Harbor Adapters are developed that port more than 80 benchmarks to evaluate arbitrary agents, and validated through rigorous code review and parity experiments, and Harbor-Index is introduced, a curated set of 82 difficult, diverse, and high-quality tasks spanning 29 benchmarks to support more reliable and comprehensiv...
Lin Shi, Hao-Wei Lin, Zi-Xuan Zhu et al.· 1 citation
STAMP is proposed, in which a reference-based verifier judges whether each cited document supports an entity or relation in a training-time evidence graph, and first-exposure attribution traces each supported citation back to the action that first surfaced it.
Ke Xu, Han Xu, Xin-Ran Chen et al.· arXiv.org· 1 citation
SCOPE-RL improves average accuracy by up to 11.2 pp and reduces reasoning tokens by up to 27.1% over outcome-only GRPO, indicating that reward-signal densification is complementary to policy-update-level RLVR advances.
Xiaojia Liu, Han Xu, Jianqiang Xia et al.· arXiv.org· 0 citations
An evaluation taxonomy, a non-instruction data taxonomy, and mapping rules form a closed loop turning a benchmark-level failure into a targeted, testable data intervention, showing the evaluation-to-data inference can be routine, auditable, and experimentally validated rather than intuitive.
Zhixuan Li, Jiangan Yuan, Han Xu· arXiv.org· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
