Notes from NeurIPS 2025
(Views expressed here are my own and do not reflect those of my employer.)
[Read More]
-
-
vLLM-Watermark
Tiny, Hackable, Lightning-fast Watermarking for Researchers
Most text-watermarking papers ship code that nobody can actually use. Each codebase is welded to a specific experimental setup, and none of them are fast. MarkLLM consolidates several watermarking methods, but Gumbel watermarking still took hours on my benchmarks. [Read More] -
Watermarking Degrades Alignment in Language Models
Analysis and Mitigation
Watermarking is now used to authenticate LLM outputs. But we know surprisingly little about how it affects model behavior. In our new paper, “Watermarking Degrades Alignment in Language Models: Analysis and Mitigation,” presented at the 1st GenAI Watermarking Workshop at ICLR 2025, we investigate how watermarking impacts key alignment properties... [Read More] -
Red-Teaming Large Language Models (LLMs)
Operationalizing a Threat Model (SoK)
Imagine asking ChatGPT for homework help and receiving the response “Please die.”; or picture discovering that an AI chatbot has leaked someone’s private medical information. These real incidents that have made headlines in recent months underscore the fragility of current AI safeguards. Slides available here [Read More]