LLM Digest

Story

arxiv_cs_lg · Oct 22, 2026 · paper

Source brief

When to Follow the Teacher: Advantage-Based On-Policy Distillation

arxiv.orgOct 22, 2026
original source linked

In brief

Online Policy Distillation (OPD) often outperforms offline methods such as Supervised Fine-Tuning (SFT) by training on trajectories sampled directly from the student. However, standard OPD trains the student to follow...

Feed lens

evaluation

Read the original at arxiv.org →Open in live feed Read that day’s brief

When to Follow the Teacher: Advantage-Based On-Policy Distillation

Earlier in this thread 4 items

Beyond Teacher Likelihood: Group-Calibrated On-Policy Distillation for Long-Context Reasoning

OPRD: On-Policy Reminder Distillation for Long-Horizon Coding Agents

Teaching Agents in their Own Style: Blackbox Teacher Distillation for Web Agents

Qwen-MT-Lumen at WMT2026: Instruction-Specialized Translation via Quality-Aware Training and On-Policy Distillation