AI Paper Atlas

Research map / Reinforcement learning / RL from feedback

Reinforcement-aware Knowledge Distillation for LLM Reasoning

Zhaoyang Zhang, Shuli Jiang, Yantao Shen, Yuting Zhang, Dhananjay Ram, Shuo Yang, Zhuowen Tu, Wei Xia, Stefano Soatto

ICML 2026 · Reinforcement learning › RL from feedback · approach: RL for LLM reasoning

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use RLAD (TRRD distillation) instead of plain GRPOQwen3-0.6B · K&K Logistics (8K)average accuracy: 0.76 -> 0.94Improved
use RLAD (TRRD distillation) instead of plain GRPOQwen3-1.7B · K&K Logistics (8K)average accuracy: 0.95 -> 0.99Improved
use RLAD (TRRD distillation) instead of plain GRPOQwen3-0.6B · K&K Logistics (2K)average accuracy: 0.70 -> 0.90Improved
use RLAD instead of GRPO on hardest splitQwen3-0.6B · K&K Logistics PPL8 (8K)accuracy: 0.63 to 0.83Improved
use RLAD instead of GRPOQwen3-8B-Base · Long-context math avg (30K)average score: 61.0 -> 66.5Improved
use RLAD instead of GRPOQwen3-1.7B-Base · Long-context math avg (30K)average score: 36.5 -> 39.0Improved
use RLAD instead of GRPOQwen3-8B-Base · AIME25-Pass@32Pass@32: 48.5 to 66.4Improved
use RLAD instead of GRPOQwen3-8B-Base · MATH500 (30K)Mean@32: 68.2 to 68.6No clear effect
use RLAD instead of GRPOQwen3-1.7B (post-trained) · Math avg (8K)average score: 51.2 to 57.8Improved
add KDRL distillation from a weaker teacher (Qwen3-0.6B) (ablation)Qwen3-1.7B · K&K Logistics (8K)average accuracy: 0.950 -> 0.910Hurt
add RLAD distillation from a weaker teacher (Qwen3-0.6B) (ablation)Qwen3-1.7B · K&K Logistics (8K)average accuracy: 0.950 -> 0.946No clear effect
raise mixture coefficient alpha from 0.5 to 0.9 (ablation)RLAD · MATHPass@1: 80.5 -> 77.4Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy