8 papers
DocAtlas: Long-Document Understanding as Mutable-State Interaction
Hongchen Wei, Yuanzhe Wang, Bei Liu +8
Long-document understanding requires models to find and combine evidence across many pages, layouts, tables, figures, and charts. Existing retrieval-augmented systems usually selec…
XL-DocBench: Benchmarking Evidence-Grounded Extra-Long Document Understanding
Hongchen Wei, Yuanzhe Wang, Bei Liu +9
Real-world document tasks often ask professionals to answer questions from annual reports, regulations, clinical guidelines, and technical manuals that span hundreds or thousands o…
Covering Human Action Space for Computer Use: Data Synthesis and Benchmark
Miaosen Zhang, Xiaohan Zhao, Zhihong Tan +14
Computer-use agents (CUAs) automate on-screen work, as illustrated by GPT-5.4 and Claude. Yet their reliability on complex, low-frequency interactions is still poor, limiting user…
Towards On-Policy SFT: Distribution Discriminant Theory and its Applications in LLM Training
Miaosen Zhang, Yishan Liu, Shuxia Lin +8
Supervised fine-tuning (SFT) is computationally efficient but often yields inferior generalization compared to reinforcement learning (RL). This gap is primarily driven by RL's use…
Controlled LLM Training on Spectral Sphere
Tian Xie, Haoming Luo, Haoyu Tang +9
Scaling large models requires optimization strategies that ensure rapid convergence grounded in stability. Maximal Update Parametrization (P) provides a theoretical…
RE-TRAC: REcursive TRAjectory Compression for Deep Search Agents
Jialiang Zhu, Gongrui Zhang, Xiaolong Ma +17
LLM-based deep research agents are largely built on the ReAct framework. This linear design makes it difficult to revisit earlier states, branch into alternative search directions,…