LMSYS's New OPD Training Makes Qwen3.5 Reason 3X Faster with Improved Accuracy
ying11231 · x · 2026-07-23
LMSYS's Miles system has introduced On-Policy Distillation (OPD) as a first-class training primitive. In its first experiment, pure OPD reduced the reasoning length of the Qwen3.5-35B-A3B model by approximately 3X while slightly improving accuracy, without using any task rewards.
Key highlights from the controlled self-distillation run include:
- Accuracy improved: held-out DAPO rose from 0.8457 to 0.8945
- 3X shorter responses: rollout length dropped from 18.6k tokens to 5.5k–6.7k
- Student converged to teacher: reverse-KL fell from 0.045 to 0.010
- Efficient scoring: uses sparse per-position teacher scoring instead of dense O(R²K) payload
- Flexible to use: supports running pure distillation, or OPD combined with GRPO/PPO rewards
Next steps for the team include exploring OPD-augmented RL, scaling studies, and multi-teacher OPD.
More from Research
- Manual checks of 105 papers find most reproducibility failures come from missing code and mismatched results — ChenhaoTan · 2026-07-23
- A proxy gateway approach to stopping agent tool-call loops — bulleykebaal · 2026-07-23
- NEJM AI: Generative AI Intervention Effectively Mitigates College Student Depression — zakkohane · 2026-07-23
- Cohere to Host Talk on LLM Agent Reliability & Uncertainty Signals — Cohere_Labs · 2026-07-23
- Writer Study: Optimizing AI Harness Reduces Costs by 41% Without Losing Accuracy — bendee983 · 2026-07-23
- Reasoning traces from math breakthroughs may reveal how models think — benno_krojer · 2026-07-23