§ feed · storyline

ReFlect: Training-Free Error Recovery for Long-Horizon LLM Reasoning

ReFlect introduces a training-free error recovery method for long-horizon LLM reasoning that detects inconsistencies and improves task success rates over existing baselines.

May 11 · 02:00:00 · primary fetch1 sourceupdated May 11 · 02:00:00

ReFlect introduces a training-free harness for error recovery in long-horizon LLM reasoning, detecting inconsistencies and improving success rates over baselines.

read full article on dev.to ↗

§ sources1 publication · timeline below

dev.toReFlect: Training-Free Error Recovery for Long-Horizon LLM Reasoningprimary02:00:00