Xiaoman Wang

Evidence retrieval from long multimodal documents can be dramatically improved by leveraging relevance diffusion over a Chunk-Page graph, leading to better question answering performance.

IA-RAG reveals that modeling knowledge as dynamic time intervals can drastically improve temporal reasoning in language models, outperforming traditional static approaches.

Accurately scoring the impact of individual edits in grammatical error correction is now possible without human annotation, thanks to a novel graph-based approach.

By forecasting compact world dynamics before taking action, DynVLA leapfrogs traditional CoT methods to achieve more informed and physically grounded autonomous driving decisions.