ACC: Compiling Agent Trajectories for Long-Context Training

Hugging Face Daily Papers Papers
agent-context-compilation long-context-training supervised-fine-tuning llm-reasoning agent-trajectories

Summary

Agent Context Compilation (ACC) enhances long-context reasoning in LLMs by converting multi-turn agent trajectories into structured QA pairs, enabling direct supervision of distant context integration without additional annotation.

Recent development of agents has renewed demand for long-context reasoning capacity of LLMs. However, training LLMs for this capacity requires costly long-document curation or heuristic context synthesis. We observe that agents produce massive trajectories when solving problems, invoking tools and receiving environment observations across many turns. The evidence needed to answer the original question is thus scattered throughout these turns, requiring integration of distant context segments. Nevertheless, standard agent SFT masks tool responses and only trains turn-level tool selection, creating a supervision blind spot where these scattered signals go unused. We propose Agent Context Compilation (ACC), which converts trajectories from search, software engineering, and database querying agents into long-context QA pairs that combine the original question with tool responses and environment observations gathered across multiple turns, training the model to answer directly without tool use. This makes the dependencies between the question and the evidence explicit, enabling direct supervision of long-context reasoning over distant segments without additional annotation. ACC is a simple but effective approach that can be combined with any existing long-context extension or training method, providing scalable supervised fine-tuning data. We validate ACC on long-range dependency modeling tasks through MRCR and GraphWalks, challenging benchmarks requiring cross-turn coreference resolution and graph traversal over extended contexts. Training Qwen3-30B-A3B with ACC achieves 68.3 on MRCR (+18.1) and 77.5 on GraphWalks (+7.6), results comparable to Qwen3-235B-A22B, while preserving general capabilities on GPQA, MMLU-Pro, AIME, and IFEval. Further mechanism analysis reveals that the ACC-trained model exhibits task-adaptive attention restructuring and expert specialization.
Original Article
View Cached Full Text

Cached at: 05/22/26, 02:30 AM

Paper page - ACC: Compiling Agent Trajectories for Long-Context Training

Source: https://huggingface.co/papers/2605.21850

Abstract

Agent Context Compilation (ACC) enhances long-context reasoning in LLMs by converting multi-turn agent trajectories into structured QA pairs, enabling direct supervision of distant context integration without additional annotation.

Recent development of agents has renewed demand forlong-context reasoningcapacity of LLMs. However, training LLMs for this capacity requires costly long-document curation or heuristic context synthesis. We observe that agents produce massive trajectories when solving problems, invoking tools and receivingenvironment observationsacross many turns. The evidence needed to answer the original question is thus scattered throughout these turns, requiring integration of distant context segments. Nevertheless, standardagent SFTmaskstool responsesand only trains turn-level tool selection, creating a supervision blind spot where these scattered signals go unused. We propose Agent Context Compilation (ACC), which converts trajectories from search, software engineering, and database querying agents into long-context QA pairs that combine the original question withtool responsesandenvironment observationsgathered across multiple turns, training the model to answer directly without tool use. This makes the dependencies between the question and the evidence explicit, enabling direct supervision oflong-context reasoningover distant segments without additional annotation. ACC is a simple but effective approach that can be combined with any existing long-context extension or training method, providing scalablesupervised fine-tuningdata. We validate ACC onlong-range dependency modelingtasks through MRCR and GraphWalks, challenging benchmarks requiringcross-turn coreference resolutionandgraph traversalover extended contexts. Training Qwen3-30B-A3B with ACC achieves 68.3 on MRCR (+18.1) and 77.5 on GraphWalks (+7.6), results comparable to Qwen3-235B-A22B, while preserving general capabilities on GPQA, MMLU-Pro, AIME, and IFEval. Further mechanism analysis reveals that the ACC-trained model exhibits task-adaptiveattention restructuringandexpert specialization.

View arXiv pageView PDFAdd to collection

Get this paper in your agent:

hf papers read 2605\.21850

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2605.21850 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2605.21850 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2605.21850 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

Locked at the Entrance, Open Inside: Where RLVR Narrows the Solution Space

arXiv cs.LG

This paper investigates how reinforcement learning with verifiable rewards (RLVR) narrows the solution space in LLM reasoning by analyzing where diversity is lost, finding it concentrated at the 'entrance' of trajectories. It demonstrates interventions to recover breadth without compromising accuracy.