is at least one of a large language model, large foundation model, or large graphical model.
12. The system of claim 11, wherein the generative model is at least one of a large language model, large foundation model, or large graphical model.
p. 2 Fig. 1 (LLM is generative model)
9. The method of claim 1, wherein the one or more machine learning models are trained via reinforcement learning based on policy-gradient-based techniques.
19. The system of claim 11, wherein the one or more machine learning models are trained via reinforcement learning based on policy-gradient-based techniques.
p. 7
DEALORNODEAL is a long-horizon task with a maximum length of 100 timesteps.
An agent Alice must come to an agreement with her partner Bob on the allocation of a set of objects (books,
hats, and balls). Agents are shown a context, which includes the counts of each item and their private
utilities for each item. In the original task, agents get rewarded based on the agreed upon split and their
utilities. If Alice and Bob reach a disagreement, both agents get nothing. We train Alice using on-policy
RL by negotiating against a fixed partner model, which we refer to as Bob. See Sec. A.4 for more details
on the domain and training.
p. 13
Agents are first trained using supervised learning on a dataset of human-human negotiations provided
by (Lewis et al., 2017) to predict the next token. We use a learning rate of 1.0 and batch size of 16. We
then fine-tune these agents using RL where they optimize the expected reward of each dialogue act using
REINFORCE (Williams, 1992). Agents are trained on 250 contexts for 1 epoch with a learning rate of
0.1. We instantiate our policy with four GRUs (Chung et al., 2014). We closely follow the implementation
outlined in (Kwon et al., 2021; Lewis et al., 2017), please refer to those papers for more training details.
p. 11
Ronald J Williams. Simple statistical gradient-following algorithms for connectionist reinforcement
learning. Machine learning, 8(3):229–256, 1992
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) 3-7 and 13-17 is/are rejected under 35 U.S.C. 103 as being unpatentable over Kwon as applied to claim 1 and 11 above and further in view of Yuan, Self-Rewarding Language Models Jan. 18 2024.
With respect to claim 3 and 13, it appears Kwon fails to explicitly teach “wherein the prompt comprises instructions for the generative model to rate a quality of the respective responses.”
However, Yuan, Self-Rewarding Language Models teaches “3. The method of claim 1, wherein the prompt comprises instructions for the generative model to rate a quality of the respective responses” on p. 3 (“In this data, the input prompt asks the model to evaluate the quality of a given response to a particular instruction.”); and p. 4 Fig. 2
Add 1 point if the response is relevant and provides some information related to the user’s inquiry, even if it is incomplete or contains some irrelevant content.- Add another point if the response addresses a substantial portion of the user’s question, but does not completely resolve the query or provide a direct answer.- Award a third point if the response answers the basic elements of the user’s question in a useful way, regardless of whether it seems to have been written by an AI Assistant or if it has elements typically found in blogs or search results.- Grant a fourth point if the response is clearly written from an AI Assistant’s perspective, addressing the user’s question directly and comprehensively, and is well-organized and helpful, even if there is slight room for improvement in clarity, conciseness or focus.- Bestow a fifth point for a response that is impeccably tailored to the user’s question by an AI Assistant, without extraneous information, reflecting expert knowledge, and demonstrating a high-quality, engaging, and insightful answer.
(Examiner finds the point system in the above teaches a “rating the quality of the responses” on a scale).
Yuan and Kwon are analogous art because they are from the same field of endeavor as the claimed invention.
It would have been obvious to one skilled in the art before the effective filing date of the invention to modify the prompt in Kwon to include “wherein the prompt comprises instructions for the generative model to rate a quality of the respective responses” as taught by Yuan.
The motivation would have been to improve the instruction following of LLMs and to improve the quality of rewards thereby increasing model performance. See Yuan abstract.
With respect to claim 4 and 14, Yuan teaches “wherein the instructions comprise rating the quality of the respective responses on a scale.” p. 3 (“In this data, the input prompt asks the model to evaluate the quality of a given response to a particular instruction.”); p. 4 Fig. 2
Add 1 point if the response is relevant and provides some information related to the user’s inquiry, even if it is incomplete or contains some irrelevant content.- Add another point if the response addresses a substantial portion of the user’s question, but does not completely resolve the query or provide a direct answer.- Award a third point if the response answers the basic elements of the user’s question in a useful way, regardless of whether it seems to have been written by an AI Assistant or if it has elements typically found in blogs or search results.- Grant a fourth point if the response is clearly written from an AI Assistant’s perspective, addressing the user’s question directly and comprehensively, and is well-organized and helpful, even if there is slight room for improvement in clarity, conciseness or focus.- Bestow a fifth point for a response that is impeccably tailored to the user’s question by an AI Assistant, without extraneous information, reflecting expert knowledge, and demonstrating a high-quality, engaging, and insightful answer.
(Examiner finds the point system in the above teaches a “rating the quality of the responses” on a scale (i.e. +1 to +5)).
The motivation to modify the prompt in Kwon to include the elements in claim 4 is the same given in claim 3.
With respect to claim 5 and 15 , Yuan teaches “wherein the instructions further comprise one or more attributes for the generative model to consider in rating the quality of the respective responses” on p. 4 Fig. 2
Add 1 point if the response is relevant and provides some information related to the user’s inquiry, even if it is incomplete or contains some irrelevant content.- Add another point if the response addresses a substantial portion of the user’s question, but does not completely resolve the query or provide a direct answer.- Award a third point if the response answers the basic elements of the user’s question in a useful way, regardless of whether it seems to have been written by an AI Assistant or if it has elements typically found in blogs or search results.- Grant a fourth point if the response is clearly written from an AI Assistant’s perspective, addressing the user’s question directly and comprehensively, and is well-organized and helpful, even if there is slight room for improvement in clarity, conciseness or focus.- Bestow a fifth point for a response that is impeccably tailored to the user’s question by an AI Assistant, without extraneous information, reflecting expert knowledge, and demonstrating a high-quality, engaging, and insightful answer.
(Examiner finds the point system in the above teaches a “rating the quality of the responses” on a scale; Examiner finds “relevancy”; “completeness”; usefulness; “helpfulness” are all attributes to consider in the rating).
The motivation to modify the prompt in Kwon to include the elements in claim 5 is the same given in claim 3.
With respect to claim 6 and 16, Yuan teaches “ wherein the instructions further comprise descriptions for the one or more attributes” p. 4 Fig. 2
Add 1 point if the response is relevant and provides some information related to the user’s inquiry, even if it is incomplete or contains some irrelevant content.- Add another point if the response addresses a substantial portion of the user’s question, but does not completely resolve the query or provide a direct answer.- Award a third point if the response answers the basic elements of the user’s question in a useful way, regardless of whether it seems to have been written by an AI Assistant or if it has elements typically found in blogs or search results.- Grant a fourth point if the response is clearly written from an AI Assistant’s perspective, addressing the user’s question directly and comprehensively, and is well-organized and helpful, even if there is slight room for improvement in clarity, conciseness or focus.- Bestow a fifth point for a response that is impeccably tailored to the user’s question by an AI Assistant, without extraneous information, reflecting expert knowledge, and demonstrating a high-quality, engaging, and insightful answer.
(Examiner finds the point system in the above teaches a “rating the quality of the responses” on a scale; Examiner finds “relevancy”; “completeness”; usefulness; “helpfulness” are all attributes to consider in the rating; Each dash “-“ in the prompt in Fig. 2 describe each attribute respectively).
The motivation to modify the prompt in Kwon to include the elements in claim 6 is the same given in claim 3.
With respect to claim 7 and 17, Yuan teaches “7. The method of claim 1, wherein processing the model-generated responses and the prompt further comprises calculating a probability weighted average of ratings to generate the reward scores” on p. 6
When evaluating candidate responses, as there is variance to these scores, in our experiments we also use sampled decoding (with the same parameters) and generate these evaluations multiple (3) times and take the average.
(Weighted average broadly includes equal weights).
The motivation to modify the prompt in Kwon to include the elements in claim 7 would have been to take into account the variance in the data thereby increasing the accuracy of the evaluations. See id.
Claim(s) 8 and 18 is/are rejected under 35 U.S.C. 103 as being unpatentable over Kwon as applied to claim 1 and 11 above and further in view of Yuan as applied to claim 7 and 17 above and further in view of Li, PRD: Peer Rank and Discussion Improve Large Language Model based Evaluation 6 July 2023.
With respect to claim 8, it appears Kwon et al. fails to explicitly teach “8. The method of claim 7, wherein processing the model-generated responses and the prompt further comprises normalizing the probability weighted average of ratings.”
However, Li, PRD: Peer Rank and Discussion Improve Large Language Model based Evaluation teaches “wherein processing the model-generated responses and the prompt further comprises normalizing the probability weighted average of ratings” p. 3 2nd paragraph under section 2.1.1
Our win rate calculation gives differing weight to the scores provided by different reviewers (A, B, C) based on the performance of the corresponding reviewers as a contestant (1, 2, 3)..
Initially, all reviewers are given the same weight. On each iteration of the calculation, the win rate for each contestant is calculated using the current weights. The win rates are scaled to the range of [0, 1] using a linear scaling, and then again scaled so that their sum is 1, and these results are used as the weights for the next round
p. 3 section 2.1.1 right column 3rd paragraph;
Let αk r be the weight assigned to reviewer r after iteration k. Initially, α0 r = 1/|R|, so that all reviewers have the same weight, and the weights add to 1.– We assume each reviewer LLM has the same capabilities to start. The score of contestant c ∈ C for iteration k is the weighted average of the raw win rates for contestant c. We set the weights for the next iteration to αk: scorek c = αk−1 r r∈R · Wc r αk = Normalize(MinMax(scorek)) (2) where the weights are scaled to a range of [0,1] and finally normalized to have sum equal to 1:
Li and Kwon et al. are analogous art because they are from the same field of endeavor as the claimed invention.
It would have been obvious to one skilled in the art before the effective filing date of the invention to modify “processing the model-generated responses” in Kwon to include “wherein processing the model-generated responses and the prompt further comprises normalizing the probability weighted average of ratings” as taught by Li.
The motivation would have been to maintain data integrity and decrease errors in data interpretation.
Claim(s) 10 is/are rejected under 35 U.S.C. 103 as being unpatentable over Kwon as applied to claim 1 above and further in view of Vu 12,632,478.
With respect to claim 10, Kwon fails to explicitly teach “10. The method of claim 1, wherein the task comprises at least one of summarization or dialogue generation.”
However, Vu 12,632,478 teaches “wherein the task comprises at least one of summarization or dialogue generation” in Col. 16:1-5:
The reward model 430 may employ reinforcement learning as described in Daniel M Ziegler et al., “Finetuning language models from human preferences”, arXiv preprint arXiv: 1909.08593 (2019); Nisan Stiennon et al., “Learning to summarize with human feedback”
And Col. 21:30-42
Dialog processing is a field of computer science that involves communication between a computing system and a human via text, audio, and/or other forms of communication. While some dialog processing involves only simple generation of a response given only a most recent input from a user (i.e., single-turn dialog), more complicated dialog processing involves determining and optionally acting on one or more goals expressed by the user over multiple turns of dialog, such as making a restaurant reservation and/or booking an airline ticket. These multi-turn “goal-oriented” dialog systems typically need to recognize, retain, and use information collected during more than one input during a back-and-forth or “multi-turn” interaction with the user
Vu and Kwon are analogous art because they are in the same field of endeavor as the claimed invention.
It would have been obvious to one skilled in the art before the effective filing date of the invention to modify the task in Kwon to include “wherein the task comprises at least one of summarization or dialogue generation” as taught by Vu.
The motivation would have been to increase the accuracy of chatbots that process airline tickets or restaurant reservations. See Vu id.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to ALBERT M PHILLIPS, III whose telephone number is (571)270-3256. The examiner can normally be reached 10a-6:30pm EST M-F.
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, Ann J Lo can be reached at (571) 272-9767. 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.
/ALBERT M PHILLIPS, III/Primary Examiner, Art Unit 2159
1 Cited on IDS 3/19/2024 Cite No. 19 under Non-Patent Literature Documents.
Prosecution Insights