paper-with-me

홈 › Papers

ISE: An Execution-Grounded Recipe for Multi-Turn OS-Agent Trajectories

2026-06-09 · Siyuan Luo, Nairong Zheng, Lin Zhou, Tiankuo Yao, Shengyou Yuan, Haojia Yu, Cong Pang, Jiapeng Luo, Lewei Lu arxiv

Training capable OS agents requires data that simultaneously captures structured user intents, multi-turn task delegation, and grounded tool execution--properties absent from existing datasets. We propose ISE (Intent -> Simulate -> Execute), a three-stage synthesis paradigm that addresses these gaps jointly. Stage 1 constructs roughly 50000 structured intents via a 4D framework (Persona x Domain x Task x Complexity); after deduplication the pool contains 43956 unique intents and attains a Vendi Score of 61.57 over the entire pool on mpnet-base-v2 embeddings (cosine kernel, q=1). Stage 2 drives multi-turn user-agent interaction through a role-locked user simulator that grounds each user turn in actual execution outcomes, producing 23132 complete trajectories averaging 8.12 user turns and 68.24 total dialogue turns. Stage 3 runs every tool call inside a live, isolated OS workspace, generating authentic failure-recovery dynamics instead of simulated responses. Fine-tuning on ISETrace improves ClawEval pass@1 from 19.3 to 37.7 using Qwen3-8B on agent tool-use tasks with a standard protocol. This result outperforms zero-shot GPT-4o and the larger Qwen3-32B base model which is four times bigger. An ablation on Stage 2 proves multi-turn simulation brings a large portion of the performance gain. We release all source code and dataset at https://github.com/Valiere01/ISE-Trace.

📄 PDF Abstract BibTeX arXiv:2606.11520

Code (0)

등록된 구현이 없습니다.

Similar Papers 제목 키워드 기반

From Self-Evolving Synthetic Data to Verifiable-Reward RL: Post-Training Multi-turn Interactive Tool-Using Agents

2026-01-30 · Jiaxuan Gao, Jiaao Chen, Chuyi He, Shusheng Xu 외 arxiv

Interactive tool-using agents must solve real-world tasks via multi-turn interaction with both humans and external environments, requiring dialogue state tracking, multi-step tool execution, while following complex instr…

Dialogue State TrackingReinforcement Learning

Drift-Bench: Diagnosing Cooperative Breakdowns in LLM Agents under Input Faults via Multi-Turn Interaction

2026-02-02 · Han Bao, Zheyuan Zhang, Pengcheng Jing, Zhengqing Yuan 외 arxiv

As Large Language Models transition to autonomous agents, user inputs frequently violate cooperative assumptions (e.g., implicit intent, missing parameters, false presuppositions, or ambiguous expressions), creating exec…

MTSQL-R1: Towards Long-Horizon Multi-Turn Text-to-SQL via Agentic Training

2025-10-12 · Taicheng Guo, Hai Wang, ChaoChun Liu, Mohsen Golalikhani 외 arxiv

Multi-turn Text-to-SQL aims to translate a user's conversational utterances into executable SQL while preserving dialogue coherence and grounding to the target schema. However, most existing systems only regard this task…

Semantic Parsing

SERL-SQL: Selective Hindsight Distillation for Text-to-SQL Reinforcement Agentic Learning

2026-08-01 · Tao Liu, Tao Feng, Xiangheng Li, Jinwang Song 외 arxiv

Recent Text-to-SQL systems increasingly rely on multi-turn interaction, execution feedback, and reinforcement learning. However, most existing methods use execution correctness only as a trajectory-level reward, which pr…

Reinforcement Learning

SIRIUS-SQL: Anchoring Multi-Candidate Text-to-SQL in Execution Feedback

2026-05-31 · Leo Luo, Haining Xie, Siqi Shen, Zhipeng Ma 외 arxiv

Text-to-SQL on complex schemas is unreliable on a single pass, so recent systems generate multiple SQL candidates and let voting filter out errors. Yet voting alone is not enough, because the multi-candidate recipe has t…