Hugging Face Daily Papers · September 3, 2026 · 4 min read

Aspire: Can Models Self-Evolve from Vague Goals?

Mirrored from Hugging Face Daily Papers for archival readability. Support the source by reading on the original site.

Like Read original ↗

We introduce ASPIRE, a benchmark that asks whether models can self-evolve from only a vague capability goal, without access to downstream evaluation tasks. Across 520 expert-authored items spanning six goals, agents can complete training and harness-editing loops, but their gains remain sparse and unstable because their chosen data and self-evaluations often fail to match the hidden target. Project page: <a href=\"https://self-developing-agents.github.io/\" rel=\"nofollow\">https://self-developing-agents.github.io/</a></p>\n","updatedAt":"2026-09-03T03:41:40.775Z","author":{"_id":"63369da91ba5d5ece24118a4","avatarUrl":"/avatars/67889e1ecadb04100a77bc8b5284c6fd.svg","fullname":"wuyuhao","name":"mozhu","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":6,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.9181742072105408},"editors":["mozhu"],"editorAvatarUrls":["/avatars/67889e1ecadb04100a77bc8b5284c6fd.svg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2608.31111","authors":[{"_id":"6a98ec4bfea818274321ff10","name":"Yuhao Wu","hidden":false},{"_id":"6a98ec4bfea818274321ff11","name":"Jingyuan Zhang","hidden":false},{"_id":"6a98ec4bfea818274321ff12","name":"Jiajun Shi","hidden":false},{"_id":"6a98ec4bfea818274321ff13","name":"Yuxuan Zhang","hidden":false},{"_id":"6a98ec4bfea818274321ff14","name":"Xinping Lei","hidden":false},{"_id":"6a98ec4bfea818274321ff15","name":"Junting Zhou","hidden":false},{"_id":"6a98ec4bfea818274321ff16","name":"Zexuan Wang","hidden":false},{"_id":"6a98ec4bfea818274321ff17","name":"Yuchen Wu","hidden":false},{"_id":"6a98ec4bfea818274321ff18","name":"Huan Zhou","hidden":false},{"_id":"6a98ec4bfea818274321ff19","name":"Duo Wang","hidden":false},{"_id":"6a98ec4bfea818274321ff1a","name":"Yinzhu Piao","hidden":false},{"_id":"6a98ec4bfea818274321ff1b","name":"Yongchang Peng","hidden":false},{"_id":"6a98ec4bfea818274321ff1c","name":"Yunfeng Shi","hidden":false},{"_id":"6a98ec4bfea818274321ff1d","name":"Jin Chen","hidden":false},{"_id":"6a98ec4bfea818274321ff1e","name":"Zuo Wang","hidden":false},{"_id":"6a98ec4bfea818274321ff1f","name":"Jinkai Liu","hidden":false},{"_id":"6a98ec4bfea818274321ff20","name":"Jiaheng Liu","hidden":false},{"_id":"6a98ec4bfea818274321ff21","name":"Wenxuan Zhang","hidden":false},{"_id":"6a98ec4bfea818274321ff22","name":"Shen Yan","hidden":false},{"_id":"6a98ec4bfea818274321ff23","name":"Wenhao Huang","hidden":false},{"_id":"6a98ec4bfea818274321ff24","name":"Ge Zhang","hidden":false}],"publishedAt":"2026-08-31T00:00:00.000Z","submittedOnDailyAt":"2026-09-03T00:00:00.000Z","title":"Aspire: Can Models Self-Evolve from Vague Goals?","submittedOnDailyBy":{"_id":"63369da91ba5d5ece24118a4","avatarUrl":"/avatars/67889e1ecadb04100a77bc8b5284c6fd.svg","isPro":false,"fullname":"wuyuhao","user":"mozhu","type":"user","name":"mozhu"},"summary":"Many important forms of human learning begin with a vague goal, such as \"become a better physicist\" or \"improve at research.\" Learners must interpret the goal, identify capability gaps, decide how to learn, and determine whether they have actually improved. In contrast, existing work on LLM self-evolution typically begins with tasks and evaluation metrics specified by humans, reducing self-evolution to optimizing an explicit objective rather than deciding what and how to learn. We introduce ASPIRE, a benchmark for vague-goal-driven self-evolution. ASPIRE provides only a natural-language capability goal while downstream evaluation tasks remain hidden. The agent must operationalize the goal by choosing data and update methods, constructing training and validation signals, and deciding when to evaluate. ASPIRE supports both model-weight and agent-harness evolution in a unified interactive environment and evaluates the resulting systems on a hidden, expert-authored set of 520 items spanning six goals. Our experiments show that vague goals redirect search effort toward goal interpretation. Current agents routinely complete training and harness-editing loops, but weight-level gains remain sparse and unstable, and the strongest evolved harness remains below the engineered Qwen-Agent reference. Agents often train on mismatched data and trust narrow self-evaluations, so local gains fail to transfer to hidden evaluation and continued search and training can erase earlier improvements.","upvotes":13,"discussionId":"6a98ec4bfea818274321ff25","projectPage":"https://self-developing-agents.github.io/","ai_summary":"ASPIRE introduces a benchmark for self-evolving LLM agents from vague natural-language goals, revealing challenges in goal interpretation, data selection, and stable weight-level improvement.","ai_keywords":["LLM self-evolution","vague-goal-driven self-evolution","ASPIRE","model-weight evolution","agent-harness evolution","hidden evaluation","self-evaluation"],"ai_summary_model":"thinkingmachines/Inkling-Small","organization":{"_id":"67d1140985ea0644e2f14b99","name":"ByteDance-Seed","fullname":"ByteDance Seed","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/6535c9e88bde2fae19b6fb25/flkDUqd_YEuFsjeNET3r-.png"}},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"63369da91ba5d5ece24118a4","avatarUrl":"/avatars/67889e1ecadb04100a77bc8b5284c6fd.svg","isPro":false,"fullname":"wuyuhao","user":"mozhu","type":"user"},{"_id":"6468ed2ae134d050a58b4c4e","avatarUrl":"/avatars/e54f43f80b5da6ab12ace5d34fef3d1f.svg","isPro":false,"fullname":"mathsfan-spark","user":"ml-llm-12345","type":"user"},{"_id":"685a9f5b6f785247e5ed562f","avatarUrl":"/avatars/b391f5c045d40ee5afc69d072a4d7f3a.svg","isPro":false,"fullname":"ZexuanWang","user":"ZexuanWang","type":"user"},{"_id":"639ae8dfb49b726255975f86","avatarUrl":"/avatars/3361477fb2de29eaea5484696b2721c6.svg","isPro":false,"fullname":"xushaoyang","user":"beiweixiaoxu","type":"user"},{"_id":"636f526a6cd69d9a36ff2b53","avatarUrl":"/avatars/8f2271a193fcac609d9be270552b5afa.svg","isPro":false,"fullname":"Qiguang Chen","user":"LightChen2333","type":"user"},{"_id":"64f0b56a87d05e740359279d","avatarUrl":"/avatars/8934a2e2bf4bcbbc3ae9d75e318c80b7.svg","isPro":false,"fullname":"Yiyang Zhang","user":"Anti-clockwise","type":"user"},{"_id":"638efcf4c67af472d316d424","avatarUrl":"/avatars/97a57859d7d87a3a8f1bb41d32a72bc2.svg","isPro":false,"fullname":"Ge Zhang","user":"zhangysk","type":"user"},{"_id":"6a2da6c8ca070ee12c6e396c","avatarUrl":"/avatars/0355287dcabaa67dbc7f0b10b87451f9.svg","isPro":false,"fullname":"Joe Mama","user":"JoeMama123123123","type":"user"},{"_id":"63ac5701c21e60a3e9b58aa7","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/63ac5701c21e60a3e9b58aa7/g6EX7diOpuA94R2ab-rZC.png","isPro":true,"fullname":"Dipankar Sarkar","user":"dipankarsarkar","type":"user"},{"_id":"66a9a55d7cda19fabeedbb89","avatarUrl":"/avatars/8e7acdd3a9c3552fbeff882bf32f245e.svg","isPro":false,"fullname":"lxp","user":"lxpp","type":"user"},{"_id":"699eab459b93bc6afbfbc436","avatarUrl":"/avatars/b7b41f8eb052d346995328da64ab799b.svg","isPro":false,"fullname":"Hu Qianyi","user":"huqianyi","type":"user"},{"_id":"65d7b983baa72790a1151923","avatarUrl":"/avatars/938531e84ca01a0c5a2a174057e3e9c5.svg","isPro":false,"fullname":"Chen Huang","user":"Albus-Chen","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"organization":{"_id":"67d1140985ea0644e2f14b99","name":"ByteDance-Seed","fullname":"ByteDance Seed","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/6535c9e88bde2fae19b6fb25/flkDUqd_YEuFsjeNET3r-.png"},"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2608/2608.31111.md","query":{}}">

Papers

arxiv:2608.31111

Aspire: Can Models Self-Evolve from Vague Goals?

Published on Aug 31

· Submitted by

wuyuhao on Sep 3

ByteDance Seed

Upvote

Authors:

Abstract

ASPIRE introduces a benchmark for self-evolving LLM agents from vague natural-language goals, revealing challenges in goal interpretation, data selection, and stable weight-level improvement.

Generated by thinkingmachines/Inkling-Small

Many important forms of human learning begin with a vague goal, such as "become a better physicist" or "improve at research." Learners must interpret the goal, identify capability gaps, decide how to learn, and determine whether they have actually improved. In contrast, existing work on LLM self-evolution typically begins with tasks and evaluation metrics specified by humans, reducing self-evolution to optimizing an explicit objective rather than deciding what and how to learn. We introduce ASPIRE, a benchmark for vague-goal-driven self-evolution. ASPIRE provides only a natural-language capability goal while downstream evaluation tasks remain hidden. The agent must operationalize the goal by choosing data and update methods, constructing training and validation signals, and deciding when to evaluate. ASPIRE supports both model-weight and agent-harness evolution in a unified interactive environment and evaluates the resulting systems on a hidden, expert-authored set of 520 items spanning six goals. Our experiments show that vague goals redirect search effort toward goal interpretation. Current agents routinely complete training and harness-editing loops, but weight-level gains remain sparse and unstable, and the strongest evolved harness remains below the engineered Qwen-Agent reference. Agents often train on mismatched data and trust narrow self-evaluations, so local gains fail to transfer to hidden evaluation and continued search and training can erase earlier improvements.