SePO is a self-evolving system prompt optimization framework that improves a prompt agent by applying the same prompt optimization procedure to the prompt agent itself.</p>\n","updatedAt":"2026-06-05T02:54:21.647Z","author":{"_id":"6520baf9a56398128b1f32b2","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6520baf9a56398128b1f32b2/BFsch4sAlpOTAq4XZk4Au.png","fullname":"Tao Wangcheng","name":"taowangcheng","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.8567072749137878},"editors":["taowangcheng"],"editorAvatarUrls":["https://cdn-avatars.huggingface.co/v1/production/uploads/6520baf9a56398128b1f32b2/BFsch4sAlpOTAq4XZk4Au.png"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2606.04465","authors":[{"_id":"6a20d00915100c5272a84614","user":{"_id":"6520baf9a56398128b1f32b2","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6520baf9a56398128b1f32b2/BFsch4sAlpOTAq4XZk4Au.png","isPro":false,"fullname":"Tao Wangcheng","user":"taowangcheng","type":"user","name":"taowangcheng"},"name":"Wangcheng Tao","status":"claimed_verified","statusLastChangedAt":"2026-06-04T12:40:02.148Z","hidden":false},{"_id":"6a20d00915100c5272a84615","name":"Han Wu","hidden":false},{"_id":"6a20d00915100c5272a84616","name":"Weng-Fai Wong","hidden":false}],"publishedAt":"2026-06-03T00:00:00.000Z","submittedOnDailyAt":"2026-06-05T00:00:00.000Z","title":"SePO: Self-Evolving Prompt Agent for System Prompt Optimization","submittedOnDailyBy":{"_id":"6520baf9a56398128b1f32b2","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6520baf9a56398128b1f32b2/BFsch4sAlpOTAq4XZk4Au.png","isPro":false,"fullname":"Tao Wangcheng","user":"taowangcheng","type":"user","name":"taowangcheng"},"summary":"System prompt optimization improves agent behavior without modifying the underlying model, yielding human-readable, model-agnostic instructions. Existing methods build a prompt agent that refines task agents' system prompts, yet leave the prompt agent's own system prompt hand-engineered and fixed. We propose Self-Evolving Prompt Optimization (SePO), which treats the prompt agent's own system prompt as an optimization target alongside task agents' system prompts. SePO adopts a self-referential design. A single prompt agent improves both task agents' system prompts and its own under an open-ended evolutionary search that maintains an archive of candidate prompts as stepping stones. Training proceeds in two stages: pre-training evolves the prompt agent on a multi-task pool, and fine-tuning then applies it to a target task. Across five benchmarks spanning math (AIME'25), abstract reasoning (ARC-AGI-1), graduate-level science (GPQA), code generation (MBPP), and logic puzzles (Sudoku), SePO consistently outperforms Manual-CoT, TextGrad, and MetaSPO, improving the average accuracy by 4.49 points compared to Manual-CoT. The prompt optimization skill from pre-training also generalizes to tasks beyond the pre-training mixture, rather than memorizing per-task prompts.","upvotes":1,"discussionId":"6a20d00a15100c5272a84617","ai_summary":"Self-Evolving Prompt Optimization (SePO) enhances agent performance by jointly optimizing both task and prompt agent system prompts through evolutionary search, demonstrating superior accuracy across diverse benchmarks.","ai_keywords":["prompt optimization","system prompt","self-referential design","evolutionary search","prompt agent","task agents","multi-task pool","fine-tuning","benchmark evaluation","generalization"],"ai_summary_model":"Qwen/Qwen2.5-Coder-32B-Instruct"},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"6520baf9a56398128b1f32b2","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6520baf9a56398128b1f32b2/BFsch4sAlpOTAq4XZk4Au.png","isPro":false,"fullname":"Tao Wangcheng","user":"taowangcheng","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2606/2606.04465.md"}">

Get this paper in your agent:

Don't have the latest CLI?

Cite arxiv.org/abs/2606.04465 in a model README.md to link it from this page.

Cite arxiv.org/abs/2606.04465 in a dataset README.md to link it from this page.

Cite arxiv.org/abs/2606.04465 in a Space README.md to link it from this page.

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