Pith. sign in

REVIEW 27 cited by

Robust Distortion-free Watermarks for Language Models

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2307.15593 v3 pith:PMFQUSM6 submitted 2023-07-28 cs.LG cs.CLcs.CR

Robust Distortion-free Watermarks for Language Models

classification cs.LG cs.CLcs.CR
keywords textlanguagemodelmodelsrandomresponsesrobustsampling
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
read the original abstract

We propose a methodology for planting watermarks in text from an autoregressive language model that are robust to perturbations without changing the distribution over text up to a certain maximum generation budget. We generate watermarked text by mapping a sequence of random numbers -- which we compute using a randomized watermark key -- to a sample from the language model. To detect watermarked text, any party who knows the key can align the text to the random number sequence. We instantiate our watermark methodology with two sampling schemes: inverse transform sampling and exponential minimum sampling. We apply these watermarks to three language models -- OPT-1.3B, LLaMA-7B and Alpaca-7B -- to experimentally validate their statistical power and robustness to various paraphrasing attacks. Notably, for both the OPT-1.3B and LLaMA-7B models, we find we can reliably detect watermarked text ($p \leq 0.01$) from $35$ tokens even after corrupting between $40$-$50\%$ of the tokens via random edits (i.e., substitutions, insertions or deletions). For the Alpaca-7B model, we conduct a case study on the feasibility of watermarking responses to typical user instructions. Due to the lower entropy of the responses, detection is more difficult: around $25\%$ of the responses -- whose median length is around $100$ tokens -- are detectable with $p \leq 0.01$, and the watermark is also less robust to certain automated paraphrasing attacks we implement.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 27 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. SLAM: Structural Linguistic Activation Marking for Language Models

    cs.CL 2026-05 unverdicted novelty 8.0

    SLAM achieves 100% detection on Gemma-2 models with only 1-2 point quality cost by causally steering SAE-identified residual-stream directions for linguistic structure.

  2. SLAM: Structural Linguistic Activation Marking for Language Models

    cs.CL 2026-05 unverdicted novelty 8.0

    SLAM achieves 100% detection accuracy on Gemma-2 models with only 1-2 points of quality loss by causally steering SAE-identified structural directions while preserving lexical sampling and semantics.

  3. Undetectable Conversations Between AI Agents via Pseudorandom Noise-Resilient Key Exchange

    cs.CR 2026-04 unverdicted novelty 8.0

    AI agents can conduct undetectable covert conversations using a new pseudorandom noise-resilient key exchange that works without shared keys and with only constant min-entropy in messages.

  4. RLCracker: Evaluating the Worst-Case Vulnerability of LLM Watermarks with Adaptive RL Attacks

    cs.CR 2025-09 conditional novelty 8.0

    RLCracker is a reinforcement learning attack that erases LLM watermarks at 98.5% success rate with minimal data and generalizes across ten schemes and multiple model sizes.

  5. TRACE: A Two-Channel Robust Attribution Watermark via Complementary Embeddings for LLM-Agent Trajectories

    cs.CR 2026-07 accept novelty 7.0

    TRACE is a two-channel, distortion-free agent watermark whose selection and tally layers jointly resist deletion and rewriting by a log-holding reseller.

  6. Beyond Heuristic Tuning: Power-Calibrated LLM Watermarking

    stat.ML 2026-07 accept novelty 7.0

    A power-calibrated statistical framework gives closed-form links from KGW watermark parameters (γ, δ) to detection power and KL distortion, enabling principled Pareto-optimal selection.

  7. Signature filtering: a lightweight enhancement for statistical watermark detection in large language models

    cs.LG 2026-06 conditional novelty 7.0

    Signature filtering learns unreliable tokens with MILP and removes them at detection time, raising true positive rates from 8-31% to 78-99% across Kgw, Sweet, Unigram, and Exp watermarks on multiple corpora and LLMs w...

  8. Unsupervised Style Representation Learning for AI-Text Detection via Paraphrase Inversion

    cs.LG 2026-06 unverdicted novelty 7.0

    Unsupervised style representations learned via paraphrase inversion enable competitive few-shot and zero-shot AI-text detection with better generalization to unseen LLMs than supervised baselines.

  9. SWAN: Semantic Watermarking with Abstract Meaning Representation

    cs.CL 2026-05 unverdicted novelty 7.0

    SWAN uses AMR to embed semantic watermarks that persist through paraphrases, matching SOTA detection on original text and improving AUC by 13.9 points on paraphrased RealNews data.

  10. Can we Watermark Low-Entropy LLM Outputs?

    cs.CR 2026-04 unverdicted novelty 7.0

    The authors give constructions for provably undetectable watermarking of constant-entropy LLM outputs that are robust to random substitutions (under subexponential LPN) and to substitutions plus random deletions (unde...

  11. RLSpoofer: A Lightweight Evaluator for LLM Watermark Spoofing Resilience

    cs.CR 2026-04 unverdicted novelty 7.0

    RLSpoofer trains a 4B model on 100 watermarked paraphrase pairs to spoof PF watermarks at 62% success rate, far exceeding baselines trained on up to 10,000 samples.

  12. Optimal Multi-bit Generative Watermarking Schemes Under Worst-Case False-Alarm Constraints

    cs.IT 2026-04 unverdicted novelty 7.0

    Two new constructions for multi-bit generative watermarking attain the established lower bound on miss-detection probability under worst-case false-alarm constraints, fully characterizing optimal performance via linea...

  13. Copyright Protection for Large Language Models: A Survey of Methods, Challenges, and Trends

    cs.CR 2025-08 accept novelty 7.0

    A survey of LLM copyright protection that unifies text watermarking, model watermarking, and model fingerprinting while presenting new coverage of fingerprint transfer and removal.

  14. Topic-Based Watermarks for Large Language Models

    cs.CR 2024-04 unverdicted novelty 7.0

    A topic-guided watermarking scheme partitions the LLM vocabulary into topic-aligned token subsets and green-lists relevant tokens based on the input prompt to embed detectable marks while preserving text quality and i...

  15. SAMark: A Self-Anchored Text Watermarking with Paragraph-Level Paraphrase Robustness

    cs.CR 2026-05 unverdicted novelty 6.0

    SAMark uses self-anchored semantic green regions, multi-channel hyperbolic scoring, and diversity-aware filtering to reach 90.2% TP@FP1% detection under paragraph paraphrasing while preserving text quality.

  16. Trustworthy AI: Ensuring Reliability and Accountability from Models to Agents

    cs.LG 2026-05 unverdicted novelty 6.0

    The thesis presents a kernel method for multiaccuracy across overlooked subpopulations, information-theoretic optimal watermarking for LLMs, and a simulator showing LLM agents outperforming humans in supply chains whi...

  17. Response Time Enhances Alignment with Heterogeneous Preferences

    cs.LG 2026-05 unverdicted novelty 6.0

    Response times modeled as drift-diffusion processes enable consistent estimation of population-average preferences from heterogeneous anonymous binary choices.

  18. Detecting Verbatim LLM Copy-Paste in Homework

    cs.CR 2026-05 unverdicted novelty 6.0

    SteganoPrompt embeds a hidden instruction in assignment prompts via the Unicode Tags block so that LLMs add a detectable signature to responses when the prompt is pasted verbatim.

  19. Block-wise Codeword Embedding for Reliable Multi-bit Text Watermarking

    cs.CR 2026-05 unverdicted novelty 6.0

    BREW uses block voting and window-shifting verification to reach TPR 0.965 and FPR 0.02 under 10% synonym substitution, addressing high false-positive issues in prior multi-bit LLM watermarking.

  20. Block-wise Codeword Embedding for Reliable Multi-bit Text Watermarking

    cs.CR 2026-05 unverdicted novelty 6.0

    BREW achieves TPR of 0.965 and FPR of 0.02 under 10% synonym substitution by shifting from ECC decoding to designated verification with block voting and local validation.

  21. Towards Robust Content Watermarking Against Removal and Forgery Attacks

    cs.CV 2026-04 unverdicted novelty 6.0

    ISTS watermarking dynamically controls injection based on prompt semantics and uses two-sided detection to resist removal and forgery attacks in diffusion models.

  22. ArcMark: Distortion-Free Multi-Byte LLM Watermark via Optimal Transport

    cs.LG 2026-02 unverdicted novelty 6.0

    ArcMark is a multi-byte LLM watermark that achieves distortion-free embedding of several bytes per few hundred tokens by treating generation as a channel coding problem and using optimal transport to match distributions.

  23. SWaRL: Safeguard Code Watermarking via Reinforcement Learning

    cs.CR 2026-01 unverdicted novelty 6.0

    SWaRL trains code LLMs with RL using compiler correctness signals and a confidential verifier reward to embed robust, functionality-preserving watermarks that resist refactoring attacks.

  24. Can AI-Generated Text be Reliably Detected?

    cs.CL 2023-03 unverdicted novelty 6.0

    Recursive paraphrasing attacks substantially lower detection rates for multiple AI text detectors with only minor quality loss, while a theoretical analysis ties best-case AUROC to total variation distance between hum...

  25. Verifiable Provenance and Watermarking for Generative AI: An Evidentiary Framework for International Operational Law and Domestic Courts

    cs.CR 2026-05 unverdicted novelty 5.0

    Develops a five-tier threat model for generative AI content, releases a 12000-item multi-modal benchmark with laundering tests, evaluates four schemes, and maps detection metrics to legal sufficiency thresholds for la...

  26. Fundamental Trade-Offs in Multi-Bit Watermarking of Stochastic Processes

    cs.IT 2026-05 unverdicted novelty 5.0

    Derives matched converse and achievability bounds that characterize optimal trade-offs among false-alarm probability, detection error probability, distortion, and information rate for multi-bit watermarking of station...

  27. Position: LLM Watermarking Should Align Stakeholders' Incentives for Practical Adoption

    cs.CR 2025-10 unverdicted novelty 4.0

    LLM watermarking adoption is limited by misaligned stakeholder incentives; incentive-aligned approaches such as in-context watermarking can enable practical use in targeted domains like education and peer review.