DistIL: Distributional DAgger for RL from Rich Feedback beyond single-bit rewards
A new arXiv preprint introduces DistIL, a distributional variant of the DAgger imitation learning algorithm designed to exploit rich feedback signals (execution traces, tool outputs, expert corrections) rather than the single-bit correctness reward used in standard RLVR. The method uses a forward cross-entropy objective that provides monotonic policy improvement guarantees, unlike reverse KL or Jensen-Shannon divergence objectives used in prior self-distillation approaches. Empirically, DistIL outperforms RLVR and self-distillation baselines on scientific reasoning, coding, and hard math benchmarks.
Related guides (3)
Related events (8)
DelTA: Discriminative Token Credit Assignment for RLVR Training
DelTA introduces a discriminative token credit assignment method for reinforcement learning from verifiable rewards (RLVR) that addresses the problem of high-frequency formatting tokens dominating policy gradient updates. The method estimates per-token coefficients to amplify side-specific gradient directions and downweight shared or weakly discriminative ones, making the effective update direction more contrastive. On seven mathematical benchmarks, DelTA outperforms same-scale baselines by 3.26 and 2.62 average points on Qwen3-8B-Base and Qwen3-14B-Base respectively, with additional gains on code generation tasks.
Two is better than one: A Collapse-free Multi-Reward RLIF Training Framework
This paper proposes a multi-reward reinforcement learning from internal feedback (RLIF) framework that decomposes training signals into an answer-level reward via cluster voting and a completion-level reward via token-wise self-certainty. To address reward hacking and entropy collapse common in single-reward RLIF, the authors introduce GDPO-based normalization and KL-Cov regularization targeting low-entropy token distributions. Evaluated on mathematical reasoning and code-generation benchmarks, the method achieves stability and performance approaching supervised RLVR methods without requiring external ground-truth supervision. The work advances scalable unsupervised RL training for LLM reasoning.
d-OPSD: First on-policy self-distillation framework tailored for diffusion LLMs
Researchers introduce d-OPSD, the first on-policy self-distillation (OPSD) framework designed specifically for diffusion large language models (dLLMs). The method addresses a fundamental mismatch between existing autoregressive OPSD approaches and dLLMs' arbitrary-order generation by using suffix conditioning on self-generated answers and step-level rather than token-level divergence supervision. Across four reasoning benchmarks, d-OPSD outperforms RLVR and SFT baselines while requiring only ~10% of the optimization steps of RLVR, suggesting strong sample efficiency gains for dLLM post-training.
ExpRL: RL-based mid-training using human QA data as reward scaffolds for LLM reasoning
ExpRL proposes an automated approach to LLM mid-training that replaces manually curated reasoning traces with large corpora of human-written QA data used as reward scaffolds rather than imitation targets. Reference solutions are hidden from the policy and used only to construct problem-specific grading rubrics, enabling dense process-level rewards that reinforce partial progress and intermediate reasoning steps. On challenging math reasoning benchmarks, ExpRL outperforms SFT, sparse-reward GRPO, and self-distillation as an RL initialization strategy, with additional mixed-domain experiments suggesting broader applicability.
Reward uncertainty as a principled mechanism for diverse RL behaviour
A new arXiv preprint proposes replacing the scalar reward in RL with a distribution over reward functions, applying a non-linear objective over sets of actions to induce calibrated behavioural diversity without sacrificing expected reward. The authors derive a principled gradient estimator in the contextual bandit setting and prove the formulation generalizes vanilla policy gradient and action-set approaches. The work is motivated by applications like language model fine-tuning where diversity is desirable but entropy regularization and diversity bonuses introduce fragile trade-offs. Empirical results support the framework as a theoretically grounded alternative to heuristic diversity methods.
Finetune Stable Diffusion Models with DDPO via TRL
Hugging Face's TRL library adds support for DDPO (Denoising Diffusion Policy Optimization), enabling reinforcement learning-based finetuning of Stable Diffusion models. This extends TRL's RLHF tooling beyond language models to image generation, allowing reward-driven optimization of diffusion models. The post demonstrates practical usage of the new DDPO trainer within the TRL ecosystem.
Reinforcement Learning with Prediction-Based Rewards (Random Network Distillation)
OpenAI introduces Random Network Distillation (RND), a curiosity-driven exploration method for reinforcement learning that uses prediction error on a fixed random neural network as an intrinsic reward signal. RND is the first method to exceed average human performance on Montezuma's Revenge, a notoriously hard-exploration Atari game. The approach is simple to implement and compatible with standard RL algorithms, offering a scalable alternative to count-based or dynamics-model exploration bonuses.
Step-aligned critique outperforms GRPO and reference-solution conditioning in self-distillation
A new arXiv paper investigates context design for self-distillation of language models, comparing binary reward (GRPO), reference solutions, and step-by-step critiques aligned to the solver's reasoning trace. Step-aligned critique yields the largest gains, outperforming GRPO by 16.11 points and reference-solution conditioning by 5.27 points on Avg@12. Per-token advantage analysis shows that step-aligned feedback targets only failing tokens, avoiding unnecessary pressure on already-correct reasoning steps. The findings suggest structural alignment between feedback and the model's reasoning trace is a key driver of self-distillation effectiveness.