DETAILED ACTION
Introduction
This office action is in response to applicant’s claims filed 8/6/2026. Claims 1-7 are currently pending and have been examined. Applicant’s IDS have been considered. There is no claim to foreign priority.
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 .
Election/Restrictions
Claims 8-20 are withdrawn from further consideration pursuant to 37 CFR 1.142(b) as being drawn to a nonelected invention, there being no allowable generic or linking claim. Election was made without traverse in the reply filed on 8/6/2026.
Applicant’s election without traverse of Group 1, claims 1-7 in the reply filed on 8/6/2026 is acknowledged. The applicant’s representative clarified to elected claim in a conversation, 8/25/2026. Claims 7-20 are to be withdrawn.
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 (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.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claim(s) 1, 3, 5 and 7 is/are rejected under 35 U.S.C. 103 as being unpatentable over Elyasi Langarani et al. (Elyasi, US 2026/0147737) in view of Chen (US 2026/0178629) and further in view of Talwar et al. (Talwar, US 2026/0170041).
As per claim 1, Elyasi teaches a system for detecting hallucinations in responses generated by a language model, comprising:
one or more processors (paragraphs [0090, 0091, 0107]-his processors, memory, and instructions);
memory storing instructions that, when executed by the one or more processors, cause the processors to (ibid):
receive, from the language model, a textual response to a natural language query (paragraph [0022-0025, 0028]-his textual query/answer, and correspond LLM);
partition the textual response into factual segments by (ibid-his answer/response split into chunks, the chunks are further fact verified, thus resulting in factual chunks):
(i) determining, for each potential segment boundary, a confidence score (ibid-his chunks, each individually scored, using likelihood, as the confidence score), and
(ii) including the potential segment as a factual segment only when the confidence score exceeds a segmentation threshold and segment length does not exceed a maximum [word] count (ibid-his factual accuracy, verification, likelihood and corresponding threshold, and his size limit, as the segment length not exceeding a maximum word count);
generate, for each factual segment, at least one query for a data repository (ibid-paragraph [0022-0029, 0029-0031]-his verification engine/process, including a query to source and/or source document(s));
retrieve evidence from the data repository based on the queries (ibid-his source verification and links to the verifiable support for answer);
compute a factual support score for each factual segment using a comparison between the segment and the retrieved evidence (ibid-paragraphs [0023-0032]-his exceeding threshold calculated value for the verification of the chunk to the document/evidence, via his “fact checking” and “comparing”, and generating of a verification value as a support score for each factual chunk); and
[identify changes in the factual support score resulting from altering the segmentation threshold or the maximum word count].
Elyasi lacks teaching that which Chen teaches, (ii) including the potential segment as a factual segment only when the confidence score exceeds a segmentation threshold and segment length does not exceed a maximum [word] count (paragraphs [0057, 0073]-his knowledge/answer document segmented into chunks, the chunks having a word count limit).
Thus, it would have been obvious to one of ordinary skill in the linguistics art, before the effective filing date of the invention, as all the claimed elements were known in the prior art and one skilled in the art could have combined the elements as claimed by known methods (computer implemented techniques and algorithms combining processes and steps in natural language processing), in view of the teachings of Elyasi and Chen to combine the prior art element of hallucination detection based on fact verification with respect to a response using chunking and a size limit, as defined as characters, and the like as taught by Elyasi with a particular word count limit as being a size limit as taught by Chen as each element performs the same function as it does separately, as the combination would yield predictable results, KSR International Co. v. Teleflex Inc., 550 US. -- 82 USPQ2nd 1385 (2007), wherein the predictable result would be using a size limit, such as word count, in determining a chunk knowledge source for a response (ibid-see also paragraphs [0006-0011, 0036]-his answer source, chunking and retrieval discussion).
The above combination lacks teaching that which Talwar teaches, identify changes in the factual support score resulting from altering the segmentation threshold or the maximum word count (paragraph [0004, 0042, 0043, 0099]-his varying chunk size, thus changing maximum chunk/word count and/or segmentation threshold, and identifying changes in the score, used as a factuality metric).
Thus, it would have been obvious to one of ordinary skill in the linguistics art, before the effective filing date of the invention, as all the claimed elements were known in the prior art and one skilled in the art could have combined the elements as claimed by known methods (computer implemented techniques and algorithms combining processes and steps in natural language processing), in view of the teachings of Elyasi and Chen and Talwar to combine the prior art element of hallucination detection based on fact verification with respect to a response using chunking and a size limit, as defined as characters, and the like as taught by Elyasi with a particular word count limit as being a size limit as taught by Chen with altering a segmentation threshold or maximum word count as taught by Talwar as each element performs the same function as it does separately, as the combination would yield predictable results, KSR International Co. v. Teleflex Inc., 550 US. -- 82 USPQ2nd 1385 (2007), wherein the predictable result would be using optimizing a size limit, as a factuality metric to optimize response accuracy (ibid-Talwar, see also abstract).
As per claim 3, Elyasi with Chen with Talwar make obvious the system of claim 1, wherein the confidence score is generated by a secondary language model trained to detect factual boundaries (ibid, Elyasi, see also paragraph [0028]-his verification via chunking engine as the secondary language model, and verification as described above with respect to the confidence score).
As per claim 5, Elyasi with Chen with Talwar make obvious the system of claim 1, wherein a factual segment is flagged for additional analysis if the factual support score is less than a predefined threshold (ibid, Elyasi-paragraphs [0022-0025]-his factual accuracy less than a threshold, and corresponding further polishing of the chunks).
As per claim 7, Elyasi with Chen with Talwar make obvious the system of claim 1, wherein output responses generated using different segmentation thresholds are stored together with corresponding factual support scores paragraph [0004, 0008, 0042, 0043, 0065, 0099]-his varying chunk size, thus changing maximum chunk/word count and/or segmentation threshold, and identifying changes in the score, used as a factuality metric, and all results stored in his CRQI, index).
Thus, it would have been obvious to one of ordinary skill in the linguistics art, before the effective filing date of the invention, as all the claimed elements were known in the prior art and one skilled in the art could have combined the elements as claimed by known methods (computer implemented techniques and algorithms combining processes and steps in natural language processing), in view of the teachings of Elyasi and Chen and Talwar to combine the prior art element of hallucination detection based on fact verification with respect to a response using chunking and a size limit, as defined as characters, and the like as taught by Elyasi with a particular word count limit as being a size limit as taught by Chen with altering a segmentation threshold or maximum word count and storing the results in an index, for evaluating a response as taught by Talwar as each element performs the same function as it does separately, as the combination would yield predictable results, KSR International Co. v. Teleflex Inc., 550 US. -- 82 USPQ2nd 1385 (2007), wherein the predictable result would be using optimizing a size limit, as a factuality metric to optimize response accuracy (ibid-Talwar, see also abstract).
Claim(s) 4, 6 is/are rejected under 35 U.S.C. 103 as being unpatentable over
Elyasi Langarani et al. (Elyasi, US 2026/0147737) in view of Chen (US 2026/0178629) and further in view of Talwar et al. (Talwar, US 2026/0170041), as applied to claim 1, and further in view of Bax et al. (Bax, US 2025/0005266).
As per claim 4, Elyasi with Chen with Talwar make obvious the system of claim 1, but lack explicitly teaching that which Bax teaches, wherein the comparison between the segment and retrieved evidence comprises a cosine similarity between vector embeddings (paragraph [0085]-his cosine similarity comparison between support statement and response).
Thus, it would have been obvious to one of ordinary skill in the linguistics art, before the effective filing date of the invention, as all the claimed elements were known in the prior art and one skilled in the art could have combined the elements as claimed by known methods (computer implemented techniques and algorithms combining processes and steps in natural language processing), in view of the teachings of Elyasi and Chen and Talwar and Bax to combine the prior art element of hallucination detection based on fact verification with respect to a response and comparison for support as taught by Elyasi with using a similarity calculation function for the comparison, cosine similarity, as taught by Bax, as each element performs the same function as it does separately, as the combination would yield predictable results, KSR International Co. v. Teleflex Inc., 550 US. -- 82 USPQ2nd 1385 (2007), wherein the predictable result would be using a similarity function for determining a score for a segment used in determining trustworthiness of a response (ibid-Bax, see also abstract).
As per claim 6, Elyasi with Chen with Talwar make obvious the system of claim 1, but lacks explicitly teaching that which Bax teaches, wherein the data repository comprises ground truth annotated data for factual verification (paragraphs [0009, 0082]-is labeled indexed documents and data sources as trustworthy, used to verify responses, as the ground truth annotated data for factual verification).
Thus, it would have been obvious to one of ordinary skill in the linguistics art, before the effective filing date of the invention, as all the claimed elements were known in the prior art and one skilled in the art could have combined the elements as claimed by known methods (computer implemented techniques and algorithms combining processes and steps in natural language processing), in view of the teachings of Elyasi and Chen and Talwar and Bax to combine the prior art element of hallucination detection based on fact verification with respect to a response and comparison for support as taught by Elyasi with using a similarity calculation function for the comparison, cosine similarity, using ground truth as verified and labeled or annotated trustworthy repository data for verification as taught by Bax, as each element performs the same function as it does separately, as the combination would yield predictable results, KSR International Co. v. Teleflex Inc., 550 US. -- 82 USPQ2nd 1385 (2007), wherein the predictable result would be using a similarity function for determining a score for a segment used in determining trustworthiness of a response (ibid-Bax, see also abstract).
Claim(s) 2 is/are rejected under 35 U.S.C. 103 as being unpatentable over
Elyasi Langarani et al. (Elyasi, US 2026/0147737) in view of Chen (US 2026/0178629) and further in view of Talwar et al. (Talwar, US 2026/0170041), as applied to claim 1, and further in view of Othene-Frempong et al. (Othene, US 2021/0035696).
As per claim 2, Elyasi with Chen with Talwar make obvious the system of claim 1, but lack explicitly teaching that which Othene-Frempong et al. wherein the maximum word count is fifteen words (paragraph [0110]-his fifteen word limit).
Thus, it would have been obvious to one of ordinary skill in the linguistics art, before the effective filing date of the invention, as all the claimed elements were known in the prior art and one skilled in the art could have combined the elements as claimed by known methods (computer implemented techniques and algorithms combining processes and steps in natural language processing), in view of the teachings of Elyasi and Chen and Talwar and Othene to combine the prior art element of hallucination detection based on fact verification with respect to a response using chunking and a size limit, as defined as characters, and the like as taught by Elyasi with a particular word count limit as being a size limit as taught by Chen with altering a segmentation threshold or maximum word count as taught by Talwar with having a particular word limit, such as fifteen, for a word length metric as each element performs the same function as it does separately, as the combination would yield predictable results, KSR International Co. v. Teleflex Inc., 550 US. -- 82 USPQ2nd 1385 (2007), wherein the predictable result would be using optimizing a size limit, as a factuality metric to optimize response accuracy, (ibid-Talwar, see also abstract).
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure (See PTO-892).
Any inquiry concerning this communication or earlier communications from the examiner should be directed to LAMONT M SPOONER whose telephone number is (571)272-7613. The examiner can normally be reached 8:00 AM -5:00 PM.
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, Daniel Washburn can be reached at (571)272-5551. 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.
/LAMONT M SPOONER/ Primary Examiner, Art Unit 2657
8/30/2026
Prosecution Insights