LLM-based autonomous agents have demonstrated strong capabilities in reasoning, planning, and tool use, yet remain limited when tasks require sustained coordination across roles, tools, and environments. Multi-agent systems address this through structured collaboration among specialized agents, but tighter coordination also amplifies a less explored risk: errors can propagate across agents and interaction rounds, producing failures that are difficult to diagnose and rarely translate into structural self-improvement. Existing surveys cover individual agent capabilities, multi-agent collaboration, or agent self-evolution separately, leaving the causal dependencies among them unexamined. This survey provides a unified review organized around four causally linked stages, which we term the LIFE progression: Lay the capability foundation, Integrate agents through collaboration, Find faults through attribution, and Evolve through autonomous self-improvement. For each stage, we provide systematic taxonomies and formally characterize the dependencies between adjacent stages, revealing how each stage both depends on and constrains the next. Beyond synthesizing existing work, we identify open challenges at stage boundaries and propose a cross-stage research agenda for closed-loop multi-agent systems capable of continuously diagnosing failures, reorganizing structures, and refining agent behaviors, extending current coordination frameworks toward more self-organizing forms of collective intelligence. By bridging these previously fragmented research threads, this survey aims to offer both a systematic reference and a conceptual roadmap toward autonomous, self-improving multi-agent intelligence.</p>\n","updatedAt":"2026-05-15T01:54:39.549Z","author":{"_id":"67a7099286a55d5569acb213","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/rYISkvTtyraUdbgSsNfpC.png","fullname":"JieMa","name":"JamesMile","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.9115707874298096},"editors":["JamesMile"],"editorAvatarUrls":["https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/rYISkvTtyraUdbgSsNfpC.png"],"reactions":[],"isReport":false}},{"id":"6a06db45f73024302cbe8406","author":{"_id":"67a7099286a55d5569acb213","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/rYISkvTtyraUdbgSsNfpC.png","fullname":"JieMa","name":"JamesMile","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"isUserFollowing":false},"createdAt":"2026-05-15T08:37:25.000Z","type":"comment","data":{"edited":true,"hidden":true,"hiddenBy":"","hiddenReason":"Off-Topic","latest":{"raw":"This comment has been hidden","html":"This comment has been hidden","updatedAt":"2026-05-15T08:38:21.603Z","author":{"_id":"67a7099286a55d5569acb213","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/rYISkvTtyraUdbgSsNfpC.png","fullname":"JieMa","name":"JamesMile","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"isUserFollowing":false}},"numEdits":1,"editors":[],"editorAvatarUrls":[],"reactions":[]}}],"primaryEmailConfirmed":false,"paper":{"id":"2605.14892","authors":[{"_id":"6a067b93b1a8cbabc9f0980d","name":"Shihao Qi","hidden":false},{"_id":"6a067b93b1a8cbabc9f0980e","name":"Jie Ma","hidden":false},{"_id":"6a067b93b1a8cbabc9f0980f","name":"Rui Xing","hidden":false},{"_id":"6a067b93b1a8cbabc9f09810","name":"Wei Guo","hidden":false},{"_id":"6a067b93b1a8cbabc9f09811","name":"Xiao Huang","hidden":false},{"_id":"6a067b93b1a8cbabc9f09812","name":"Zhitao Gao","hidden":false},{"_id":"6a067b93b1a8cbabc9f09813","name":"Jianhao Deng","hidden":false},{"_id":"6a067b93b1a8cbabc9f09814","name":"Jun Liu","hidden":false},{"_id":"6a067b93b1a8cbabc9f09815","name":"Lingling Zhang","hidden":false},{"_id":"6a067b93b1a8cbabc9f09816","name":"Bifan Wei","hidden":false},{"_id":"6a067b93b1a8cbabc9f09817","name":"Boqian Yang","hidden":false},{"_id":"6a067b93b1a8cbabc9f09818","name":"Pinghui Wang","hidden":false},{"_id":"6a067b93b1a8cbabc9f09819","name":"Jianwen Sun","hidden":false},{"_id":"6a067b93b1a8cbabc9f0981a","name":"Jing Tao","hidden":false},{"_id":"6a067b93b1a8cbabc9f0981b","name":"Yaqiang Wu","hidden":false},{"_id":"6a067b93b1a8cbabc9f0981c","name":"Hui Liu","hidden":false},{"_id":"6a067b93b1a8cbabc9f0981d","name":"Yu Yao","hidden":false},{"_id":"6a067b93b1a8cbabc9f0981e","name":"Tongliang Liu","hidden":false}],"publishedAt":"2026-05-14T00:00:00.000Z","submittedOnDailyAt":"2026-05-15T00:00:00.000Z","title":"Beyond Individual Intelligence: Surveying Collaboration, Failure Attribution, and Self-Evolution in LLM-based Multi-Agent Systems","submittedOnDailyBy":{"_id":"67a7099286a55d5569acb213","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/rYISkvTtyraUdbgSsNfpC.png","isPro":false,"fullname":"JieMa","user":"JamesMile","type":"user","name":"JamesMile"},"summary":"LLM-based autonomous agents have demonstrated strong capabilities in reasoning, planning, and tool use, yet remain limited when tasks require sustained coordination across roles, tools, and environments. Multi-agent systems address this through structured collaboration among specialized agents, but tighter coordination also amplifies a less explored risk: errors can propagate across agents and interaction rounds, producing failures that are difficult to diagnose and rarely translate into structural self-improvement. Existing surveys cover individual agent capabilities, multi-agent collaboration, or agent self-evolution separately, leaving the causal dependencies among them unexamined. This survey provides a unified review organized around four causally linked stages, which we term the LIFE progression: Lay the capability foundation, Integrate agents through collaboration, Find faults through attribution, and Evolve through autonomous self-improvement. For each stage, we provide systematic taxonomies and formally characterize the dependencies between adjacent stages, revealing how each stage both depends on and constrains the next. Beyond synthesizing existing work, we identify open challenges at stage boundaries and propose a cross-stage research agenda for closed-loop multi-agent systems capable of continuously diagnosing failures, reorganizing structures, and refining agent behaviors, extending current coordination frameworks toward more self-organizing forms of collective intelligence. By bridging these previously fragmented research threads, this survey aims to offer both a systematic reference and a conceptual roadmap toward autonomous, self-improving multi-agent intelligence.","upvotes":40,"discussionId":"6a067b93b1a8cbabc9f0981f","githubRepo":"https://github.com/mira-ai-lab/awesome-mas-life","githubRepoAddedBy":"user","ai_summary":"Multi-agent systems face challenges in sustained coordination and error propagation, requiring integrated approaches that enable continuous diagnosis, reorganization, and behavioral refinement across structured collaboration stages.","ai_keywords":["multi-agent systems","autonomous agents","coordination","error propagation","self-improvement","agent collaboration","fault attribution","closed-loop systems","collective intelligence","self-organizing systems"],"githubStars":10,"organization":{"_id":"66a92d5a58cff488d93ab512","name":"XianJiaotongUniversity","fullname":"Xi'an Jiaotong University","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/66a92ba2f351acac61ba119c/6zLTkLwBLMbRLR1y7tfpC.png"}},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"67a7099286a55d5569acb213","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/rYISkvTtyraUdbgSsNfpC.png","isPro":false,"fullname":"JieMa","user":"JamesMile","type":"user"},{"_id":"69f2377ff97922d5b29e327e","avatarUrl":"/avatars/4f7fc1f2a63d2b7c8ce507f19ac25149.svg","isPro":false,"fullname":"Hanchi Zhang","user":"NianSa","type":"user"},{"_id":"67f779a6451518714c5a362f","avatarUrl":"/avatars/e32761a7d3f865ad03e2c9c5b3e3cba7.svg","isPro":false,"fullname":"Zimo Yin","user":"yzmmomo","type":"user"},{"_id":"6720a2c462fe7024459061c1","avatarUrl":"/avatars/1e9d858c7d005087b51e21bf4516e540.svg","isPro":false,"fullname":"xue haoquan","user":"tuanchan","type":"user"},{"_id":"67ebce6f58aa630d3a0eb7ab","avatarUrl":"/avatars/6ed7d9978cdb10f7501a8c54befc9bcf.svg","isPro":false,"fullname":"ZIANG YIN","user":"gyggurrtydey333","type":"user"},{"_id":"6a0417457e5c9803c741efa4","avatarUrl":"/avatars/8d4c7771295df861b48d636f41b85354.svg","isPro":false,"fullname":"tsukiaya","user":"tsukiaya","type":"user"},{"_id":"654f87ea1ac896152b19d0db","avatarUrl":"/avatars/a2f56eac14fba1bcc42adffb07e127f0.svg","isPro":false,"fullname":"CHIHO","user":"QIsHIHAO","type":"user"},{"_id":"624aa217102fcdff87bd1c47","avatarUrl":"/avatars/fbd1df830d0c9c697b218cc843a77a95.svg","isPro":false,"fullname":"ren","user":"19z1211","type":"user"},{"_id":"64da1654d0645cd60c6b1842","avatarUrl":"/avatars/147a6997aa68694178a38375e711e083.svg","isPro":false,"fullname":"LHHH","user":"LH0521","type":"user"},{"_id":"680df97412b5c23358784166","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/3Ryfb_9VnOGTvW1qbdhYM.png","isPro":false,"fullname":"sx","user":"2329269386qqcom","type":"user"},{"_id":"6826af2f466651b08078ff87","avatarUrl":"/avatars/7bb70b66e392b02f0bd5a696d28bc1c5.svg","isPro":false,"fullname":"fei","user":"xqls1210","type":"user"},{"_id":"6530da87da81958b0f5ed293","avatarUrl":"/avatars/5a8e9e96cd06597011b01461102a4719.svg","isPro":false,"fullname":"duany049","user":"duany049","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"organization":{"_id":"66a92d5a58cff488d93ab512","name":"XianJiaotongUniversity","fullname":"Xi'an Jiaotong University","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/66a92ba2f351acac61ba119c/6zLTkLwBLMbRLR1y7tfpC.png"}}">
Beyond Individual Intelligence: Surveying Collaboration, Failure Attribution, and Self-Evolution in LLM-based Multi-Agent Systems
Published on May 14
· Submitted by JieMa on May 15 Authors: ,
,
,
,
,
,
,
,
,
,
,
,
,
,
,
,
,
Abstract
Multi-agent systems face challenges in sustained coordination and error propagation, requiring integrated approaches that enable continuous diagnosis, reorganization, and behavioral refinement across structured collaboration stages.
AI-generated summary
LLM-based autonomous agents have demonstrated strong capabilities in reasoning, planning, and tool use, yet remain limited when tasks require sustained coordination across roles, tools, and environments. Multi-agent systems address this through structured collaboration among specialized agents, but tighter coordination also amplifies a less explored risk: errors can propagate across agents and interaction rounds, producing failures that are difficult to diagnose and rarely translate into structural self-improvement. Existing surveys cover individual agent capabilities, multi-agent collaboration, or agent self-evolution separately, leaving the causal dependencies among them unexamined. This survey provides a unified review organized around four causally linked stages, which we term the LIFE progression: Lay the capability foundation, Integrate agents through collaboration, Find faults through attribution, and Evolve through autonomous self-improvement. For each stage, we provide systematic taxonomies and formally characterize the dependencies between adjacent stages, revealing how each stage both depends on and constrains the next. Beyond synthesizing existing work, we identify open challenges at stage boundaries and propose a cross-stage research agenda for closed-loop multi-agent systems capable of continuously diagnosing failures, reorganizing structures, and refining agent behaviors, extending current coordination frameworks toward more self-organizing forms of collective intelligence. By bridging these previously fragmented research threads, this survey aims to offer both a systematic reference and a conceptual roadmap toward autonomous, self-improving multi-agent intelligence.
Community
LLM-based autonomous agents have demonstrated strong capabilities in reasoning, planning, and tool use, yet remain limited when tasks require sustained coordination across roles, tools, and environments. Multi-agent systems address this through structured collaboration among specialized agents, but tighter coordination also amplifies a less explored risk: errors can propagate across agents and interaction rounds, producing failures that are difficult to diagnose and rarely translate into structural self-improvement. Existing surveys cover individual agent capabilities, multi-agent collaboration, or agent self-evolution separately, leaving the causal dependencies among them unexamined. This survey provides a unified review organized around four causally linked stages, which we term the LIFE progression: Lay the capability foundation, Integrate agents through collaboration, Find faults through attribution, and Evolve through autonomous self-improvement. For each stage, we provide systematic taxonomies and formally characterize the dependencies between adjacent stages, revealing how each stage both depends on and constrains the next. Beyond synthesizing existing work, we identify open challenges at stage boundaries and propose a cross-stage research agenda for closed-loop multi-agent systems capable of continuously diagnosing failures, reorganizing structures, and refining agent behaviors, extending current coordination frameworks toward more self-organizing forms of collective intelligence. By bridging these previously fragmented research threads, this survey aims to offer both a systematic reference and a conceptual roadmap toward autonomous, self-improving multi-agent intelligence.
This comment has been hidden (marked as Off-Topic) 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/2605.14892 in a model README.md to link it from this page.
Cite arxiv.org/abs/2605.14892 in a dataset README.md to link it from this page.
Cite arxiv.org/abs/2605.14892 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.