Tag
This paper proposes JUMP, a single-pass membership inference attack for fine-tuned discrete diffusion language models that exploits their any-order and parallel decodability to improve detection accuracy with fewer queries.
This paper proposes Global-Local Uncertainty (GLU), an unsupervised single-pass score that fuses token-level local entropy with hidden-state geometric global entropy for uncertainty quantification in LLMs, showing that the two are near-orthogonal and together capture confident-but-wrong failures.
The author built a fully autonomous AI agent named KawaiiBaka that runs in a Discord group chat, using a single-pass cognitive loop with native time perception and real access to execute Python on a Windows machine. The system uses Mistral API for the LLM and a local image generation pipeline to create context-aware selfies.