arXiv 2610.19376 · cs.AI · 2026-10-14
DeLoop: Repeated Actions as Self-Supervision for Long-Horizon Agents
Tianle Chen, Pushkar Singh, Ketan Todi, Biswajit Paria, Qijia He, Wenqi Wang, Yuwen Tan, Jike Zhong 等 11 人
在最新榜 #305 本站收录于 2026-10-24
一句话要点
Google 的 DeLoop 处理长任务智能体“兜圈子”：重复发出已经调用过、结果完全相同的工具调用。这种循环在 Qwen、Llama、Gemini 智能体里都会出现，历史被压缩后更常见；DeLoop 用智能体自己的轨迹做自监督，把循环动作当负例来训练。
Google's DeLoop tackles loops in long-horizon agents — re-issuing earlier tool calls that return identical observations — seen natively in Qwen, Llama and Gemini agents and worse under history compression; it self-supervises from the agent's own rollouts, using looping actions as negative examples.
阅读 arXiv 原文 PDF Hugging Face 讨论 代码仓库 项目页
主题与学科
同主题：智能体
- Raven: The Harness of Harnesses for Composable Agentic Intelligence ▲748
- False Frontiers: Diagnosing and Mitigating Co-Cheating in Self-Evolving Search Agents ▲642
- Omni-IO Skills: Harnessing Your Agent Omni-Native ▲485
- BizAgentBench: Unveiling Hidden Challenges in Deploying Long-Horizon Agents for Real-World Business ▲388
- Player-1: Learning to Play from Synthetic Experience ▲347
同主题：大语言模型
- Jeva: System One Models from First Principles 2026-10-13 · ▲371
- Post-Training Leaves Behavioral Shadows on Unrelated Decisions 2026-09-24 · ▲273
- WeVisOne-VL: Unified Visual Perception from Objects to Pixels 2026-10-22 · ▲247
- Thinking Inertia: LLMs Keep Thinking When Told Not To 2026-10-08 · ▲241
- OmniEdu: Open Foundation Models for Learning and Teaching 2026-09-19 · ▲240
同机构：Google
- Harness^2: Recursive Agent Harnessing for an Open World 2026-10-09
- VIS-Ground: Video Interactive Storytelling with Contextual Grounding 2026-10-07
- VeriHarness: Scaling Agentic Verification for Long-Horizon Tasks 2026-10-01
- AIM: Agentic Idea Management for Automated Research 2026-09-29
- TabFM: A Zero-Shot Foundation Model for Tabular Data 2026-09-29
数据说明
元数据来自 arXiv API（已核实）；点赞来自 Hugging Face 每日论文；引用来自 Semantic Scholar。最近刷新 2026-10-26。
要点由本站根据摘要自写，不转载摘要原文，请阅读原文。