NJU-LINK Lab · 10 records

dateidtitleauthorsvenue↑
2026-10-222610.17860Socratic Skill Learning: Interleaving Teaching and Practice for Transferable Agent SkillsLetian Zhu et al.arXiv45
2026-10-092610.12289TestPrism: Rethinking Test Evaluation Beyond a Single ReferenceHan Li et al.arXiv35
2026-08-062607.24821AVE-Compass: Towards Holistic Evaluation for Audio-Video Editing AbilitiesYuqing Wen et al.arXiv11
2026-06-232606.22883CLI-Universe: Towards Verifiable Task Synthesis Engine for Terminal AgentsZhanbo Hua et al.arXiv38
2026-06-092606.08415CoVEBench: Can Video Editing Models Handle Complex Instructions?Jiangtao Wu et al.arXiv47
2026-06-092606.08572OmniCap-IF: Benchmarking and Improving Instruction Following Abilities for Omni-Video CaptioningJiahao Wang et al.arXiv13
2026-06-042606.02060Where Do Deep-Research Agents Go Wrong? Span-Level Error Localization in Agent TrajectoriesJiaming Wang et al.arXiv58
2026-06-042606.01993MMG2Skill: Can Agents Distill In-the-Wild Guides into Self-Evolving Skills?Xinyu Che et al.arXiv16
2026-06-022606.02320TVIR: Building Deep Research Agents Towards Text--Visual Interleaved Report GenerationXinkai Ma et al.arXiv11
2026-05-182605.15301Solvita: Enhancing Large Language Models for Competitive Programming via Agentic EvolutionHan Li et al.arXiv21