CTNF 18/934,756 CTNF 99728 Notice of Pre-AIA or AIA Status 07-03-aia AIA 15-10-aia The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA. Information Disclosure Statement 06-52 The information disclosure statement (IDS) was filed on . The submission is in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the examiner. Claim Rejections - 35 USC § 103 07-06 AIA 15-10-15 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 (i.e., changing from AIA to pre-AIA) 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. 07-20-aia AIA 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. 07-23-aia AIA The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows: 1. Determining the scope and contents of the prior art. 2. Ascertaining the differences between the prior art and the claims at issue. 3. Resolving the level of ordinary skill in the pertinent art. 4. Considering objective evidence present in the application indicating obviousness or nonobviousness. 07-21-aia AIA Claim s 1-3, 5-10, 12-17 and 19-20 are rejected under 35 U.S.C. 103 as being unpatentable over Anandkumar et. al. (US Pub. No. 20240265690 A1) in view of Grenader et. al. (US Pub. No. 20240362213 A1) . As per claim 1, Anandkumar teaches “as per claim 1, Anandkumar teaches “a computer-implemented method for leveraging semantic information for a multi-domain visual agent, comprising: sampling questions from question templates for domain-specific label spaces to obtain a unified label space;” (See paragraph 30, the resampler integrates multi-modal labels by the experts (domain-specific label spaces) to obtain the multi-modal feature (unified label space) which has cross attention (makes them unified). “[0030] The expert ensemble vision-language system 100 includes the framework which is divided into a vision encoder 130 that receives multi-modal features and an autoregressive language decoder 140. The framework comprises an experts resampler 120 that integrates multi-modal labels generated by the pre-trained experts 110 into a fixed length token that defines the multi-modal features (embeddings). . In an embodiment, the experts resampler 120 receives multi-modal labels of variable length as input, and outputs a fixed number of tokens via cross attention . The experts resampler 120 applies trainable parameters to the multi-modal labels to generate the multi-modal features … ” See also paragraphs 31-40. Paragraph 25 shows the sampled questions “For example , given multi-modal tokens (encoded from an input image and corresponding multi-modal labels) and a question as the prefix , the expert ensemble vision-language system 100 generates an answer for the visual question answering task; given the multi-modal tokens, the expert ensemble vision-language system 100 generates a caption for the image captioning task. In response to a prefix prompt, the predicted text may be sampled in an autoregressive manner, as in an open-ended setting; or a ranking of the log-likelihood may be performed from a fixed set of completions, as in a closed-ended setting. ”. See also paragraphs 21-25. Paragraphs 39-43 shows that the resampler learns pre-defined (template) input queries (questions) to cross attend and concatenate them. Therefore the “resampler” samples the input questions. “[0039] The experts resampler 120 learns a pre-defined number of latent input queries , to cross-attend a flattened embedding concatenated from the multi-modal labels . The experts resampler 120 then compresses the multi-modal labels into a much smaller number of tokens (multi-modal features) equal to the number of learned latent queries, as a form of auxiliary knowledge distillation .” Therefore within a BRI (broadest reasonable interpretation) it samples input queries. Paragraph 40 also shows that the sampler processes learned queries (questions from question templates) with the use of bi-direction attention. See also paragraphs 46 and 38, shows randomly sampled embeddings (queries). Anandkumar) mapping domain-specific labels from the domain-specific label spaces into natural language descriptions (NLD) to obtain mapped NLD; (The mapped NLD are the experts in the presented reference, the language descriptions such as those seen in paragraphs 4 and 22-26 “[0004] The vision-language model learns skills and domain knowledge via the distinct and separate task-specific expert neural networks . Each expert is independently optimized for a specific task , facilitating the use of domain-specific data and architectures that are not feasible with a single large neural network trained for multiple tasks.” “[0022] The expert ensemble vision-language system 100 receives an input image 105 and multi-modal labels 101, 102, 103, 104, 106, and 107 generated by a variety of pre-trained vision experts (not shown). The experts perform domain-specific vision tasks, processing the input image 105 to predict depth data, object labels, surface normal vector data, edge data, text labels, or segmentation labels corresponding to the multi-modal labels 101, 102, 103, 104, 106, and 107, respectively . In an embodiment at least two experts process the input image 105 to produce multi-modal labels that are input the expert ensemble vision-language system 100 . In an embodiment, additional, fewer, or different experts perform other domain-specific vision tasks to generate multi-modal labels for input to the expert ensemble vision-language system 100.” Paragraph 23 shows a clear descriptions working as a natural language description “[0023] For image captioning, the expert ensemble vision-language system 100 processes the input image 105 and multi-modal labels 101, 102, 103, 104, 106, and 107 to predict “A man is playing baseball in a field.” For visual question answering, the expert ensemble vision-language system 100 receives a question input and processes the input question, the input image 105, and the multi-modal labels 101, 102, 103, 104, 106, and 107 to predict an answer to the question. For example, in response to the input question “What's this person doing,” the expert ensemble vision-language system 100 predicts the answer “Playing baseball.” In another example, in response to the input question “What's the number of this player,” the expert ensemble vision-language system 100 predicts the answer “21.” See also paragraphs 31-32. See also paragraph 24. Anandkumar. ) generating… by combining the questions sampled from the unified label space and the mapped NLD; (See paragraphs 31 and 35, the experts (NLD) are used with cross attention with the features/tokens from the unified space to generate answers or text “[0035] Once trained, the expert ensemble vision-language system 100 comprises a generative model and the loss function 150 may be removed. Vision-language reasoning tasks may be formulated as a language modelling or prefix language modelling problem for the expert ensemble vision-language system 100. For example, given multi-modal tokens (encoded from an input image and corresponding multi-modal labels) and a question as the prefix, the expert ensemble vision-language system 100 generates an answer for the visual question answering task; given the multi-modal tokens, the expert ensemble vision-language system 100 generates a caption for the image captioning task . In response to a prefix prompt, the predicted text may be sampled in an autoregressive manner, as in an open-ended setting; or a ranking of the log-likelihood may be performed from a fixed set of completions, as in a closed-ended setting.” See also paragraph 42, 49-53. ) learning the semantic information by iteratively generating outputs from tokens extracted from the prompts using a large-language model (LLM); and (Paragraphs 37-38 show that the labels/features/embeddings include semantic information, “[0038] In an embodiment, the resolution of high-level semantic labels such as those in object detection, semantic segmentation, and OCR detection, is downsampled by a factor of 4 to conserve running memory. Furthermore, for each object instance, a trainable and randomly sampled embedding is added to distinguish among different object instances… ”, these are learned by outputting the tokens as seen in paragraph 31 “[0031] The vision encoder 130 takes the input image and the corresponding multi-modal labels as input and outputs a sequence of vision tokens . The language decoder 140 processes the vision tokens and an optional word embedding (input question for visual question answering) and outputs predicted text (e.g., caption or answer) . In an embodiment, the language decoder 140 generates text tokens that are converted into the predicted text . In an embodiment, during training, the language decoder 140 is conditioned on the multi-modal features via cross attention and produces a sequence of text tokens as the predicted text. In an embodiment, the vision encoder 130 and the language decoder 140 are based on a transformer architecture. To preserve the rich domain specific knowledge encoded in the network parameters of the vision encoder 130 and the language decoder 140, the majority or all of the learned parameters (weights) are frozen during training of the expert ensemble vision-language system 100.” They are learned through the use of trainable parameters concerning the knowledge of each expert. See also paragraphs 30, 37-40 and 44-47 as it shows learned parameters. Examiner interprets “large language model” as any learning model that learns text data. See paragraphs 64, 134, 135 and 127. See also paragraph 24. ) training the multi-domain visual agent (MDVA) using the semantic information to obtain a trained MDVA. (See paragraphs 19-32, examiner interprets trained MDVA as the trained “expert ensemble vision system” which is multidomain “[0022] The expert ensemble vision-language system 100 receives an input image 105 and multi-modal labels 101, 102, 103, 104, 106, and 107 generated by a variety of pre-trained vision experts (not shown) . The experts perform domain-specific vision tasks , processing the input image 105 to predict depth data, object labels, surface normal vector data, edge data, text labels, or segmentation labels corresponding to the multi-modal labels 101, 102, 103, 104, 106, and 107, respectively. In an embodiment at least two experts process the input image 105 to produce multi-modal labels that are input the expert ensemble vision-language system 100. In an embodiment, additional, fewer, or different experts perform other domain-specific vision tasks to generate multi-modal labels for input to the expert ensemble vision-language system 100.” See also paragraph 24 “0024] In an embodiment, the expert ensemble vision-language system 100 is a visually conditioned autoregressive text generation model , trained to use diverse pre-trained domain experts for open-ended vision-language reasoning tasks . In an embodiment, the expert ensemble vision-language system 100 incorporates powerful vision-only and language-only neural network models for web-scale knowledge to implement a backbone or framework for integration of at least two pre-trained modality-specific vision expert neural networks. The experts encode multiple types of visual information, including low-level vision signals such as depth, surface normal vectors, and edges, and high-level vision signals such as object (instance) labels, segmentation labels, and text labels, as a form of auxiliary knowledge, as the multi-modal labels 101, 102, 103, 104, 106, and 107 …” See also paragraphs 29-32. Anandkumar) , however Anandkumar does not teach “sampling questions from questions templates…”, “generating prompts…” Grenader teaches “sampling questions from questions templates…” (See paragraphs 15 and 64 “[0015] In various embodiments, any combinations of any of the above can be organized to perform any variation of acts for selecting a prompt generation technique based on a corpus of example question embeddings . Many further combinations of aspects of the above elements are contemplated. For example, in addition to the forgoing prompt generation techniques, combinations of any of the above can be organized to implement (1) choosing a large language model interfacing mechanism based on sample question embeddings, (2) generating a large language model prompt based on collaboration activities of a user , and (3) using sample question embeddings when choosing between an LLM interfacing model and a non-LLM interfacing model .” “0064] In some embodiments, it is possible to calculate an embedding Ex for each manually-posed question by asking an LLM model (e.g., ChatGPT4) to classify the question. This can be accomplished by prompting the LLM model with a request in the form of, “ Please classify the question: “{question}” into one of the following categories: “{a sample category list}. ” The sample category list might include names such as “Summarization.” Other possible sample categories may include “Policy”, “Legal”, “Health”, “Architecture”, etc. The LLM will return a classification selected from the provided sample category list. This can be repeated for each manually-posed question until such time as all manually-posed questions have a corresponding embedding as well as a corresponding designated class. ) See also, paragraph 9 “In particular, the herein-disclosed techniques provide technical solutions that address the technical problems that arise when choosing a large language model interfacing mechanism based on a set of candidate sample question embedding vectors ”. The templates are those shown in paragraphs 94-99. ) , Grenader also teaches “generating prompts” (See paragraph 15 “[0015] In various embodiments, any combinations of any of the above can be organized to perform any variation of acts for selecting a prompt generation technique based on a corpus of example question embeddings. Many further combinations of aspects of the above elements are contemplated. For example, in addition to the forgoing prompt generation techniques, combinations of any of the above can be organized to implement (1) choosing a large language model interfacing mechanism based on sample question embeddings, (2) generating a large language model prompt based on collaboration activities of a user, and (3) using sample question embeddings when choosing between an LLM interfacing model and a non-LLM interfacing model.” See also paragraphs 161-163. Grenader) ¸ Grenader also teaches “LLMs” (See paragraphs 3 and 45.) and “natural language description” (See paragraphs 81-82, 98-99 and 93, “[0093]… As shown, this can be done by employing a local instance of natural language processor 206. In some cases, a generative LLM AI entity is prompted to convert local query results into a specified representation (e .g., from one query language syntax into a different query language syntax, and/or to convert from natural language into a specified query language syntax ).” Grenader) It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Anandkumar with the teachings of Grenader to sample questions from templates and generate prompts based on questions and NLDs. The modification would have been motivated by the desire to reduce demand for computer processing power, bandwidth by reducing ambiguity, and also to produce useful answers, therefore it is an improvement, as suggested by Grenader (See paragraph 10 “[0010] Various applications of the herein-disclosed improvements in computer functionality serve to reduce demand for computer memory, reduce demand for computer processing power, reduce network bandwidth usage, and reduce demand for intercomponent communication . For example, when performing computer operations that address the various technical problems underlying how to use data of a content management system to synthesize generative artificial intelligence prompts, both memory usage and CPU cycles demanded are significantly reduced as compared to the memory usage and CPU cycles that would be needed but for practice of the herein-disclosed techniques. This is because the foregoing synthesized generative artificial intelligence prompts are more likely to generate answers useful to the seeker, thus avoiding retries and retries and further retries. ” See also paragraph 46. See also paragraphs 94-99. Grenader ) Claim 8 is rejected under the same analysis as claim 1. Claim 15 is rejected under the same analysis as claim 1. As per claim 2, Anandkumar in view of Grenader teaches “the computer-implemented method of claim 1, further comprising, generating a trajectory for a traffic scene that includes detected objects by the trained MDVA to control a vehicle.” (See paragraph 127, it shows that the system (already detects objects in the image), it is well known that vehicles are normally used in traffic scenes, the system provides navigation assistance. “In at least one embodiment, training data (or data to be analyzed by a trained neural network) can be provided by a provider, a user, or a third party content provider 524. In at least one embodiment , client device 502 may be a vehicle or object that is to be navigated on behalf of a user, for example, which can submit requests and/or receive instructions that assist in navigation of a device .” See paragraph 53 “0053] In an embodiment, at least one of steps 310, 320, and 325 is performed on a server or in a data center to generate the predicted text. In an embodiment, at least one of steps 310, 320, and 325 is performed within a cloud computing environment. In an embodiment, at least one of steps 310, 320, and 325 is performed for training, testing, or certifying a neural network employed in a machine, robot, or autonomous vehicle .” It is well known that an autonomous vehicle can navigate itself through traffic. See also paragraphs 120-124 “[0123]… The second layer assembles the lines to look for higher level patterns such as wheels, windshields, and mirrors . The next layer identifies the type of vehicle, and the final few layers generate a label for the input image, identifying the model of a specific automobile brand .” “0124] Once the DNN is trained, the DNN can be deployed and used to identify and classify objects or patterns in a process known as inference . Examples of inference (the process through which a DNN extracts useful information from a given input) include identifying handwritten numbers on checks deposited into ATM machines, identifying images of friends in photos, delivering movie recommendations to over fifty million users, identifying and classifying different types of automobiles, pedestrians, and road hazards in driverless cars , or translating human speech in real-time.” Anandkumar. ) Claim 9 is rejected under the same analysis as claim 2. Claim 16 is rejected under the same analysis as claim 2. As per claim 3, Anandkumar in view of Grenader teaches “the computer-implemented method of claim 1, further comprising transferring learned knowledge from one domain to another.” (See paragraphs 25 and 45, it shows transfer learning for the domains of the experts. “0045] The adaptor 215 or 245 learns to smoothly transition the pre-trained experts 110 to new tasks and modalities, improving expressivity and conditioning on multi-modal features. The adaptors 215 and 245 have an encoder-decoder design, which has proven to be successful for efficient transfer learning . Combined with a standard cross attention block in the language decoder 130 , the expert ensemble vision-language system 100 is able to smoothly transition from the domain-specific vision-only and language-only backbones to a vision-language model during pre-training with paired image-text data .” See paragraph 61, examiner interprets the ensemble vision language as a bigger domain, and this domains obtains knowledge from all the other domains “[0061] The ensemble vision-language system 100 is a data- and parameter-efficient vision-language model that leverages an ensemble of domain experts. The ensemble vision-language system 100 is capable of generating captions that are semantically coherent and aligned with the visual content of the images. The ensemble vision-language system 100 only requires training of a small number of components, with the majority of network weights inherited from readily-available, pre-trained domain experts, and kept frozen during training. By leveraging experts from a wide range of domains, the ensemble vision-language system 100 can efficiently pool this expert knowledge and adapt it to various vision-language reasoning tasks, achieving competitive performance in image captioning, VQA (visual question answering), and image classification benchmarks.” The expert resampler also counts as a bigger domain that obtains information from several domains as seen in paragraphs 30, 32, and 37-40. Anandkumar) Claim 10 is rejected under the same analysis as claim 3. Claim 17 is rejected under the same analysis as claim 3. As per claim 5, Anankumar in view of Grenader teaches “The computer-implemented method of claim 1, wherein mapping the domain-specific labels further comprises determining similarities between word embeddings of the NLD and the domain-specific labels.” (See paragraphs 20-24, the experts (which contain the NLD) also include the specific labels. See also paragraph 37, it shows a comparison between text embedding and specific labels “In an embodiment, for all multi-modal labels encoding high-level semantic signals, each pixel is tiled with its corresponding text embedding from a pre-trained CLIP (contrastive language-image pre-training) text neural network model, and then PCA (principal component analysis) is applied to down-sample the dimensionality to C=64 for efficient training. In an embodiment, multi- modal labels for object detection are post-processed to tile each pixel with its corresponding multi-modal label parameterized by CLIP text embedding , and then the multi-modal labels for overlapping pixels are further determined by a depth expert . In an embodiment, multi-modal labels for text detection are post-processed to tile each pixel with its corresponding multi-modal label parameterized by CLIP text embedding . In an embodiment, multi-modal labels for depth estimation, surface normal vectors, and edge detection are renormalized to [−1,1].” Examiner also interprets the text embedding with its corresponding label parametrized as performing a detection of similarities, for something to correspond to another, they must match. “[0022] T he expert ensemble vision-language system 100 receives an input image 105 and multi-modal labels 101, 102, 103, 104, 106, and 107 generated by a variety of pre-trained vision experts (not shown). The experts perform domain-specific vision tasks , processing the input image 105 to predict depth data, object labels, surface normal vector data, edge data, text labels, or segmentation labels corresponding to the multi-modal labels 101, 102, 103, 104, 106, and 107, respectively. In an embodiment at least two experts process the input image 105 to produce multi-modal labels that are input the expert ensemble vision-language system In an embodiment, additional, fewer, or different experts perform other domain-specific vision tasks to generate multi-modal labels for input to the expert ensemble vision-language system 100. ” [0024] In an embodiment , the expert ensemble vision-language system 100 is a visually conditioned autoregressive text generation model, trained to use diverse pre-trained domain experts for open-ended vision-language reasoning tasks. In an embodiment, the expert ensemble vision-language system 100 incorporates powerful vision-only and language-only neural network models for web-scale knowledge to implement a backbone or framework for integration of at least two pre-trained modality-specific vision expert neural networks . The experts encode multiple types of visual information, including low-level vision signals such as depth, surface normal vectors, and edges, and high-level vision signals such as object (instance) labels, segmentation labels, and text labels, as a form of auxiliary knowledge, as the multi-modal labels 101, 102, 103, 104, 106, and 107 . See also paragraph 32. Paragraph 52 also shows that parameters applied by mapping, are adjusted to reduce differences. Therefore there is a comparison. “[0052]… In an embodiment, a question about content depicted in the image is input to the language decoder and the predicted text is an answer to the question. In an embodiment, parameters applied by the mapping are adjusted to reduce differences between the predicted text and reference text associated with the image. In an embodiment, the differences are reduced according to a loss function .” See also paragraphs 46, 32 and 52, all parameters utilize a loss function, which has the purpose of comparing, it also shows a likelihood. Anandkumar) Claim 12 is rejected under the same analysis as claim 5. Claim 19 is rejected under the same analysis as claim 5. As per claim 6, Anandkumar in view of Grenader teaches “the computer-implemented method of claim 1, wherein generating the prompts further comprises generating prompt templates based on learned semantic information between each question, NLD, and input data.” (See paragraphs 161-163 “[0161] Now, continuing with this particular embodiment, prompt generation involves use of a prompt template 367 in combination with natural language that is derived from all or portions of the foregoing instances of selected chunks 375. As shown, a natural language processor 206 can be used to identify and/or extract specific portions from the selected chunk s, which extracted specific portions can in turn be used to populate the fields of prompt template 367 . As shown, performance of the operations within prompt generation module 374 result in one or more LLM prompts 327 , which in turn are provided to a selected LLM (not shown) with the expectation that the LLM will return a large language model system answer (e.g., LLM answer 226, as shown) corresponding to the user question.” See also paragraph 168 and 170-176 “As shown, the administrative user develops preconfigured prompts 397 and interacts with a template generation module 390 to generate a prompt template, possibly involving prompt substitution variables, any of which can be defined through use of a user interface (e.g., the user interface 3C00 of FIG. 3C).” See also fig. 3C it shows a generated prompt on the screen. Grenader) Claim 13 is rejected under the same analysis as claim 6. Claim 20 is rejected under the same analysis as claim 6. As per claim 7, Anandkumar in view of Grenader teaches “the computer-implemented method of claim 1, wherein learning the semantic information further comprises verifying the learned semantic information by computing a loss function between generated outputs and ground truth data from the unified label space.” (See paragraph 32, the system uses a loss between outputs (prediction) and the groundtruth. “[0032] In an embodiment, the expert ensemble vision-language system 100 is trained in a supervised manner using a training dataset that includes images (and optional input questions) paired with ground truth text (reference text). A loss function 150 receives the predicted text output by the language decoder 140 and reference text corresponding to the image . The loss function 150 compares the predicted text with the reference text and updates parameters of the experts resampler 120 and adaptors within the vision encoder 130 and the language decoder 140.” See also paragraph 52 “In an embodiment, a question about content depicted in the image is input to the language decoder and the predicted text is an answer to the question. In an embodiment, parameters applied by the mapping are adjusted to reduce differences between the predicted text and reference text associated with the image . In an embodiment, the differences are reduced according to a loss function.” See also paragraph 24 and 30-38. Anandkumar. ) Claim 14 is rejected under the same analysis as claim 7 . 07-21-aia AIA Claim s 4, 11 and 18 are rejected under 35 U.S.C. 103 as being unpatentable over Anandkumar in view of Grenader and further in view of Zhang et. al. (CN 117290487 A) . As per claim 4, Anandkumar in view of Grenader already teaches “the computer-implemented method of claim 1, wherein sampling the questions further comprises determining the questions with sampling rules”, however Anandkumar in view of Grenader does not teach “based on a commonality threshold.” Zhao teaches “determining the questions with sampling rules based on a commonality threshold.” (See pages 8 last paragraph and page 10 first paragraph. Examiner interprets “similarity measure” as “commonality”. “… if the difference type between any new test question and the old test question is pen error and/or language sequence adjustment, the new test question and the old test question form a sample, which is added to the first sample base of the post-processing module aiming at the large language model. otherwise, the new test question and the old test question are respectively vectored, and the similarity of the two vectors is calculated , and the difference degree between the two is measured by the similarity . if the similarity is less than the set threshold value , the sample is composed of the old test question and the new test question and added to the first sample base of the post-processing module aiming at the large language model; if the similarity is greater than or equal to the set threshold value, the sample is composed of the whole examination outline of the volume and the final examination questions , and added to the second sample base aiming at the large language model.” Zhang) It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Anandkumar with the teachings of Grenader and Zhang to determine questions based on sampling rules based on a commonality threshold. The modification would have been motivated by the desire to facilitate distinguishing the sample questions, therefore it is an improvement, as suggested by Zhang ( “ In order to facilitate distinguishing and description, the sample library for the post-processing module is called as the first sample librar y, the sample library for the LLM is called as the second sample library, in a specific embodiment, after each automatic volume assembly, firstly judging the difference type of each pair of new and old test questions, if the difference type between any new test question and the old test question is pen error and/or language sequence adjustment, the new test question and the old test question form a sample, which is added to the first sample base of the post-processing module aiming at the large language model. otherwise, the new test question and the old test question are respectively vectored, and the similarity of the two vectors is calculated, and the difference degree between the two is measured by the similarity . if the similarity is less than the set threshold value, the sample is composed of the old test question and the new test question and added to the first sample base of the post-processing module aiming at the large language model;” Zhang) Claim 11 is rejected under the same analysis as claim 4. Claim 18 is rejected under the same analysis as claim 4. Conclusion Any inquiry concerning this communication or earlier communications from the examiner should be directed to DYLAN J MENDEZ MUNIZ whose telephone number is (703)756-5672. The examiner can normally be reached M-F, 8AM - 5PM ET. 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, Andrew Moyer can be reached at (571) 272-9523. 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. /DYLAN JOHN MENDEZ MUNIZ/Examiner, Art Unit 2675 /ANDREW M MOYER/Supervisory Patent Examiner, Art Unit 2675 Application/Control Number: 18/934,756 Page 2 Art Unit: 2675 Application/Control Number: 18/934,756 Page 3 Art Unit: 2675 Application/Control Number: 18/934,756 Page 4 Art Unit: 2675 Application/Control Number: 18/934,756 Page 5 Art Unit: 2675 Application/Control Number: 18/934,756 Page 6 Art Unit: 2675 Application/Control Number: 18/934,756 Page 7 Art Unit: 2675 Application/Control Number: 18/934,756 Page 8 Art Unit: 2675 Application/Control Number: 18/934,756 Page 9 Art Unit: 2675 Application/Control Number: 18/934,756 Page 10 Art Unit: 2675 Application/Control Number: 18/934,756 Page 11 Art Unit: 2675 Application/Control Number: 18/934,756 Page 12 Art Unit: 2675 Application/Control Number: 18/934,756 Page 13 Art Unit: 2675 Application/Control Number: 18/934,756 Page 14 Art Unit: 2675 Application/Control Number: 18/934,756 Page 15 Art Unit: 2675 Application/Control Number: 18/934,756 Page 16 Art Unit: 2675 Application/Control Number: 18/934,756 Page 17 Art Unit: 2675 Application/Control Number: 18/934,756 Page 18 Art Unit: 2675 Application/Control Number: 18/934,756 Page 19 Art Unit: 2675 Application/Control Number: 18/934,756 Page 20 Art Unit: 2675 Application/Control Number: 18/934,756 Page 21 Art Unit: 2675
Prosecution Insights