SG
Siyuan Gan
cs.AIcs.LGcs.CLstat.ML
On Valency
published · living versionsW_qc83whe2·v1 · currentpublished
Adaptive Pullback: Reweighting Supervision at Penalty Jumps in On-Policy Distillation
with Guangyu Liu, Shilong Zhou, Linjian Meng, Boyan Wang +1
1 version
W_eycyjqq8·v1published
Preprints & journals
6 papers in the corpus · 2026Adaptive Pullback: Reweighting Supervision at Penalty Jumps in On-Policy Distillation2610.28194v1 · Guangyu Liu, Siyuan Gan, Shilong Zhou et al.2026 · 0 citationsarXivon Valency
Nash-MOPD: Optimizing Task Trade-offs in Multi-Teacher On-Policy Distillation2610.26124v1 · Siyuan Gan, Bo Yang, Linjian Meng et al.2026 · 0 citationsarXivon Valency
Unbiased Top- Estimation for On-Policy Distillation2609.34447v2 · Linjian Meng, Siyuan Gan, YuHan Li et al.2026 · 0 citationsarXiv
When Teacher Guidance Misleads: Reward-Aligned On-Policy Distillation2608.27960v1 · Siyuan Gan, Yuhan Li, Xiran Wang et al.2026 · 0 citationsarXiv
Learning from Hard Prompts: Difficulty-aware Advantage Amplification in Dynamic Sampling2608.27982v1 · Siyuan Gan, Yuhan Li, Xiran Wang et al.2026 · 0 citationsarXiv
Thinking-Based Non-Thinking: Solving the Reward Hacking Problem in Training Hybrid Reasoning Models via Reinforcement Learning2601.04805v2 · Siyuan Gan, Jiaheng Liu, Boyan Wang et al.2026 · 0 citationsarXiv
Career total: 7 works. 6 are in this corpus.
Profile built from the corpus for this byline.
Author records are still filling in while the Hub is in alpha. If this is your page, you'll be able to claim it soon. Spot a mistake? Tell us.