DETAILED ACTION
Continued Examination Under 37 CFR 1.114
1. A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114. Applicant's submission filed on 12/09/2025 has been entered.
Response to Amendment
2. This office action is in response to applicant’s communication filed on 12/09/2025 in response to PTO Office Action mailed 09/11/2025. The Applicant’s remarks and amendments to the claims and/or the specification were considered with the results as follows.
3. In response to the last Office Action, claims 1-6, 8-13 and 15-20 are amended. New claim 22 is added. Claims 7 and 14 are canceled. As a result, claims 1-6, 8-13 and 15-22 are pending in this office action.
4. The information disclosure statements (IDS) submitted on 01/02/2026 and 02/22/2026 are in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statements are being considered by the examiner.
Response to Arguments
5. Applicant’s arguments with respect to claims 1-6, 8-13 and 15-22 have been considered but are moot in view of the new ground of rejection(s).
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1, 3, 6, 8, 10, 13, 15, 17, 20 and 22 are rejected under 35 U.S.C. 103 as being unpatentable by Mueller (US 2025/0077870 A1) and in view of in view of Zhang (CN 105809473 B).
Referring to claims 1, 8 and 15, Mueller discloses an apparatus comprising: a memory configured to store an in-context learning model comprising a memory and an artificial intelligence AI model (See para. [0012], para [0045] and Figures 1-4, a storage system includes a personal computer with a memory system to store a classification-based machine learning model 200); and a processor coupled to the memory (See para. [0045] and para. [0047] and Figure 5, the storage system includes one or more processors and one or more memory), the processor configured to:
store a table comprising a plurality of columns corresponding to a plurality of attributes and a plurality of rows of data corresponding to a plurality of records, respectively (See para. [0013]- para. [0014], storing tabular data [e.g., tables with various rows and columns {attributes}] includes the data in tables),
receive an inference task (See para. [0017] and Figure 2, processing an approximate Bayesian inference on a plurality of tabular datasets in a Bayesian framework for supervised learning) for a row of data of a target record (See para. [0012] and para. [0035], para. [0044] and Figures 2 and 3, receiving an input data 400 [e.g., a tabular data] to be processed and/or executed by a classification-based machine learning model),
convert the row of data into a target embedding (See para. [0013], para. [0016] and Figure 2, converting the tabular data [e.g. tabular data] into a large language model into embeddings 220, 222, 224);
retrieve, by retriever, a subset of rows from the table […that are similar to the target record based on comparison of attribute values within the subset of records] (See para. [0013], para. [0016], para. [0044] identifying and retrieving columns, rows, numbers, names or other characteristics between different portions within the input data 400 which includes portions 402, 404, 406, 408, 410),
execute the AI model on the subset of rows to tune the AI model (See para. [0035], para. [0044], executing the classification-based machine learning model 200 to generate a dataset-specific trained neural network), and
execute the tuned AI model on the row of target record to generate a predicted result for the inference task (See para. [0044], processing the trained neural network classifies each portion of the target record to generate predictions for subsequent portion [e.g., 412, 414]).
Mueller discloses everything except retrieve a subset of data records rows based on a size of the memory and respective distances between respective embeddings of the subset of rows and a target embedding being within a distance threshold in vector space.
Zhang discloses retrieve a subset of data records rows(See abstract, obtaining a portion of search record) based on a size of the memory and respective distances between respective embeddings of the subset of rows and a target embedding being within a distance threshold in vector space (See Zhang, translation pages 10-12 and Figure 4, obtaining a search record based on word embedding matrix, calculating a distance between the normalized word embedding corresponding to the search record [e., target embedding] and the normalized word embedding corresponding to the word embedding matrix from each service provider [e.g., respective embeddings], the distance between the normalized word embedding corresponding to the search record and the normalized word embedding corresponding to each service provider is determined to determine the matching degree, note each word of all sample data after word segmentation is contained in a word embedding matrix, the row height of the matrix is word embedding V |, the column width of the matrix is emb-size. Here, | V | is the number of all possible words, namely the dictionary size. the size of the emb-size is a preset value, the preset value is an empirical value, normally set locating 50 to 1000. Each row of parameters in the matrix is a vector with a length of emb-size, which is referred to as the word embedding word of this row).
Therefore, it would have been obvious to a person of ordinary skill in the computer art before the effective filing date of the claimed invention to modify the AI model of Mueller to retrieve a subset of data records rows based on a size of the memory and respective distances between respective embeddings of the subset of rows and a target embedding being within a distance threshold in vector space, as taught by Zhang. Skilled artisan would have been motivated to construct a training dataset that can provide matching model parameters in order to obtain the optimal solution of the word embedding matrix (See Zhang’s translation pages 6 and 7]). In addition, both references (Zhang and Mueller) teach features that are directed to analogous art and they are directed to the same field of endeavor, such as using a machine-learning model to match features in data records. This close relation between all references highly suggests an expectation of success.
As to claims 3, 10 and 17, Mueller discloses wherein the plurality of records corresponds to a plurality of users, the target record corresponds to a target user, and the processor is configured to train the AI model on a subset of records of other users with respect to the target user to generate the trained AI model (See para. [0051], 534, 536, 538, 540 may access storage system 500 directly through network 502 or through secondary network 506. Further, storage system 500 may be connected to network 502 through secondary network 506, as illustrated with link line 542).
As to claims 6, 13 and 20, Mueller discloses wherein the processor is configured to modify parameters of the AI model based on execution of the AI model on the subset of records prior to execution of the AI model on the target record (See para. [0008], producing hyper-parameters or candidate machine learning model architectures, the foundation machine learning model directly generates a trained neural network and training a universal machine learning model on millions of tabular classification tasks to generate a trained neural network with its corresponding weights specific to one “target” classification task (provided as a prompt)).
As to claim 22, Mueller in view of Zhang discloses the processor is configured to execute nearest-neighbor functionality on the plurality of embeddings and the target embeddings to identify the distance threshold in vector space (See Zhang, translation pages 10-12 and Figure 4, obtaining a search record based on word embedding matrix, calculating a distance between the normalized word embedding corresponding to the search record [e., target embedding] and the normalized word embedding corresponding to the word embedding matrix from each service provider [e.g., respective embeddings], the distance between the normalized word embedding corresponding to the search record and the normalized word embedding corresponding to each service provider is determined to determine the matching degree, note the matching degree of each service provider and the user is determined by calculating the distance of rep (S) and rep (H) and the rep (P) corresponding to each service provider using the included angle between the vector (Cosine included angle) to obtain the distance between rep (S) and rep (P) and rep (H) and rep (P), the two distances are added to obtain the final distance MatchScore: MatchScore = Cosine (rep (S), rep (P)) + Cosine (rep (H), rep (P))).
Therefore, it would have been obvious to a person of ordinary skill in the computer art before the effective filing date of the claimed invention to modify the AI model of Mueller to retrieve a subset of data records rows based on a size of the memory and respective distances between respective embeddings of the subset of rows and a target embedding being within a distance threshold in vector space, as taught by Zhang. Skilled artisan would have been motivated to construct a training dataset that can provide matching model parameters in order to obtain the optimal solution of the word embedding matrix (See Zhang’s translation pages 6 and 7]). In addition, both references (Zhang and Mueller) teach features that are directed to analogous art and they are directed to the same field of endeavor, such as using a machine-learning model to match features in data records. This close relation between all references highly suggests an expectation of success.
Claim 21 is rejected under 35 U.S.C. 103 as being unpatentable by Mueller (US 2025/0077870 A1) in view of Zhang (CN 105809473 B) and further in view of Parham (US 2024/0248963 A1).
As to claim 21, Mueller does not explicitly disclose identify the subset of records based on a location of a subset of embeddings within the vector space being within a distance threshold radius of the target embedding within the vector space.
Parham discloses identify the data records in the table based on a location of the target embedding within the vector space and a distance threshold radius around the target embedding within the vector space (See para. [0154], determining, for each data record portion, can be done with one or more nearest neighbors based on the embedding. The method includes searching each nearest neighbor for a particular data record portion based on one or more keywords associated with the set of topics; determining one or more topic predictions associated with each data record portion of the set of data record portions based on the searching; and outputting the one or more topic predictions to the ensemble as the second set of predictions. The method can further determine, based on the searching, whether a particular keyword is present in a threshold number of nearest neighbors, wherein a particular topic of the set of topics is identified for a particular data record portion when one or more keywords are identified within at least the threshold number of nearest neighbors).
Therefore, it would have been obvious to a person of ordinary skill in the computer art before the effective filing date of the claimed invention to modify the AI model of Mueller to include a RAG model to identify the data records in the table based on location of the target embedding within the vector space, as taught by Parham. Skilled artisan would have been motivated to identify relevant information that are pertinent to measuring and analysis of topics to attain higher quality and more accurate insights (See Parham, para. [0019]). In addition, all references (Parham, Zhang and Mueller) teach features that are directed to analogous art and they are directed to the same field of endeavor, such as using a machine-learning model to match features in data records. This close relation between all references highly suggests an expectation of success.
Claims 2, 5, 9, 12, 16 and 19 are rejected under 35 U.S.C. 103 as being unpatentable by Mueller (US 2025/007870 A1) and in view of Zhang (CN 105809473 B) and further in view of Vu (2023/0115321 A1).
As to claims 2, 9 and 16, Mueller does not explicitly disclose retrain the pre-trained AI model based on execution of the pre-trained AI model on the subset of records to generate a fine-tuned AI model.
Vu discloses the AI model comprises a pre-trained AI model, and the processor is configured to retrain the pre-trained AI model based on execution of the pre-trained AI model on the subset of records to generate a fine-tuned AI model (See para. [0028] and Figure 11, defining a model by fine-tuning a portion of a pre-trained model using client specific data)).
Therefore, it would have been obvious to a person of ordinary skill in the computer art before the effective filing date of the claimed invention to modify the model of Mueller to retrain the pre-trained AI model based on execution of the pre-trained AI model on the subset of records to generate a fine-tuned AI model, as taught by Vu. Skilled artisan would have been motivated to conduct a training that includes updating parameters within an incomplete subset of a set of layers within the pre-trained language model to be updated, while freezing parameters within one or more other layers of the pre-trained language model during the training (See Vu, para. [0002]). In addition, all references (Zhang, Vu and Mueller) teach features that are directed to analogous art and they are directed to the same field of endeavor, such as pre-training a language model. This close relation between all references highly suggests an expectation of success.
As to claims 5, 12 and 19, Mueller does not explicitly disclose concatenate the subset of records and the target record to generate an augmented set of records
Vu discloses wherein the processor is configured to concatenate the subset of records and the target record to generate an augmented set of records and execute the AI model on the augmented set of records to generate the trained AI model (See para. [0022], para. [0137] and para. [0145] and Figure 5, most models explicitly define hyperparameters that control different aspects of the models such as memory or cost of execution. However, additional hyperparameters may be defined to adapt a model to a specific scenario. For example, the hyperparameters may include the number of hidden units or layers of a model, the learning rate of a model, the convolution kernel width, or the number of parameters for a model. Each iteration of training can involve finding a set of model parameters for the model 408 (configured with a defined set of hyperparameters) so that the value of the objective function using the set of model parameters is smaller than the value of the objective function using a different set of model parameters in a previous iteration. The objective function can be constructed to measure the difference between the outputs inferred using the model 408 and the ground truth annotated to the subset of augmented pre-training data assets 414 using the labels 416).
Therefore, it would have been obvious to a person of ordinary skill in the computer art before the effective filing date of the claimed invention to modify the model of Mueller to concatenate the subset of records and the target record to generate an augmented set of records, as taught by Vu. Skilled artisan would have been motivated to conduct a training that includes updating parameters within an incomplete subset of a set of layers within the pre-trained language model to be updated, while freezing parameters within one or more other layers of the pre-trained language model during the training (See Vu, para. [0002]). In addition, all references (Zhang, Vu and Mueller) teach features that are directed to analogous art and they are directed to the same field of endeavor, such as pre-training a language model. This close relation between all references highly suggests an expectation of success.
Allowable Subject Matter
Claims 4, 11 and 18 are objected to as being dependent upon a rejected base claim, but would be allowable if rewritten in independent form including all of the limitations of the base claim and any intervening claims.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to YUK TING CHOI whose telephone number is (571)270-1637. The examiner can normally be reached Monday-Friday 9am-6pm.
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, AMY NG can be reached at 5712701698. 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.
/YUK TING CHOI/Primary Examiner, Art Unit 2164