The paper proposes CANOPY, a minimalist reinforcement learning protocol that addresses two common pitfalls—signal starvation and policy drift—in outcome‑only RL for long‑horizon interactive tasks. By scaling same‑task exploration, keeping updates on‑policy, and anchoring updates with KL divergence, CANOPY enables a Qwen3‑14B agent to achieve top leaderboard results on the AppWorld coding benchmark without auxiliary supervision or elaborate scaffolding. The approach also improves performance on SWE‑bench for a Qwen3.5‑9B model.
arXiv:2610.18868v1 Announce Type: cross Abstract: Long-horizon agents have drawn growing attention for their ability to work autonomously -- from hours to days -- on economically valuable tasks, a ca...
arXiv:2510.15047v2 Announce Type: replace-cross Abstract: Large Language Models (LLMs) as agents often fail to improve in new environments. We identify and characterize a failure mode we call explora...
arXiv:2607.08837v4 Announce Type: replace-cross Abstract: Exploration is essential to RL since a policy cannot improve by repeatedly sampling the behaviors it already prefers. Standard methods inject...
arXiv:2607. 08837v1 Announce Type: cross Abstract: Exploration is essential to RL since a policy cannot improve by repeatedly sampling the behaviors it already prefers.
arXiv:2605.11467v2 Announce Type: replace-cross Abstract: Reasoning models post-hoc rationalize answers they have already committed to internally, producing chains of *reasoning theater*: deliberativ...
arXiv:2607. 22724v1 Announce Type: cross Abstract: Group-based policy optimization has been increasingly used to train large language model (LLM) agents from sparse outcome rewards by comparing trajectories or steps within a group.
The paper introduces Self‑Sparse Speculative Rollouts (S$^3$R), a method that uses sparse attention to propose tokens and then verifies them with full attention, thereby reducing KV‑cache traffic while preserving the dense‑policy distribution for reinforcement learning rollouts. S$^3$R eliminates the need for a separate drafter by sharing policy parameters between drafting and verification, and a KV‑aware scheduler limits cache thrashing to maintain parallelism. Experiments on long‑context mathematical reasoning with the Qwen3 model show that S$^3$R achieves 1.35–1.39× higher end‑to‑end training throughput with comparable task performance.
The paper introduces Coverage-conditioned Group Policy Optimization (CovGPO), a method that addresses sampling imbalance in group-based policy optimization for long-horizon agentic tasks. CovGPO applies first-visit observation coverage only when all samples in a group receive zero outcome reward, giving higher relative advantages to state-exploring actions without altering the original reward signal. Experiments on ALFWorld and WebShop benchmarks demonstrate that CovGPO consistently outperforms group-based baselines, especially on hard tasks.
arXiv:2607. 16204v1 Announce Type: new Abstract: Recent growth in reinforcement learning (RL) has surfaced a need for diverse, specialized training environments.
arXiv:2607. 07435v1 Announce Type: cross Abstract: Agents acting on our behalf in the real world (e.
arXiv:2608. 06735v1 Announce Type: new Abstract: Reinforcement learning (RL) has achieved strong results in improving large language models (LLMs) on tasks with stationary, verifiable rewards, such as mathematical reasoning and code execution.
The paper introduces Feedback‑Enriched Environments (FEEs) as a new approach to training large language models as autonomous agents for long‑horizon tasks. By shifting from action guidance to observation enrichment during later stages of exploration, FEEs improve performance across SciWorld and BFCL benchmarks with various Qwen3 model scales and RL algorithms. The study shows that FEEs stabilize training, promote proactive exploration, embed environmental guidance into policy weights, and highlight intra‑group feedback consistency as key for stable optimization.