…Furthermore, it estimates an adaptive per-problem target length online based on the model's own correct rollouts, applying a symmetric efficiency reward that penalizes both overthinking and over-compression. Evaluated on…
…We propose On-Policy Representation Distillation (OPRD), which lifts distillation into hidden-state space by aligning student and teacher representations across selected layers on the same rollouts , bypassing the LM head entirely…
…sparse sequence-level reward should train models where exploration is productive, while dense token-level teacher reward should be used where the aim is to compress behavior into a smaller model. In…
…For each condition, S GRPO samples a supergroup of candidate sets, compares their diversity under the same condition, and redistributes the group diversity reward to individual rollouts through leave-one-out diversity…
…Learning Budget-Efficient Thinking for Adaptive Reasoning (2026) Where to Spend Rollouts: Hit-Utility Optimal Rollout Allocation for Group-Based RLVR (2026) Reasoning Is Not Free: Robust Adaptive Cost-Efficient Routing for…
…and recovers by treating a VLA / WAM as an interruptible tool it steers mid-rollout alongside vision models and action primitives. To evaluate these long-horizon capabilities, we introduce RoboVoLo, a high…
…Instead of materializing each policy as a merged full checkpoint , MinT keeps the base model resident and moves exported LoRA adapter revisions through rollout, update, export, evaluation, serving , and rollback, hiding distributed…
…Zihao Han , Tiangang Zhang , , Abstract Adaptive Teacher Exposure for Self-Distillation (ATESD) improves large language model reasoning by dynamically adjusting teacher exposure during training through a learnable policy controller. AI-generated summary…
…BraveGuard mines recent research sources to identify emerging risks and attack patterns, instantiates them as executable computer-use tasks , collects agent rollouts, and derives trajectory-level supervision for guard model training. As…
…Reinforcement Learning Unlocks Parametric Knowledge in LLMs Published on May 8 Submitted by Wanli Yang on May 13 Authors: , , , , , , , Abstract Reinforcement learning improves large language model recall of parametric knowledge by redistributing…