Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Angles Don't Lie: Unlocking Training-Efficient RL Through the Model's Own Signals

About

Current Reinforcement Fine-tuning (RFT) paradigms for Large Language Models (LLMs) suffer from sample inefficiency due to the redundant exposure of identical queries under uniform data sampling. While previous work has explored curriculum learning via heuristic difficulty metrics, these strategies exhibit limitations by neglecting the intrinsic learning signals generated by the model itself, thus leading to suboptimal training regimes. In this paper, we identify a model-inherent signal termed angle concentration that effectively reflects an LLM's capacity to learn from specific data. We theoretically and empirically demonstrate a correlation between the angular distribution of token hidden state vectors and the resulting gradient, revealing a learning preference for data exhibiting higher angle concentration. Inspired by this finding, we propose GAIN-RL, a Gradient-driven Angle-Informed Navigated RL framework. By leveraging the model's intrinsic angle concentration signal, GAIN-RL dynamically selects training data in each epoch, ensuring consistently impactful gradient updates and thus significantly enhancing overall training efficiency. Empirical evaluations show that GAIN-RL (GRPO) achieves over a 2.5x acceleration in training efficiency across diverse mathematical and coding tasks and varying model scales. Furthermore, GAIN-RL (GRPO)'s efficient sampling yields data-efficient training, achieving better performance with half the original data compared to vanilla GRPO with full training data. Code is realsed at https://github.com/wangqinsi1/GAINRL/tree/main.

Qinsi Wang, Jinghan Ke, Hancheng Ye, Yueqian Lin, Yuzhe Fu, Jianyi Zhang, Kurt Keutzer, Chenfeng Xu, Yiran Chen• 2025

Related benchmarks

TaskDatasetResultRank
Mathematical ReasoningMATH500 (test)--
895
Mathematical ReasoningAIME 2024 (test)--
209
Mathematical ReasoningOlympiadBench (test)--
40
Mathematical ReasoningAMC 2023 (test)
Avg@8 Success Rate68.4
12
Mathematical ReasoningGSM8K
Training Steps to Target Accuracy220
12
Mathematical ReasoningMATH500
Training Steps to Target Accuracy180
12
Mathematical ReasoningMinervaMath
Steps to Target Accuracy220
12
Mathematical ReasoningAggregate AIME, AMC, GSM8K, MATH, MNV, OLPD
Avg Training Steps to Target Acc193
12
Mathematical ReasoningGSM8K (test)
Accuracy (Avg@8)90.1
12
Mathematical ReasoningAMC23
Training Steps240
12
Showing 10 of 12 rows

Other info

Follow for update