AI Paper Atlas

Research map / Large language models / Instruction and fine-tuning

Reinforcement Learning Fine-Tuning Enhances Activation Intensity and Diversity in the Internal Circuitry of LLMs

Honglin Zhang, Qianyue Hao, Fengli Xu, Yong Li

ICLR 2026 · Large language models › Instruction and fine-tuning · approach: Fine-tuning generalization and delta compression · 6 citations (OpenAlex)

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
RL fine-tuning (PPO/GRPO)SFT 7B math models · MATH/GSM8K/College Mathactivation intensity: Mistral MATH alpha=0.5: 3.58e-4 -> 4.05e-4Improved
online RL fine-tuningSFT models · MATH, GSM8Kinformation complexity: GSM8K Deepseek-Math alpha=0.5: 1.37e-1 -> 3.23e-1Improved
online RL fine-tuningSFT models · MATHdistribution kurtosis: Deepseek-Math alpha=0.5: 3.03e+2 -> 1.89e+2Improved
DPO instead of online RLQwen2.5-7B-SFT · MATH, College Mathactivation intensity / information complexity: MATH alpha=0.1 intensity: 6.95e-4 -> 6.90e-4No clear effect
DPO fine-tuningQwen2.5-7B-SFT · all datasetsdistribution kurtosis: MATH alpha=0.5 not reported; alpha=0.1: 5.44e+2 -> 4.83e+2Improved
lower sampling temperature (0.6 vs 1.0) during RL (ablation)GRPO variant on Qwen2.5-3B-Instruct · GSM8K testaccuracy: 36.5 -> 51.2Improved
RL post-trainingMistral-7B SFT · Minerva mathaccuracy: 12.1 -> 11.8Hurt
RL post-trainingDeepSeek-Math, DS-Distill-Qwen · MATHaccuracy: 46.2 -> 52.6; 88.4 -> 95.4Improved

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