Adaptive Layerwise Perturbation: Unifying Off-Policy Corrections for LLM RL
This addresses a critical bottleneck in LLM RL training stability for researchers and practitioners, though it appears incremental as it builds on existing off-policy correction methods.
The paper tackles the problem of training instability in LLM reinforcement learning caused by heavy-tailed importance ratios due to policy distribution gaps, proposing Adaptive Layerwise Perturbation (ALP) which injects learnable perturbations into hidden states to flatten the distribution and reduce ratio tails. Experiments on math and reasoning tasks show ALP improves final performance, avoids KL spikes, and boosts exploration while maintaining training stability.
Off-policy problems such as policy staleness and training-inference mismatch, has become a major bottleneck for training stability and further exploration for LLM RL. To enhance inference efficiency, the distribution gap between the inference and updated policy grows, leading to heavy-tailed importance ratios. Heavy-tailed ratios arise when the policy is locally sharp, which further inflates sharp gradients and can push updates outside the trust region. To address this, we propose Adaptive Layerwise Perturbation(ALP) by injecting small learnable perturbations into input hidden states of each layer during updates, which is used as the numerator of the importance ratio against the unchanged inference policy in the objective. Intuitively, by adding controlled noise to intermediate representations, ALP prevents the updated policy from deviating too sharply from the inference policy, and enlarges the policy family to cover the inference policy family with mismatch noises. Hence, the flattened distribution can naturally tighten the updated and inference policy gap and reduce the tail of importance ratios, thus maintaining training stability. This is further validated empirically. Experiments on single-turn math and multi-turn tool-integrated reasoning tasks show that ALP not only improves final performance, but also avoid blow up of importance ratio tail and KL spikes during iterative training, along with boosted exploration. Ablations show that representation-level perturbations across all layers are most effective, substantially outperforming partial-layer and logits-only variants.