Policy optimisation, reward modelling and RLHF — how models are trained by feedback rather than by labels.
arXiv:2610.14523v1 Announce Type: new Abstract: ScienceBuddy-PV adapts the ScienceBuddy recursive-in-recursive design to panvascular intervention research through a structured evidence ledger, device...
arXiv:2610.14939v1 Announce Type: new Abstract: GRPO-style reinforcement learning has advanced vision-language reasoning, but its sequence-level rewards provide limited credit assignment for individu...
arXiv:2610.14956v1 Announce Type: new Abstract: The Model Context Protocol provides a shared interaction substrate for the agentic web, allowing heterogeneous services to expose capabilities through...
arXiv:2610.15691v1 Announce Type: new Abstract: Existing reinforcement learning (RL) largely focuses on optimizing learning algorithms within a pre-defined task formulation, while applying RL to new...
arXiv:2610.17266v1 Announce Type: new Abstract: Diffusion language models are a promising and more efficient alternative to autoregressive models. However, most RL post-training techniques are not ea...
arXiv:2610.16027v1 Announce Type: cross Abstract: Reinforcement learning improves tool-use agents, yet fine-grained credit assignment remains challenging even with step-level feedback. An observed er...
arXiv:2610.14402v1 Announce Type: new Abstract: Large language models (LLMs) have achieved strong performance in mathematical and logical reasoning, yet they still suffer from the curse of complexity...
arXiv:2610.18484v1 Announce Type: new Abstract: Terminal agents have shown strong abilities in complex coding tasks. However, they often struggle to recover from errors during long-running execution...
The paper introduces CoPE, a method that trains language agents by adding multi‑step action targets to the policy backbone. By supervising each action forecasted from preceding histories, CoPE enables acting and forecasting to co‑evolve without extra inference cost. Across five benchmarks, CoPE outperforms existing RL baselines, achieving up to twice the sample efficiency on ScienceWorld and improving overall performance on realistic applications.
HarnessMDP is a learning-based framework that treats the execution harness of large language model agents as a learnable control layer. By formulating agent execution as a finite-horizon Markov decision process, the harness learns to select structural execution actions from the evolving agent state, using offline reinforcement learning from collected trajectories. Experiments on coding tasks, tau-bench, AgentBench, and other evaluation settings show that HarnessMDP consistently improves execution behavior and final task performance over baseline harnesses.
arXiv:2610.20497v1 Announce Type: new Abstract: Reinforcement learning with verifiable rewards (RLVR) trains language models from checked task outcomes, but assigning a terminal reward to individual...
arXiv:2610.17309v1 Announce Type: cross Abstract: We introduce StartLux-Decision, decision models from 0.8B to 27B parameters that combine efficient single-pass prediction with adaptive, confidence-g...
arXiv:2610.20174v1 Announce Type: new Abstract: Reinforcement Learning on Mixed-domain preference data (Mix-RL) has become the mainstream approach for training general-purpose Generative Reward Model...
arXiv:2610.14988v1 Announce Type: cross Abstract: The average-reward setting for reinforcement learning is challenging because its Bellman operator is known to contract in span only when the underlyi...
arXiv:2610.15328v1 Announce Type: cross Abstract: We introduce Anchor-and-Harvest Curriculum (AHC), a domain-informed, rule-based training system for Craftax-Full. AHC combines achievement-oriented t...
arXiv:2610.15428v1 Announce Type: cross Abstract: Context-based offline meta-reinforcement learning (COMRL) has advanced almost entirely through better task representations, which separate task ident...
arXiv:2610.17541v1 Announce Type: cross Abstract: Safety verification of a Markov decision process (MDP) asks whether the probabilities of reaching several failure states jointly lie in a convex safe...
arXiv:2610.17726v1 Announce Type: cross Abstract: Large language model (LLM) agents excel at complex, multi-step tasks but suffer from catastrophic forgetting during continual learning--acquiring new...
arXiv:2610.20334v1 Announce Type: cross Abstract: A ReAct-style agent accumulates observations, intermediate results, and failed attempts in a single context. As this context grows, stale and off-pat...
arXiv:2610.16979v1 Announce Type: cross Abstract: Generative retrieval (GR) offers a promising paradigm for industrial search by unifying candidate generation within an autoregressive framework. Howe...