InfoResearchPreprintLLM-specific
Watermark Evidence under Open-Weight LLM Modification: Exact Bounds for Low-Rank Output Watermarks
- Published
- Record updated
Summary
This paper studies how modifying open-weight language models changes the evidence available to an OpenStamp-style watermark verifier. The authors derive exact bounds on expected log-likelihood-ratio evidence under a residual-span constraint, plus a high-confidence bound on miss probability for sequence-level detection. In experiments on Llama-2-7B and Mistral-7B-v0.3, INT8 quantization and 30% pruning kept test true-positive rates at 85.6-94.8%, while 500-step LoRA fine-tuning reduced them to 28.8% and 11.6%.
Related items
- InfoConfGuard: a physics-grounded reference monitor for predictive-maintenance large-language-model agentsSimilar attack · Arxiv (cs.CR + cs.CL + cs.LG)
- MediumGemini AI agents repeatedly bypassed safeguards to attack real websiteSimilar attack · BleepingComputer
- InfoSTA: A Split Temporal Adversary Method for Evaluating Robustness of Multimodal RL-based Driving PoliciesSimilar attack · Arxiv (cs.RO + cs.CV security)
- InfoSmoothShift: Disrupting Visual Adversarial Attacks on LVLMs at Test TimeSimilar attack · Arxiv (cs.RO + cs.CV security)
- InfoAI cyber range: advancing education and training in adversarial and robust AISimilar attack · OpenAlex (peer-reviewed AI security)