4 papers
Is Better Teacher Supervision Enough? Unlocking Student-side Learning in Multimodal On-Policy Distillation
Siyuan Liu, Kanghui Tian, Yue Duan +4
On-policy distillation (OPD) improves reasoning by providing token-level supervision from a teacher on a student's own trajectories. Existing methods primarily focus on enhancing t…
AgenticVAU: Multi-Agent Explore-Verify Reasoning for Video Anomaly Understanding
Yuxiang Duan, Huining Li, Ao Li +6
Video anomaly understanding (VAU) focuses on comprehensively interpreting abnormal events in videos, requiring models to identify anomalous occurrences, discover their supporting e…
DiagLoop: A Counterfactual Data Flywheel with Stage-Localized Reinforcement for Diagnostic LLMs
Jian Zhang, Bingyi Wang, Yizhi Liu
Causal diagnostic models must explain how conclusions follow from evidence because diagnoses guide repairs and treatments. Yet serious cases are scarce, records rarely contain reas…
CausalOPD: First-Wrong-Step Supervision for Distilling Causal Chain Reasoning
Jian Zhang, Bingyi Wang, Yizhi Liu
Many critical reasoning tasks, including clinical diagnosis, legal judgment, and industrial fault diagnosis, require step-dependent causal chains in which early errors propagate an…