Showing cs.LGShow all
3 papers · 1 filter
cs.LG2026
When Are Teacher Tokens Reliable? Position-Weighted On-Policy Self-Distillation for Reasoning
Xiaogeng Liu, Xinyan Wang, Yingzi Ma +2
On-policy self-distillation (OPSD) trains a student on its own rollouts using a privileged teacher, but its standard objective weights all generated tokens equally, implicitly trea…
cs.LG2026
ROM: Real-time Overthinking Mitigation via Streaming Detection and Intervention
Xinyan Wang, Xiaogeng Liu, Ming Pei +1
Large reasoning models (LRMs) often suffer from overthinking, continuing with redundant verification and repeated attempts even after reaching a correct solution. This wastes compu…
cs.LG2024
Pruner: A Draft-then-Verify Exploration Mechanism to Accelerate Tensor Program Tuning
Liang Qiao, Jun Shi, Xiaoyu Hao +10
Tensor program tuning is essential for the efficient deployment of deep neural networks. Search-based approaches have demonstrated scalability and effectiveness in automatically fi…