Paper page - JIT-Agent: Scaling Harness Intelligence via Just-in-Time Harness Evolution

Source: https://huggingface.co/papers/2608.25593 Published on Aug 26

·

Submitted byhttps://huggingface.co/greeky

gbzon Aug 27

#3 Paper of the day Authors:

,

,

,

,

,

,

,

,

,

,

,

,

,

,

Abstract

JIT-Agent is a trainable model that synthesizes adaptive agent harnesses for off-the-shelf LLMs, improving performance across diverse models and tasks.

Agent capability is not determined by the model alone. Theagent harness, encompassing memory management, planning strategy, action protocol, and tool/skill orchestration, can dominate the contribution of the underlying foundation model. Yet harness design remains manual, task-specific, and fundamentally unscalable. We presentJIT-Agent, aharness intelligencemodel trained to synthesize task-adaptiveagent harnesses on the fly for arbitrary off-the-shelf agentic LLMs. We formalize theagent harnessas a composable, machine-generatable artifact governed by a fixedfour-module protocol, and trainJIT-Agentto customize harnesses for a given task at hand, repair harnesses for stable and reliable execution, andself-evolveby distilling performance signals from an expanding archive of prior harness configurations. Equipped withJIT-Agentas a harness helper, DeepSeek-V4-Flash surpasses GPT-5.6 onDeepSearchQA(+9.1) andOdysseyBench(+4.3), while the already strong GLM-5.2 gains up to +20.2 points. Across controlled evaluations,JIT-Agent-generated harnesses are performance-competitive with mature agent runtimes such asOpenCodeandClaude Codeand consistently improve multi-scale model families of DeepSeek V4, Mimo-V2.5, and Qwen3.6. To our knowledge,JIT-Agentis the first model purpose-built for just-in-time harness generation, establishingharness intelligenceas a trainable, transferable, and compounding dimension of agent capability orthogonal to model scaling.

View arXiv pageView PDFProject pageGitHub0Add to collection

Get this paper in your agent:

hf papers read 2608\.25593

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/2608.25593 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2608.25593 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2608.25593 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.