这篇论文解决了GRPO训练时难问题学不到东西的痛点，用自适应前缀让模型在数学推理上准确率翻倍，而且模型越小效果越明显，值得做强化学习的人看看。
AdaPrefix-GRPO针对GRPO在困难问题上梯度消失的问题，提出自适应调整正确前缀长度的机制。方法通过反馈控制器将问题的成功率维持在50%附近以最大化梯度信号，训练后完全移除前缀。在0.6B模型上，该方法在保留训练分布的难题上将准确率提升2.1倍；在Qwen3-1.7B上提升1.6倍，在AIME基准上提升1.7倍，同时将追踪长度减半。模型越小，增益越大。
Max Out GRPO Signal: Adaptive Trace Prefix Control for Hard Reasoning Problems
Group Relative Policy Optimization (GRPO) stalls on a model's hardest problems: when no rollout in a group succeeds, the group-relative advantages vanish and the problem contributes no gradient, wasting the frontier examples we most want to learn from. Prepending a correct prefix of a reference solution raises the success rate, making prefix length a continuous knob on difficulty. Concurrent methods set the knob once; AdaPrefix-GRPO turns it into a feedback controller: throughout training it adjusts how much of the solution each problem gets, holding its success rate near 50%, where GRPO's gradient signal is largest, then withdraws the assistance entirely, so the deployed model solves problems unaided. On hard math, at matched training FLOPs, it more than doubles GRPO's accuracy on held-out problems from the training distribution for a 0.6B model (2.1x), with 1.6x on Qwen3-1.7B and 1.7x on AIME, while roughly halving trace length. The method is implemented in data preparation plus a loss mask on prefix tokens; the trainer is otherwise stock. The smaller the model, the larger the gain.