The paper introduces ReTRAIL, a framework that allows long‑horizon agents to rewind the environment to earlier states while preserving forward progress in knowledge. By training a shared policy with State‑Anchored GRPO and a branch‑advancement reward, ReTRAIL enables agents to recover from unproductive states and reuse insights from abandoned exploration. Experiments on MineSweeper, Sokoban, ALFWorld, and ScienceWorld show that this combined rewinding and experience‑reuse strategy outperforms simple rewind primitives and improves long‑horizon task solving.
arXiv:2608.21946v1 Announce Type: cross Abstract: Reinforcement learning with outcome-based objectives such as GRPO enables LLM-based agents to solve complex, long-horizon tasks, yet the reusable exp...
arXiv:2609.18304v1 Announce Type: new Abstract: Large language model (LLM) agents increasingly tackle long-horizon tasks through multi-step environment interaction, yet a single erroneous action can...
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...
The paper introduces Recuris, a recursive Experiential‑Working Memory architecture that lets long‑horizon agents track task progress and select skills based on current needs rather than full history. By coupling working memory with experiential memory, execution becomes structured evidence that localizes failures to specific memory components, enabling a bounded recursive memory‑evolution loop. Across four benchmarks and ten models, Recuris improves task success in 35 of 37 model‑benchmark pairs, raising state‑of‑the‑art performance on tau‑bench and SkillFlow and reducing common long‑horizon failures by up to 80%.
Recursive self-improvement (RSI) remains hard in long-horizon tasks, where growing histories obscure the task state and misalign skill invocation. We introduce Recuris, a recursive Experiential-Workin...
The paper introduces checkpoint handoff, an evaluation protocol that separates an agent’s ability to reach useful states from its ability to complete tasks in reinforcement learning. By using one checkpoint as a reacher up to a handoff point and another as a solver from the same replayed history, the authors can measure Reach (how often states within a fixed number of actions from success are achieved) and Solve (how often the task is completed from those states). Experiments on TravelPlanner and ALFWorld show that switching the solver from supervised fine‑tuning to RL yields larger gains when RL is used as the reacher, indicating that RL more effectively finds solvable states.
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:2609.36864v1 Announce Type: new Abstract: Group-relative methods for reinforcement learning with verifiable rewards (RLVR) learn from differences in rollout outcomes. Independently sampling com...
arXiv:2607. 08716v1 Announce Type: new Abstract: In long-horizon tasks, decision-relevant state is often scattered across an expanding trajectory, while the action agent must surface it and act.
A ReAct-style agent accumulates observations, intermediate results, and failed attempts in a single context. As this context grows, stale and off-path information consumes its limited capacity and obs...
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...
The paper introduces NavMCP, a scaffolding framework that couples vision‑language models (VLMs) with navigation foundation models (NFMs) to enable long‑horizon physical‑world agents. NavMCP orchestrates three communication channels—intent, observation, and memory—to allow the VLM to decide what evidence to seek and the NFM to ground semantic sub‑goals into closed‑loop navigation, without retraining either model. The approach achieves state‑of‑the‑art results on several embodied question‑answering benchmarks and significantly outperforms episodic interfaces on the Unitree Go2 robot as task horizons lengthen.