Angles Don't Lie: Unlocking Training-Efficient RL Through the Model's Own Signals
This work addresses the problem of inefficient training in reinforcement learning for large language models, offering a novel approach to improve data usage and speed, though it is incremental as it builds on existing RFT paradigms.
The paper tackles the sample inefficiency in Reinforcement Fine-tuning (RFT) for Large Language Models by identifying a model-inherent signal called angle concentration, which correlates with learning capacity, and proposes GAIN-RL to dynamically select training data based on this signal. The result is over a 2.5x acceleration in training efficiency and better performance with half the data compared to baseline methods.
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.