LLM agents are increasingly expected to operate across heterogeneous task regimes that require distinct execution paradigms. This challenges fixed agent systems and motivates system-level meta-adaptation beyond isolated component updates. While existing works have adapted external harness or trained underlying reasoning policies, full-system adaptation remains insufficiently characterized. The adaptation space between structure and execution is rarely made explicit, and the compatibility between the external harness and the internal reasoner is not optimized jointly. We propose HarnessForge, a meta-adaptive framework for evolving LLM agent systems. HarnessForge formulates an agent system as a harness--policy pair, defining a stable adaptation space that separates harness-level execution structure from policy-level reasoning behavior. It then performs harness--policy co-evolution through fault-guided harness tailoring and harness-conditioned policy alignment. Experiments across five benchmarks from diverse domains show that HarnessForge consistently improves both Qwen3-4B and Qwen3-8B backbones, outperforming harness-only and policy-only baselines with gains of up to 12.0% over the strongest baseline and achieving favorable rollout-efficiency tradeoffs, demonstrating that harness--policy co-evolution is effective, and that executable compatibility between the harness and reasoning policy is essential for agent-system adaptation.</p>\n","updatedAt":"2026-06-08T08:00:48.824Z","author":{"_id":"68a41112324114890b7324e9","avatarUrl":"/avatars/35051b00d6d2b2ec9325cc630d6c1e52.svg","fullname":"Chen Mingju","name":"mingju0303","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.900617778301239},"editors":["mingju0303"],"editorAvatarUrls":["/avatars/35051b00d6d2b2ec9325cc630d6c1e52.svg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2606.01779","authors":[{"_id":"6a1fa699e292c1c78ecb1381","name":"Mingju Chen","hidden":false},{"_id":"6a1fa699e292c1c78ecb1382","name":"Can Lv","hidden":false},{"_id":"6a1fa699e292c1c78ecb1383","name":"Guibin Zhang","hidden":false},{"_id":"6a1fa699e292c1c78ecb1384","name":"Heng Chang","hidden":false},{"_id":"6a1fa699e292c1c78ecb1385","name":"Shiji Zhou","hidden":false}],"publishedAt":"2026-06-01T07:00:35.000Z","submittedOnDailyAt":"2026-06-08T00:00:00.000Z","title":"HarnessForge: Joint Harness and Policy Evolution for Adaptive Agent Systems","submittedOnDailyBy":{"_id":"68a41112324114890b7324e9","avatarUrl":"/avatars/35051b00d6d2b2ec9325cc630d6c1e52.svg","isPro":false,"fullname":"Chen Mingju","user":"mingju0303","type":"user","name":"mingju0303"},"summary":"LLM agents are increasingly expected to operate across heterogeneous task regimes that require distinct execution paradigms. This challenges fixed agent systems and motivates system-level meta-adaptation beyond isolated component updates. While existing works have adapted external harness or trained underlying reasoning policies, full-system adaptation remains insufficiently characterized. The adaptation space between structure and execution is rarely made explicit, and the compatibility between the external harness and the internal reasoner is not optimized jointly. We propose HarnessForge, a meta-adaptive framework for evolving LLM agent systems. HarnessForge formulates an agent system as a harness--policy pair, defining a stable adaptation space that separates harness-level execution structure from policy-level reasoning behavior. It then performs harness--policy co-evolution through fault-guided harness tailoring and harness-conditioned policy alignment. Experiments across five benchmarks from diverse domains show that HarnessForge consistently improves both Qwen3-4B and Qwen3-8B backbones, outperforming harness-only and policy-only baselines with gains of up to 12.0\\% over the strongest baseline and achieving favorable rollout-efficiency tradeoffs, demonstrating that harness--policy co-evolution is effective, and that executable compatibility between the harness and reasoning policy is essential for agent-system adaptation. The code is available at https://github.com/mingju-c/HarnessForge.","upvotes":4,"discussionId":"6a1fa699e292c1c78ecb1386","githubRepo":"https://github.com/mingju-c/HarnessForge","githubRepoAddedBy":"user","ai_summary":"LLM agents face challenges in heterogeneous task regimes requiring distinct execution paradigms, prompting the need for system-level meta-adaptation that goes beyond component updates.","ai_keywords":["LLM agents","meta-adaptation","agent systems","harness-policy pair","co-evolution","fault-guided harness tailoring","harness-conditioned policy alignment"],"ai_summary_model":"Qwen/Qwen2.5-Coder-32B-Instruct","githubStars":8},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"68a41112324114890b7324e9","avatarUrl":"/avatars/35051b00d6d2b2ec9325cc630d6c1e52.svg","isPro":false,"fullname":"Chen Mingju","user":"mingju0303","type":"user"},{"_id":"6407e5294edf9f5c4fd32228","avatarUrl":"/avatars/8e2d55460e9fe9c426eb552baf4b2cb0.svg","isPro":false,"fullname":"Stoney Kang","user":"sikang99","type":"user"},{"_id":"620783f24e28382272337ba4","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/620783f24e28382272337ba4/zkUveQPNiDfYjgGhuFErj.jpeg","isPro":false,"fullname":"GuoLiangTang","user":"Tommy930","type":"user"},{"_id":"651c80a26ba9ab9b9582c273","avatarUrl":"/avatars/e963452eafd21f517d800f2e58e0f918.svg","isPro":false,"fullname":"siyeng feng","user":"siyengfeng","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2606/2606.01779.md"}">
HarnessForge: Joint Harness and Policy Evolution for Adaptive Agent Systems
Abstract
LLM agents face challenges in heterogeneous task regimes requiring distinct execution paradigms, prompting the need for system-level meta-adaptation that goes beyond component updates.
LLM agents are increasingly expected to operate across heterogeneous task regimes that require distinct execution paradigms. This challenges fixed agent systems and motivates system-level meta-adaptation beyond isolated component updates. While existing works have adapted external harness or trained underlying reasoning policies, full-system adaptation remains insufficiently characterized. The adaptation space between structure and execution is rarely made explicit, and the compatibility between the external harness and the internal reasoner is not optimized jointly. We propose HarnessForge, a meta-adaptive framework for evolving LLM agent systems. HarnessForge formulates an agent system as a harness--policy pair, defining a stable adaptation space that separates harness-level execution structure from policy-level reasoning behavior. It then performs harness--policy co-evolution through fault-guided harness tailoring and harness-conditioned policy alignment. Experiments across five benchmarks from diverse domains show that HarnessForge consistently improves both Qwen3-4B and Qwen3-8B backbones, outperforming harness-only and policy-only baselines with gains of up to 12.0\% over the strongest baseline and achieving favorable rollout-efficiency tradeoffs, demonstrating that harness--policy co-evolution is effective, and that executable compatibility between the harness and reasoning policy is essential for agent-system adaptation. The code is available at https://github.com/mingju-c/HarnessForge.
Community
LLM agents are increasingly expected to operate across heterogeneous task regimes that require distinct execution paradigms. This challenges fixed agent systems and motivates system-level meta-adaptation beyond isolated component updates. While existing works have adapted external harness or trained underlying reasoning policies, full-system adaptation remains insufficiently characterized. The adaptation space between structure and execution is rarely made explicit, and the compatibility between the external harness and the internal reasoner is not optimized jointly. We propose HarnessForge, a meta-adaptive framework for evolving LLM agent systems. HarnessForge formulates an agent system as a harness--policy pair, defining a stable adaptation space that separates harness-level execution structure from policy-level reasoning behavior. It then performs harness--policy co-evolution through fault-guided harness tailoring and harness-conditioned policy alignment. Experiments across five benchmarks from diverse domains show that HarnessForge consistently improves both Qwen3-4B and Qwen3-8B backbones, outperforming harness-only and policy-only baselines with gains of up to 12.0% over the strongest baseline and achieving favorable rollout-efficiency tradeoffs, demonstrating that harness--policy co-evolution is effective, and that executable compatibility between the harness and reasoning policy is essential for agent-system adaptation.
Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.
Tap or paste here to upload images
Cite arxiv.org/abs/2606.01779 in a model README.md to link it from this page.
Cite arxiv.org/abs/2606.01779 in a dataset README.md to link it from this page.
Cite arxiv.org/abs/2606.01779 in a Space README.md to link it from this page.
Discussion (0)
Sign in to join the discussion. Free account, 30 seconds — email code or GitHub.
Sign in →No comments yet. Sign in and be the first to say something.