AI Paper Atlas

Research map / Multi-agent systems and games / LLM agents

FIRE: Learning to Navigate and Act on Real-World Files via Stateful Reinforcement Learning

Jingyuan Ma, Zheng Li, Rui Li, Siyao Song, Ying Zeng, Minghao Li, Zhifang Sui

ICML 2026 · Multi-agent systems and games › LLM agents · approach: Agentic benchmarks and orchestration

Read the paperCheck your idea against 137,217 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Formalizes File Reasoning, where agents interact with raw XLSX, PDF, DOCX, and PPTX files in a persistent sandbox, and proposes FIRE, a stateful reinforcement learning approach for it.

Problem

LLMs struggle to answer questions grounded in real-world files such as spreadsheets and slides, where evidence is spread across irregular layouts and heterogeneous formats.

Approach

Agents interact directly with unprocessed files inside a persistent sandbox, and are trained with stateful reinforcement learning on a unified data pipeline spanning multiple file formats.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Add stateful RL (GRPO in persistent sandbox) after cold-start SFT (ablation)Qwen3-32B-FIRE · File reasoning benchmarkOverall accuracy (%): 42.76 -> 50.33Improved
Add stateful RL after SFT (ablation)Qwen2.5-32B-FIRE · File reasoning benchmarkOverall accuracy (%): 40.57 -> 48.54Improved
Add stateful RL after SFT (ablation)Qwen2.5-7B-FIRE · File reasoning benchmarkOverall accuracy (%): 22.68 -> 30.81Improved
Add SFT and RL training to base Qwen3-32BQwen3-32B · File reasoning benchmarkOverall accuracy (%): 11.14 -> 50.33Improved
Add RL after SFT on tool-call count (ablation)Qwen3-32B-FIRE · File reasoning benchmarkAverage tool calls: 7.43 -> 6.49Improved
Add RL after SFT on out-of-distribution file benchmarks (ablation)FIRE models · GAIA, MEET, Spider, SSTQAAccuracyImproved
Increase maximum interaction rounds beyond 15 (ablation)GPT-5, GLM-5, Gemini-2.5-Pro, Qwen3-RL · File reasoning benchmarkCumulative accuracyNo clear effect
Do not penalize intermediate tool failures in rewardFIRE reward design · File reasoning benchmarktool call success gap between correct and incorrect tasksImproved
Compare benchmark difficulty with existing benchmarksGPT-5 · SSTQA, GAIA vs FIRE benchmarkAccuracy (%): 76.7 (SSTQA) -> 60.2 (ours)Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy