Junbo Zhao

Context-sensitive hallucinations in LLMs can be mitigated by a new optimization technique that rewards step-level consistency, leading to significant improvements in reasoning accuracy.

Dense retrievers miss the mark on answerability, dropping QA performance drastically despite high semantic relevance, revealing a critical oversight in RAG systems.

OPSD can backfire, leading to rote memorization instead of enhancing reasoning, but a novel decomposition approach reveals a path to meaningful improvements.

ISPO reduces critical reasoning failures in RLVR by transforming reward structures, leading to superior performance on complex reasoning tasks.

SkillComposer enables language models to self-evolve skills in real-time, achieving up to +4.5 improvements on agent tasks compared to larger models.

Synthetic data that looks good can still tank your model's performance – Optimsyn uses influence functions to find the *actually* useful synthetic examples and optimize your generation rubrics.

Forget quadratic attention: FEAT achieves state-of-the-art performance on structured data with linear complexity and 40x faster inference.

LLMs can match SOTA supervised theorem provers without training, if you give them the right structural scaffolding.