跳到正文
HuggingFace Daily Papers(社區熱門論文)·· 2026-10-14AI 評分37

Interlock:讓長時程智能體從局部正確走向任務完成

Interlock: From Local Correctness to Task Completion in Long-Horizon Agents

AI 導讀

Interlock 將長時程任務合成為具型別的完成圖,並透過智能體驗證及自我訓練改善複雜基準測試的表現。研究以各自獲接納的軌跡微調兩個 Qwen3.6 主幹模型;AutomationBench 通過率提升 4.00-4.50 個百分點,JobBench 加權分數提高 3.20-6.64 分,Harvey LAB 評分準則通過率提升 3.89-4.16 個百分點。

正文

Published on Oct 14

Authors:

,

,

,

,

,

,

,

Abstract

Interlock synthesizes long-horizon tasks as typed completion graphs to enforce cross-component constraints, then uses agent-based verification and self-training to improve agent performance on complex benchmarks.

Large language model agents increasingly undertake long-horizon tasks in complex environments characterized by fragmented evidence, partially completed states, and dependencies across resources and artifacts. Existing agents may satisfy individual requirements while failing to maintain these relationships, producing locally correct but incomplete outcomes. These failures arise because task completion depends on relationships among environmental evidence, consequential decisions, and required effects. To address this completion gap, we propose Interlock, a framework for task synthesis that represents each task as a typed completion graph. Grounding edges connect public evidence to decisions, coupling hyperedges impose joint constraints across components, and realization edges connect decisions to all required downstream effects. Interlock compiles this graph into a resettable workspace, a public request, and an evaluation rubric. A source-coverage review checks whether the original requirements survive generation, while a public-input review checks whether the rubric is supported by the materials available to the solver. An agent-as-judge inspects execution artifacts and runs verification scripts, retaining trajectories judged to satisfy all mandatory requirements. To evaluate training utility, we fine-tune two Qwen3.6 backbones on their own accepted trajectories. Across the two backbones, training improves AutomationBench task pass rates by 4.00-4.50 percentage points, JobBench weighted scores by 3.20-6.64 points, and Harvey LAB criterion pass rates by 3.89-4.16 percentage points. These results demonstrate the training utility of the synthesized experience across external task environments.

View arXiv page View PDF Add to collection

Get this paper in your agent:

hf papers read 2610.19686

Don't have the latest CLI?

curl -LsSf https://hf.co/cli/install.sh | bash

Models citing this paper 0

No model linking this paper

Cite arxiv.org/abs/2610.19686 in a model README.md to link it from this page.

Datasets citing this paper 0

No dataset linking this paper

Cite arxiv.org/abs/2610.19686 in a dataset README.md to link it from this page.

Spaces citing this paper 0

No Space linking this paper

Cite arxiv.org/abs/2610.19686 in a Space README.md to link it from this page.

Collections including this paper 0

No Collection including this paper

Add this paper to a collection to link it from this page.

來源:HuggingFace Daily Papers(社區熱門論文) · huggingface.co