---
title: "Meituan Trains AI Agents to Stop Repeating the Same Tool Mistakes"
subtitle: "Meituan researchers turn error recovery into a trainable skill, doubling multi-turn tool-call accuracy on BFCL v3 and outperforming GPT-4o-mini on repair."
url: "https://alphasignal.ai/news/meituan-trains-ai-agents-to-stop-repeating-the-same-tool-mistakes"
author: "AlphaSignal Newsroom"
author_url: "https://alphasignal.ai/editorial-team"
company: "AlphaSignal"
company_url: "https://alphasignal.ai/company/alphasignal"
published: "2026-04-15T00:00:00.000Z"
updated: "2026-08-23T22:15:55.941Z"
type: "Paper"
topics:
  - "Agents"
  - "Post Training"
subtopics:
  - "Tool Use"
  - "Dpo"
  - "Rl"
upvotes: 1863
source_url: "https://arxiv.org/abs/2509.18847"
comments: 0
truncated: true
---

# Meituan Trains AI Agents to Stop Repeating the Same Tool Mistakes

> Meituan researchers turn error recovery into a trainable skill, doubling multi-turn tool-call accuracy on BFCL v3 and outperforming GPT-4o-mini on repair.

Anyone who has watched an LLM agent trip over a tool call knows the pattern: it makes a mistake, gets an error back, and then confidently makes the exact same mistake again. A new paper from Meituan's Vision Agent Team argues this is a training problem, not a prompting one. Their fix is to treat the path from failure to repair as a first-class action the model learns to produce, rather than a heuristic side-effect of chain-of-thought.

The work, titled [Failure Makes the Agent Stronger](https://arxiv.org/abs/2509.18847), introduces a technique called structured reflection along with a benchmark and training recipe for tool-using agents. [Code and data](https://github.com/MeiGen-AI/Tool-Reflection-Bench) are on GitHub under an MIT license.

## Why one-shot tool training breaks down

Most tool-augmented LLMs today are trained with supervised fine-tuning or coarse-grained RL that optimizes a single tool call in isolation. That works for a first attempt, but multi-turn interactions are a different beast. A tiny typo in a parameter, a swapped argument, or a call issued in the wrong order invalidates the whole function invocation, and the reward signal collapses to zero. Existing self-correction approaches lean on prompts that tell the model to think harder, which is fragile because the model never actually learned to diagnose an error, only to keep generating.

Reflection, the authors argue, should be an explicit, controllable, trainable step rather than a hopeful byproduct of longer reasoning traces.

## The Reflect, Call, Final loop

Structured reflection reformats every training trajectory as three tagged segments the policy must produce in order:

_Preview only. Full text requires an AlphaSignal Pro account._

## Sources

- [Failure Makes the Agent Stronger: Enhancing Accuracy through Structured Reflection for Reliable Tool Interactions](https://arxiv.org/abs/2509.18847)
- [Tool-Reflection-Bench GitHub repository](https://github.com/MeiGen-AI/Tool-Reflection-Bench)
- [ACL Findings PDF of the paper](https://aclanthology.org/2026.findings-acl.618.pdf)
- [Berkeley Function Calling Leaderboard (BFCL)](https://gorilla.cs.berkeley.edu/leaderboard.html)