mlsanityData to 14 Oct 2026

ReflexGrad: Within-Episode Failure Recovery in LLM Agents via Progress-Gated Dual-Process Routing

arXiv 2511.14584 ↗First posted 18 Nov 2025cs.LG cs.AIcited by 2

On the map: LLM agents and tool use → Software engineering coding agents

Abstract (arXiv)

We present ReflexGrad, a dual-process architecture for within-episode failure recovery in LLM agents without demonstrations. When agents commit to a wrong approach early and exhaust the step budget, the post-failure trajectory contains the information to escape -- but no published architecture acts on it within a single episode. ReflexGrad routes between a fast process (TextGrad-style continuous refinement every k=3 steps) and a slow process (Reflexion-style causal diagnosis when m=5 consecutive low-progress scores fire a routing gate). A deterministic priority merge keeps the natural-language policy coherent, and each slow activation emits three observable artifacts: a reproducible trigger, a causal diagnostic, and a verified fix. On ALFWorld 134 tasks, n=10 seeds, no demonstrations, ReflexGrad lifts Qwen-3-8B from 35.1% to 75.4% (+40.3pp), beating compute-matched 1-shot LATS by +2.7pp (p≈0.01), ToT by +5.7pp (p<10-4), and Self-Refine by +6.7pp (p<10-5); on GPT-5 the lift is 46.3→88.1% (+41.8pp). The 1.5pp cross-model difference is within seed noise (p≈0.13), suggesting that the routing mechanism, rather than model scale, is the primary source of the gain. Code, prompts, per-seed logs, and sensitivity sweeps are released.

Key facts. Cited by 2 papers in the map (data through 14 Oct 2026). Placed in Software engineering coding agents, one of 16 directions in LLM agents and tool use. Builds on 13 of its 23 references that are in the map. First posted on arXiv 18 Nov 2025.

Builds on

13 of 23 references.

References · 13

  1. 2025Qwen3 Technical Reportcited by 8,820
  2. 2023Reflexion: Language Agents with Verbal Reinforcement Learningcited by 4,276
  3. 2023Self-Refine: Iterative Refinement with Self-Feedbackcited by 3,442
  4. 2023Voyager: An Open-Ended Embodied Agent with Large Language Modelscited by 2,191
  5. 2020ALFWorld: Aligning Text and Embodied Environments for Interactive Learningcited by 1,205
  6. 2024OSWorld: Benchmarking Multimodal Agents for Open-Ended Tasks in Real Computer Environmentscited by 1,195
  7. 2023ExpeL: LLM Agents Are Experiential Learnerscited by 876
  8. 2023Large Language Models as Optimizerscited by 864
  9. 2023DSPy: Compiling Declarative Language Model Calls into Self-Improving Pipelinescited by 832
  10. 2023CRITIC: Large Language Models Can Self-Correct with Tool-Interactive Critiquingcited by 783
  11. 2018TextWorld: A Learning Environment for Text-based Gamescited by 310
  12. 2023AdaPlanner: Adaptive Planning from Feedback with Language Modelscited by 196
1 more
  1. 2025ReflAct: World-Grounded Decision Making in LLM Agents via Goal-State Reflectioncited by 31

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 2.

2 papers.

  1. 2026Graph Engineering in the Era of LLM Agents: From Individual Intelligence to System Intelligencecited by 3
  2. 2026What Can Component-Replacement Evidence Establish? A Critical Scoping Review of Local Decisions in LLM Agentscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026Sentry: Learning to Recover from LLM Agent Failures at Test Time
  2. 2026When Harnesses Lose the Signal: Causal Evaluation of Recovery in LLM Agents
  3. 2026Failure Recovery via Belief Updates and World Modeling for Long-Horizon Agents
  4. 2026ANNEAL: Adapting LLM Agents via Governed Symbolic Patch Learning
  5. 2025Conditional Multi-Stage Failure Recovery for Embodied Agents
  6. 2026Doomed from the Start: Early Abort of LLM Agent Episodes via a Recall-Controlled Probe Cascade
  7. 2026SiLR: Structure-Preserving Admission and Process Reward for LLM Tool Agents
  8. 2026Diagnosis Before Recovery: Turning Agent Failures into Selective Self-Correction