Agent skills have emerged as an effective means of enabling LLM agents to solve complex, interactive tasks by encoding reusable knowledge without modifying model parameters. Recent approaches further improve skill adaptation by distilling reusable experience from agent execution trajectories and evolving skills based o...
Jia-Hao Wang, Yu-Qi Xiong, Qing Yang et al.· 0 citations
Across eight multimodal benchmarks, SCOPED consistently outperforms the backbone and mixed-capability RL, demonstrating the effectiveness of skill-decoupled on-policy expert distillation for integrating independently acquired capabilities.
Yu-Qi Xiong, Chun-Yi Peng, Zhi-Peng Xu et al.· 0 citations
Dense retrievers have made significant progress on text and image corpora, but whether these capabilities extend reliably to mixed corpora containing text, image, and fused text-image documents remains unclear. In this paper, we systematically examine retrievers across architectures and find that their performance is h...
Yu-Bo Sun, Chun-Yi Peng, Yu-Kun Yan et al.· 0 citations
The MiMo-V2.6 series is introduced, an omni-modal family that pushes the frontier of model intelligence by scaling RL compute by building infrastructure for mixed-task agentic RL, including a unified trajectory representation, high-concurrency multi-framework rollout, decoupled control and data planes, and training-inf...
Xiao-Zhen Qiao, Zi-Yue Hua, Zi-Rui Ou et al.· 10 citations
AUV-Bench is introduced, developed in collaboration with professional UI designers around 1,395 executable web interfaces and four tasks: aesthetic scoring, diagnosis, repair, and text-to-UI generation, which exhibits partial aesthetic competence, but still lack the fine-grained understanding and judgment-action cohere...
Zhi-Jie Deng, Ling Li, Jun-Hao Ji et al.· 0 citations
ConceptFormer models query-relevant evidence as continuous, query-conditioned latent concepts that explicitly bridge localized visual evidence and semantic relevance, without requiring either textual intermediate representations or direct reliance on raw visual annotations.
Chunyi Peng, Zhi-Peng Xu, Yu-Kun Yan et al.· 0 citations
PAPT++ is introduced, a risk-aware adversarial generation-training framework for SDG that progressively exposes the classifier to challenging yet semantically consistent variations.
Zhi-Peng Xu, De Cheng, Xinyang Jiang et al.· 0 citations
A paired ablation that removes explicit scientific guidance while preserving the repository and executable engineering context shows that scientific knowledge is not uniformly beneficial: well-grounded information can constrain repair and improve average performance and token efficiency, whereas poorly aligned guidance...
Zhi-Peng Xu, Jia-Hao Lu, Yi-Ning Zheng et al.· 4 citations
SAYRE is presented, a scene-aware document synthesis framework for generating scalable KIE training data without hand-crafted template design, and error analysis shows that synthesized training reduces field-level errors by improving schema-aware extraction over dense tables, business identifiers, and contract clauses.
Zhipeng Xu, Zulong Chen, Qing Liu et al.· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.