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 .
Status of Claims
Claims 1, 3, 6, 8, 10, 13, 15, 17, 20, and 21 have been amended. Claims 2, 7, and 14 has been canceled. Claims 22-23 has been added. Claims 1, 3-6, 8-13, and 15-23 are pending and rejected in the application. This action is Final.
Arguments
Applicant Argues
The original specification also fully supports executing the ICL model using the
populated local context to generate a predicted output. For example, paragraphs [0031], [0032], [0072], [0108], [0109], and [0110] describe generating a local context based on a target record and task, executing the in-context learning model using the local context rather than the global context, and generating a predictive output with respect to the task. As such, the amended claim language is expressly supported by the original disclosure.
Accordingly, reconsideration and withdrawal of the rejection are requested.
Examiner Responds:
Applicant's 35 U.S.C. 112(a) or 35 U.S.C. 112 (pre-AIA ), first paragraph, arguments have been considered and are persuasive for claims 1, 3-6, 8-13, and 15-23. Thus, the 35 U.S.C. 112(a) or 35 U.S.C. 112 (pre-AIA ), first paragraph, rejection has been withdrawn.
Applicant Argues
The combination also fails to describe populating a limited-size memory of an ICL model with the local context in place of a global context. Merrick merely retrieves neighboring examples. Those examples are not described as replacing a global context supplied to an ICL model, nor does Merrick discuss constructing an inference prompt from only the retrieved subset. Lee is even farther removed, as the partitions are not related to global context nor any prompt supplied to AI model. Instead, Lee distributes different partitions of a RegEx rule set across different edge devices for parallel execution.
Furthermore, the combination also fails to describe executing the ICL model using the populated local context in place of the global context to generate a predicted output. Merrick performs similarity retrieval, but does not describe executing an ICL model using local context rather than a global context. Lee likewise fails to describe in-context learning whatsoever.
Examiner Responds:
Applicant's 35 USC § 103 arguments, noted above, with respect to claims 1, 3-6, 8-13, and 15-23 have been considered but are moot in view of the new ground(s) of rejection.
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 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, 2, 3, 4, 8, 9, 10, 11, 15, 16, 17, 18, and 21 are rejected under 35 U.S.C. 103 as being unpatentable over Bodenstab et al. U.S. Patent Publication (2025/0291780; hereinafter: Bodenstab) in view of Merrick et al. U.S. Patent Publication (2025/0045634; hereinafter: Merrick) and further in view of Breejen et al. Non-Patent Publication (“Fine-Tuning the Retrieval Mechanism for Tabular Deep Learning”, 2023; hereinafter: Breejen)
Claims 1, 8, and 15
As to claims 1, 8, and 15, Bodenstab discloses an apparatus, comprising:
a memory (paragraph[0022], the reference describes using a retrieval augmented generation model.); and
a processor coupled to the memory, the processor configured to (paragraph[0056], the reference describes using one or more processors.):
store a table comprising a plurality of rows (paragraph[0012], the reference describes a flowsheets containing table and rows.);
receive a target record of a task (paragraph[0033], the reference describes the RAG model receiving records.);
Bodenstab does not appear to explicitly disclose
convert the target record into a target embedding and convert the plurality of rows into a plurality of embeddings, in vector space;
determine a distance around the target embedding within vector space based on execution of a nearest neighbor algorithm;
retrieve a subset of rows from the table which have a size that fits within a limited-size memory of an in-context learning (ICL) model to form a local context, the subset of rows corresponding to embeddings that are within the distance around the target embedding;
populate the limited-size memory of the ICL model with the local context comprising the subset of rows in place of a global context; and
execute the ICL model using the populated local context in place of the global context on the target record and the task to generate a predicted output with respect to the task.
However, Merrick discloses covert the target record into a target embedding and convert the plurality of rows into a plurality of embeddings, in vector space (paragraph[0023]-paragraph[0024], the reference describes a transforming row data.);
determine a distance around the target embedding within vector space based on execution of a nearest neighbor algorithm (paragraph[0085], the reference describes using nearest neighbor to determine the distance of the embeddings.);
retrieve, via an artificial intelligence (AI) model, a subset of rows from the table corresponding to embeddings that are within the distance around the target embedding within the vector space (paragraph[0086], the reference describes determining a set of rows from a table to gather based on the distances that satisfy a threshold.). It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to a person having ordinary skill in the art to which said subject matter pertains to have modified the teachings of Bodenstab with the teachings of Merrick to determine the distance of embeddings which would result in the claim invention. The skilled artisan would have been motivated to improve the teachings of Bodenstab with the teachings of Merrick to efficiently generate a mathematical representation of data table rows that capture semantic similarities and difference between data table rows (Merrick: paragraph[0017]).
The combination of Bodenstab and Merrick do not appear to explicit disclose retrieve a subset of rows from the table which have a size that fits within a limited-size memory of an in-context learning (ICL) model to form a local context, the subset of rows corresponding to embeddings that are within the distance around the target embedding;
populate the limited-size memory of the ICL model with the local context comprising the subset of rows in place of a global context; and
execute the ICL model using the populated local context in place of the global context on the target record and the task to generate a predicted output with respect to the task.
However, Breejen discloses retrieve a subset of rows from the table which have a size that fits within a limited-size memory of an in-context learning (ICL) model to form a local context, the subset of rows corresponding to embeddings that are within the distance around the target embedding (Introduction, the reference describes the TabPFN retrieving a dataset for the model.);
populate the limited-size memory of the ICL model with the local context comprising the subset of rows in place of a global context (pages 1 and 2 of PDF, the reference describes retrieving the rows that is specific for an in-context learning model.); and
execute the ICL model using the populated local context in place of the global context on the target record and the task to generate a predicted output with respect to the task (page 2 and 3 of the PDF, the reference describes using the in-context learning model.). It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to a person having ordinary skill in the art to which said subject matter pertains to have modified the teachings of Bodenstab with the teachings of Merrick and Breejen to determine partition of data to be used TabPFN system which would result in the claim invention. The skilled artisan would have been motivated to improve the teachings of Bodenstab with the teachings of Merrick and Breejen to efficiently improve fine-tuning the pretrained TabPFN model (Breejen: Abstract).
Claims 9 and 16
As to claims 9 and 16, the combination of Bodenstab, Merrick, and Breejen discloses all the elements in claim 1, as noted above, and Bodenstab further disclose wherein the plurality of rows comprise data from a plurality of columns of the table (paragraph[0012], the reference describes using plurality of rows.).
Claims 3, 10, and 17
As to claims 3, 10, and 17, the combination of Bodenstab, Merrick, and Breejen discloses all the elements in claim 1, as noted above, and Merrick further disclose wherein the processor is configured to transform content values in the plurality of rows to generate the plurality of embedding, respectively, and transform a corresponding set of content values of the target record to generate the target embedding (paragraph[0025], the reference describes the transformer that encodes data.).
Claims 4, 11, and 18
As to claims 4, 11, and 18, the combination of Bodenstab, Merrick, and Breejen discloses all the elements in claim 3, as noted above, and Merrick further disclose wherein the processor is configured to generate a predetermined radius around the target embedding within vector space based on execution of the nearest neighbor algorithm on the plurality of embeddings and the target embedding (paragraph[0085], the reference describes using a threshold (i.e., a predetermined radius, as claimed) around the nearest neighbor calculation of embedded data.).
Claim 21
As to claim 21, the combination of Bodenstab, Merrick, and Breejen discloses all the elements in claim 1, as noted above, and Merrick further disclose wherein the limited-size memory comprises a fixed capacity, and the processor is configured to select the subset of rows such that the subset of rows fit within the fixed capacity during a single execution of the ICL model (paragraph[0086], the reference describes generating a set of rows to be processed by the model in one execution.).
Claims 5, 12, and 19 are rejected under 35 U.S.C. 103 as being unpatentable over Bodenstab et al. U.S. Patent Publication (2025/0291780; hereinafter: Bodenstab) in view of Merrick et al. U.S. Patent Publication (2025/0045634; hereinafter: Merrick) and further in view of Breejen et al. Non-Patent Publication (“Fine-Tuning the Retrieval Mechanism for Tabular Deep Learning”, 2023; hereinafter: Breejen) and further in view of Mauer et al. U.S. Patent Publication (2022/0141263; hereinafter: Mauer)
Claims 5, 12, and 19
As to claims 5, 12, and 19, the combination of Bodenstab, Merrick, and Breejen discloses all the elements in claim 1, as noted above, but do not appear to explicitly disclose wherein the processor is configured to determine distances between the plurality of embeddings and the target embedding based on execution of a dot product algorithm on the plurality of embeddings and the target embedding.
However, Mauer discloses wherein the processor is configured to determine distances between the plurality of embeddings and the target embedding based on execution of a dot product algorithm on the plurality of embeddings and the target embedding (paragraph[0018], the reference describes using dot product to determine distances.). It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to a person having ordinary skill in the art to which said subject matter pertains to have modified the teachings of Bodenstab with the teachings of Merrick, Breejen, and Mauer to use a dot product which would result in the claim invention. The skilled artisan would have been motivated to improve the teachings of Bodenstab with the teachings of Merrick, Breejen, and Mauer to efficiently facilitate user discovery of communication platform features to the user without explicit prompting or request by the user (Mauer: paragraph[0009[).
Claims 6, 13, and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Bodenstab et al. U.S. Patent Publication (2025/0291780; hereinafter: Bodenstab) in view of Merrick et al. U.S. Patent Publication (2025/0045634; hereinafter: Merrick) and further in view of Breejen et al. Non-Patent Publication (“Fine-Tuning the Retrieval Mechanism for Tabular Deep Learning”, 2023; hereinafter: Breejen) and further in view of Bourdev U.S. Patent Publication (7,343,551; hereinafter: Bourdev)
Claims 6, 13, and 20
As to claims 6, 13, and 20, the combination of Bodenstab, Merrick, and Breejen discloses all the elements in claim 1, as noted above, but do not appear to explicitly disclose wherein the processor is further configured to receive a request to execute the ICL model on the target record via a software application, and display the predicted output via a graphical user interface (GUI) of the software application.
However, Bourdev discloses wherein the processor is further configured to receive a request to execute the ICL model on the target record via a software application, and display the predicted output via a graphical user interface (GUI) of the software application (Column 9, lines 15-36, the reference describes display predicted values on a user interface.). It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to a person having ordinary skill in the art to which said subject matter pertains to have modified the teachings of Bodenstab with the teachings of Merrick, Breejen, and Bourdev to show predictions on an interface which would result in the claim invention. The skilled artisan would have been motivated to improve the teachings of Bodenstab with the teachings of Merrick, Breejen, and Bourdev to efficiently predict values being selected from the proposed values based on the likelihood assessments (Bourdev: Column 2, lines 1-5).
Final Action
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 extension fee 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 date of this final action.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to DAWAUNE A CONYERS whose telephone number is (571)270-3552. The examiner can normally be reached on M-F 8:00am-4:30pm EST. EST.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Ann J. Lo can be reached on (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 an application may be obtained from the Patent Application Information Retrieval (PAIR) system. Status information for published applications may be obtained from either Private PAIR or Public PAIR. Status information for unpublished applications is available through Private PAIR only. For more information about the PAIR system, see http://pair-direct.uspto.gov. Should you have questions on access to the Private PAIR system, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative or access to the automated information system, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/DAWAUNE A CONYERS/Primary Examiner, Art Unit 2159
/DAWAUNE A CONYERS/Primary Examiner, Art Unit 2152 February 24, 2024
Prosecution Insights