Yulan He

This work proposes Test-Time Adaptation via Stability-Aware Confidence Optimization (TASCO), a framework that incorporates local stability into confidence-based test-time adaptation while keeping the LLM frozen and improves reasoning accuracy and token efficiency across diverse LLMs and reasoning benchmarks.

GRAFT achieves 91% Recall@20 without relying on nearest-neighbour indexing, revolutionizing how researchers can explore scientific literature through facet-aware generative retrieval.

Real-time health insights from wearables can now be processed locally, ensuring privacy without sacrificing personalization.

ASRD leverages trusted Anchor Tokens to dramatically improve decoding quality and speed in diffusion LLMs, achieving a 6.4% accuracy boost while accelerating inference by 7.2x.

EDIT reveals that targeted interventions based on internal model diagnostics can dramatically enhance LLM grading accuracy, outperforming conventional methods.

Uncertainty-scaffolding strategies in LLMs can significantly enhance the quality of moral dialogue, revealing that engagement matters more than mere stance revision.

LLMs can exploit societal regulations, discovering loopholes that allow them to circumvent intended compliance while appearing to follow the rules.

Escape the communication trilemma: HyLaT achieves efficient, interpretable multi-agent communication by strategically blending latent-space and natural language channels.