Prosecution Insights
Last updated: November 20, 2026
Application No. 19/088,888

REMARK-LLM: A ROBUST AND EFFICIENT WATERMARKING FRAMEWORK FOR GENERATIVE LARGE LANGUAGE MODELS

Non-Final OA §101§103
Filed
Mar 24, 2025
Priority
Mar 22, 2024 — provisional 63/568,643
Examiner
REN, ZHUBING
Art Unit
2658
Tech Center
2600 — Communications
Assignee
The Regents of the University of California
OA Round
1 (Non-Final)
72%
Grant Probability
Favorable
1-2
OA Rounds
1y 4m
Est. Remaining
99%
With Interview

Examiner Intelligence

Grants 72% — above average
72%
Career Allowance Rate
293 granted / 407 resolved
+10.0% vs TC avg
Strong +42% interview lift
Without
With
+41.7%
Interview Lift
resolved cases with interview
Typical timeline
3y 0m
Avg Prosecution
20 currently pending
Career history
438
Total Applications
across all art units

Statute-Specific Performance

§101
6.5%
-33.5% vs TC avg
§103
72.5%
+32.5% vs TC avg
§102
9.5%
-30.5% vs TC avg
§112
2.7%
-37.3% vs TC avg
Black line = Tech Center average estimate • Based on career data from 407 resolved cases

Office Action

§101 §103
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 . DETAIL ACTION Priority This application claims priority to U.S provisional Patent Application No. 63568643, filed on 3/22/2024 and is hereby incorporated by references. 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 U.S.C. 101 because the claimed invention is directed to non-statutory subject matter. The claimed invention is directed to non-statutory subject matter because the claim(s) as a whole, considering all claim elements both individually and in combination, do not amount to significantly more than an abstract idea. As summarized in the 2019 Revised Patent Subject Matter Eligibility Guidance, examiners must perform a Two-Part Analysis for Judicial Exceptions. Step 1 In Step 1, it must be determined whether the claimed invention is directed to a process, machine, manufacture or composition of matter. The instant invention encompasses three sets of claims: a system in claims 1-9 (i.e., a manufacture), a method in claims 10-18 (i.e., a process) and a non-transitory machine-readable medium in claims 19-20 (i.e., a manufacture). All claims are directed to one of the four statutory categories and meet the requirements of step 1. Step 2A Prong One The claimed invention is directed to an abstract idea without significant more. The instant invention is broadly directed to “generating a perturbed watermarked text distribution from an output text sequence from a trained large language model”. Claim 1 recites the following (with emphasis added): Claim 1: A system comprising: at least one processor; and at least one memory including instructions which when executed by the at least one processor causes operations comprising: receiving an output text sequence from a trained large language model; converting the output text sequence into a token representation of the output text sequence; generating a dense watermarked text distribution over a token vocabulary of the output text sequence, the generating based on the token representation of the output text sequence and on a binary signature sequence; perturbing the dense watermarked text distribution to yield a perturbed distribution; and mapping the perturbed distribution to an encoded output text sequence. The bold portions of claim 1 encompass the abstract idea, which is also encompassed by the dependent claims 2-9, and substantially also encompassed by claims 10-16 and 17-20. Claims 1, 10, and 12 recite the steps to generate a perturbed watermarked text distribution from an output text sequence from a trained large language model including a natural language processing. These limitations, when given their broadest reasonable interpretation, are directed to certain performing of organizing human activity and mental processes, which is abstract idea. Prong Two This judicial exception is not integrated into a practical application because mere instruction to implement on computers (i.e. storage medium or processors in claim 1) or a computer model (language model here in claim 1), or merely using computers as a tool to perform the abstract idea, adding insignificant extra solution activity, and/or generally linking the use of the abstract idea to a technological environment for field of use is not considered integration into a practical application. Claim 1 recites using trained language model to generate text output data then generating a perturbed watermarked text distribution. Using language model to generate text output data then generating a perturbed watermarked text distribution is a generic feature of natural language process, which does not represent a technological improvement. The using of the computer and natural language process does not add improvement to the functioning of a computer or to any other technology field, which failed to enable the abstract idea to integrate into a practical application. The claims are drafted in a result-oriented fashion, without the requisite specificity needed to provide a nonabstract technological solution. The computing system and natural language process are directed to the components of a system amount to merely field of use type limitations and/or extra solution activity to implement the abstract idea as presented. Step 2B Step 2B in the analysis requires us to determine whether the claims do significantly more than simply describe that abstract method. Mayo, 132 S. Ct. at 1297. We must examine the limitations of the claims to determine whether the claims contain an "inventive concept" to "transform" the claimed abstract idea into patent-eligible subject matter. Alice, 134 S. Ct. at 2357 (quoting Mayo, 132 S. Ct. at 1294, 1298). The transformation of an abstract idea into patent-eligible subject matter "requires 'more than simply stat[ing] the [abstract idea] while adding the words 'apply it."' Id. (quoting Mayo, 132 S. Ct. at 1294) (alterations in original). "A claim that recites an abstract idea must include 'additional features' to ensure 'that the [claim] is more than a drafting effort designed to monopolize the [abstract idea].'" Id. (quoting Mayo, 132 S. Ct. at 1297) (alterations in original). Those "additional features" must be more than "well-understood, routine, conventional activity." Mayo, 132 S. Ct. at 1298. The present claims include the additional elements other than the abstract idea which include a processor, storage medium, trained large language model (in claim 1). These additional elements are merely conventional computer and computer model. Any potentially technical aspects of the claims are well-known generic computer components performing conventional functions (e.g., a processor performing a mental process). The present claims have been analyzed both individually and in combination and, the instant claims do not provide any improvement of the functioning of the computer or improvement to computer technology or any other technical field. There do not appear to be any meaningful limitations other than those that are well-understood, routine and conventional in the field. Thus, the present claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception. Thus, the claims 1-9 are not patent eligible. Claims 10-18 and method claim 19-20 recite similar limitations of claims 1-9, thus are abstract idea and not patent eligible. 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) 1-5, 10-14 and 19-20 is/are rejected under 35 U.S.C. 103 as being unpatentable over LIU et al (CN 116956906 A) in view of Dathathri et al (US 20240320529 A1) and MELLOR et al (US 20250005115 A1). Regarding claim 1, LIU discloses a system [e.g. FIG. 1-2 and 10; text generation device] comprising: at least one processor [e.g. FIG. 10; page 4; processor]; and at least one memory [e.g. memory] including instructions [e.g. executable instructions] which when executed by the at least one processor causes operations [e.g. FIG. 1-2; text generation operation] comprising: receiving an output text sequence from a trained large language model [e.g. FIG. 1-2; page 8-9; obtain text from LLM]; converting the output text sequence into a token representation of the output text sequence [e.g. FIG. 1-2 and 4; obtaining token sequence (word)]; LIU fails to explicitly disclose the detail of the token representation. However, Dathathri teaches the well-known concept of generating a dense watermarked text distribution over a token vocabulary of the output text sequence [e.g. FIG. 1; watermarked token sequence], the generating based on the token representation of the output text sequence and on a binary signature sequence [e.g. FIG. 5; token sequence with binary values 1, 0, 0, 1]; and mapping the distribution to an encoded output text sequence [e.g. FIG. 1; encoder to encode the input token sequence]. It would have been obvious to a person of ordinary skill in the art before the effective filing date of the invention to modify the text generation system disclosed by LIU to exploit the well-known generating watermarked token sequence technique taught by Dathathri as above, in order to provide robust and detectable watermarking to be added to a digital object with relatively low computational overhead (depending on the number of watermarking stages applied) [See Dathathri; [0045]]. Moreover, MELLOR teaches the well-known concept of perturbing the dense watermarked text distribution to yield a perturbed distribution [e.g. FIG. 18 and 29-31; watermark token with noise]; and mapping the perturbed distribution to an output text sequence [e.g. FIG. 18 and 27; mapping watermark token]. It would have been obvious to a person of ordinary skill in the art before the effective filing date of the invention to modify the text generation system disclosed by LIU to exploit the well-known generating watermarked token sequence technique taught by Dathathri and the week-known concept of embedding a digital watermark within tokenised data taught by MELLOR as above, in order to provide robust and detectable watermarking to be added to a digital object with relatively low computational overhead (depending on the number of watermarking stages applied) [See Dathathri; [0045]] and a solution that uses deterministic tokenization, the process can also achieve lower latency [See MELLOR; [0012]]. Regarding claim 2, LIU, Dathathri and MELLOR further disclose the receiving, the converting, and the generating are caused to be performed by an encoding module [e.g. LIU: FIG 1 and 10, Dathathri: FIG. 1; MELLOR: FIG. ; [0298]], and wherein the perturbing and the mapping are caused to be performed by an optimization beam search module [e.g. LIU: FIG 1 and 10, Dathathri: FIG. 1]. It would have been obvious to a person of ordinary skill in the art before the effective filing date of the invention to modify the text generation system disclosed by LIU to exploit the well-known generating watermarked token sequence technique taught by Dathathri and the week-known concept of embedding a digital watermark within tokenised data taught by MELLOR as above, in order to provide robust and detectable watermarking to be added to a digital object with relatively low computational overhead (depending on the number of watermarking stages applied) [See Dathathri; [0045]] and a solution that uses deterministic tokenization, the process can also achieve lower latency [See MELLOR; [0012]]. Regarding claim 3, LIU, Dathathri and MELLOR further disclose the token representation of the output text sequence comprises a plurality of tokens [e.g. LIU: FIG 2-3 and 6, Dathathri: FIG. 1 and 5]; each of the plurality of tokens representing a corresponding portion of text of the output text sequence [e.g. LIU: FIG 2-3 and 6; Dathathri: FIG. 1 and 5; MELLOR: tokenised data]. It would have been obvious to a person of ordinary skill in the art before the effective filing date of the invention to modify the text generation system disclosed by LIU to exploit the well-known generating watermarked token sequence technique taught by Dathathri and the week-known concept of embedding a digital watermark within tokenised data taught by MELLOR as above, in order to provide robust and detectable watermarking to be added to a digital object with relatively low computational overhead (depending on the number of watermarking stages applied) [See Dathathri; [0045]] and a solution that uses deterministic tokenization, the process can also achieve lower latency [See MELLOR; [0012]]. Regarding claim 4, LIU, Dathathri and MELLOR further disclose for each of the plurality of tokens, an associated probability indicative of how the corresponding portion of text maps to the encoded output text sequence [e.g. LIU: FIG 2-3 and 6; sequence of the probability value corresponding to the prediction output text; Dathathri: FIG. 1 and 5; The watermarking involves modifying a probability distribution of the token MELLOR: tokenised data; probability of segment]. It would have been obvious to a person of ordinary skill in the art before the effective filing date of the invention to modify the text generation system disclosed by LIU to exploit the well-known generating watermarked token sequence technique taught by Dathathri and the week-known concept of embedding a digital watermark within tokenised data taught by MELLOR as above, in order to provide robust and detectable watermarking to be added to a digital object with relatively low computational overhead (depending on the number of watermarking stages applied) [See Dathathri; [0045]] and a solution that uses deterministic tokenization, the process can also achieve lower latency [See MELLOR; [0012]]. Regarding claim 5, LIU, Dathathri and MELLOR further disclose adding noise to the dense watermarked text distribution [e.g. MELLOR: FIG. 18 and 29-31; watermark token with noise]. It would have been obvious to a person of ordinary skill in the art before the effective filing date of the invention to modify the text generation system disclosed by LIU to exploit the well-known generating watermarked token sequence technique taught by Dathathri and the week-known concept of embedding a digital watermark within tokenised data taught by MELLOR as above, in order to provide robust and detectable watermarking to be added to a digital object with relatively low computational overhead (depending on the number of watermarking stages applied) [See Dathathri; [0045]] and a solution that uses deterministic tokenization, the process can also achieve lower latency [See MELLOR; [0012]]. Regarding claim 10-14, this is a computer-implemented method that includes same limitation as in claim 1-5 above, the rejection of which are incorporated herein. Regarding claim 19-20, this is a non-transitory computer-readable storage medium that includes same limitation as in claim 1-2 above, the rejection of which are incorporated herein. Claim(s) 6 and 15 is/are rejected under 35 U.S.C. 103 as being unpatentable over LIU et al (CN 116956906 A) in view of Dathathri et al (US 20240320529 A1), MELLOR et al (US 20250005115 A1) and LI et al (US 20210241099 A1). Regarding claim 6, LIU, Dathathri and MELLOR further disclose adding noise to the dense watermarked text distribution [e.g. MELLOR: FIG. 18 and 29-31; watermark token with noise], but LIU, Dathathri and MELLOR fail to explicitly disclose the detail of the noise. However, LI teaches the well-known concept of the noise comprises Gumbel-Softmax noise [e.g. [0042-0044]; introduce Gumbel-Softmax noise]. It would have been obvious to a person of ordinary skill in the art before the effective filing date of the invention to modify the text generation system disclosed by LIU to exploit the well-known generating watermarked token sequence technique taught by Dathathri, the week-known concept of embedding a digital watermark within tokenised data taught by MELLOR and the well-known concept of modeling the noise as Gumbel distribution technique taught by LI as above, in order to provide robust and detectable watermarking to be added to a digital object with relatively low computational overhead (depending on the number of watermarking stages applied) [See Dathathri; [0045]], a solution that uses deterministic tokenization, the process can also achieve lower latency [See MELLOR; [0012]] and the effect of sampling as introducing noise to the input so that the outputs become continuous and differentiable [See LI; [0042]]. Regarding claim 15, this is a computer-implemented method that includes same limitation as in claim 6 above, the rejection of which are incorporated herein. Claim(s) 7-8 and 16-17 is/are rejected under 35 U.S.C. 103 as being unpatentable over LIU et al (CN 116956906 A) in view of Dathathri et al (US 20240320529 A1), MELLOR et al (US 20250005115 A1) and BELLEGARDA (US 20230394248 A1). Regarding claim 7, LIU, Dathathri and MELLOR further disclose reparametrizing the dense watermarked text distribution over the token vocabulary of the output text sequence [e.g. LIU: FIG 2-3 and 6; Dathathri: FIG. 1 and 5; MELLOR: FIG. 18 and 30-31; tokenised data], but LIU, Dathathri and MELLOR fails to explicitly disclose the output text sequence to a yield a sparse distribution. However, 7. BELLEGARDA teaches the well-known concept of reparametrizing the text distribution over the token vocabulary of the output text sequence to a yield a sparse distribution [e.g. FIG. 1 and 8; output token probability distribution may be a sparse output token probability distribution]. It would have been obvious to a person of ordinary skill in the art before the effective filing date of the invention to modify the text generation system disclosed by LIU to exploit the well-known generating watermarked token sequence technique taught by Dathathri, the week-known concept of embedding a digital watermark within tokenised data taught by MELLOR and the well-known concept of generating token probability distribution technique taught by BELLEGARDA as above, in order to provide robust and detectable watermarking to be added to a digital object with relatively low computational overhead (depending on the number of watermarking stages applied) [See Dathathri; [0045]], a solution that uses deterministic tokenization, the process can also achieve lower latency [See MELLOR; [0012]] and more accurate and useful text prediction for the specific user [See BELLEGARDA ; [0005]]. Regarding claim 8, LIU, Dathathri, MELLOR and BELLEGARDA further disclose 8. The system of claim 7, wherein a reparameterization module causes the reparametrizing [e.g. LIU: FIG 2-3 and 6; Dathathri: FIG. 1 and 5; MELLOR: FIG. 18 and 30-31; BELLEGARDA : FIG. 1 and 8]. It would have been obvious to a person of ordinary skill in the art before the effective filing date of the invention to modify the text generation system disclosed by LIU to exploit the well-known generating watermarked token sequence technique taught by Dathathri, the week-known concept of embedding a digital watermark within tokenised data taught by MELLOR and the well-known concept of generating token probability distribution technique taught by BELLEGARDA as above, in order to provide robust and detectable watermarking to be added to a digital object with relatively low computational overhead (depending on the number of watermarking stages applied) [See Dathathri; [0045]], a solution that uses deterministic tokenization, the process can also achieve lower latency [See MELLOR; [0012]] and more accurate and useful text prediction for the specific user [See BELLEGARDA ; [0005]]. Regarding claim 16-17, this is a computer-implemented method that includes same limitation as in claim 7-8 above, the rejection of which are incorporated herein. Claim(s) 9 and 18 is/are rejected under 35 U.S.C. 103 as being unpatentable over LIU et al (CN 116956906 A) in view of Dathathri et al (US 20240320529 A1), MELLOR et al (US 20250005115 A1) and Paul (US 20240236076 A1). Regarding claim 9, LIU, Dathathri and MELLOR further disclose decoding the encoded output text sequence [e.g. LIU: FIG 2-3 and 6; Dathathri: FIG. 1 and 5; MELLOR: FIG. 18 and 30-31; tokenised data], but LIU, Dathathri and MELLOR fails to explicitly disclose enable ownership verification of the output text sequence. However, Paul teaches the well-known concept of the output text sequence to enable ownership verification of the output text sequence [e.g. FIG. 1 and 7; [0130]; the domain authentication can be visually watermarked into the NFT in the case of images and videos via plain text]. It would have been obvious to a person of ordinary skill in the art before the effective filing date of the invention to modify the text generation system disclosed by LIU to exploit the well-known generating watermarked token sequence technique taught by Dathathri, the week-known concept of embedding a digital watermark within tokenised data taught by MELLOR and the well-known concept of generating domain authentication technique taught by Paul as above, in order to provide robust and detectable watermarking to be added to a digital object with relatively low computational overhead (depending on the number of watermarking stages applied) [See Dathathri; [0045]], a solution that uses deterministic tokenization, the process can also achieve lower latency [See MELLOR; [0012]] and more accurate and a secured website and storage [See Paul; [0130-1032]]. Regarding claim 18, this is a computer-implemented method that includes same limitation as in claim 9 above, the rejection of which are incorporated herein. Conclusion The prior art made of record and not relied upon is considered pertinent to applicant's disclosure. MELLOR et al (US 20260252668 A1). COLLADO UMANA et al (US 20250272089 A1). Any inquiry concerning this communication or earlier communications from the examiner should be directed to ZHUBING REN whose telephone number is (571)272-2788. The examiner can normally be reached Monday-Friday 9am-5pm. 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, Richemond Dorvil can be reached at 571-272-7602. 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. /ZHUBING REN/Primary Examiner, Art Unit 2658
Read full office action

Prosecution Timeline

Mar 24, 2025
Application Filed
Oct 30, 2026
Non-Final Rejection mailed — §101, §103 (current)

Precedent Cases

Applications granted by this same examiner with similar technology

Patent 12701253
BLOCK BASED WEIGHTING FACTOR FOR JOINT MOTION VECTOR DIFFERENCE CODING MODE
1y 5m to grant Granted Aug 04, 2026
Patent 12689746
SYSTEMS AND METHODS FOR EXTENDED MULTI-RESIDUE BLOCK CODING
2y 1m to grant Granted Jul 21, 2026
Patent 12659490
IN-LOOP FILTERING METHOD AND APPARATUS FOR SAME
1y 5m to grant Granted Jun 16, 2026
Patent 12651499
USER ACCOUNTABILITY RESTROOM
2y 7m to grant Granted Jun 09, 2026
Patent 12647308
METHOD FOR TRANSMITTING ORTHOGONAL FREQUENCY DIVISION MULTIPLEXING SIGNAL ON BASIS OF RIS IN WIRELESS COMMUNICATION SYSTEM, AND DEVICE THEREFOR
2y 4m to grant Granted Jun 02, 2026
Study what changed to get past this examiner. Based on 5 most recent grants.

Strategy Recommendation AI-generated — please review before filing

Get a prosecution strategy drawn from examiner precedents, rejection analysis, and claim mapping.
Typically takes 5-10 seconds — AI-generated, attorney review required before filing

Prosecution Projections

1-2
Expected OA Rounds
72%
Grant Probability
99%
With Interview (+41.7%)
3y 0m (~1y 4m remaining)
Median Time to Grant
Low
PTA Risk
Based on 407 resolved cases by this examiner. Grant probability derived from career allowance rate.

Sign in with your work email

Enter your email to receive a magic link. No password needed.

Personal email addresses (Gmail, Yahoo, etc.) are not accepted.

Free tier: 3 strategy analyses per month