Hugging Face Daily Papers · June 23, 2026 · 4 min read

EvoEmbedding: Evolvable Representations for Long-Context Retrieval and Agentic Memory

#model-release #agents #rag #long-context

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

Like Read original ↗

EvoEmbedding: Evolvable Representations for Long-Context Retrieval and Agentic Memory\n","updatedAt":"2026-06-23T02:32:27.136Z","author":{"_id":"653b268cd1041ca9188954da","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/653b268cd1041ca9188954da/5Qy0fyQdRCIUUAzV6fg1k.png","fullname":"Chaoyou Fu","name":"BradyFU","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":8,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.5416788458824158},"editors":["BradyFU"],"editorAvatarUrls":["https://cdn-avatars.huggingface.co/v1/production/uploads/653b268cd1041ca9188954da/5Qy0fyQdRCIUUAzV6fg1k.png"],"reactions":[],"isReport":false}},{"id":"6a39f0ab09d5e5f730971b64","author":{"_id":"69b14cdeeed94a88011ce095","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/1LRe2njtqTGIsBTkOSIOT.jpeg","fullname":"Nie","name":"ClareNie","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":2,"isUserFollowing":false},"createdAt":"2026-06-23T02:34:19.000Z","type":"comment","data":{"edited":false,"hidden":false,"latest":{"raw":"paper: https://github.com/MiG-NJU/EvoEmbedding/tree/main\ndataset: https://huggingface.co/datasets/MiG-NJU/EvoTrain-180K\nmodel: https://huggingface.co/MiG-NJU/EvoEmbedding-4B","html":"paper: <a href=\"https://github.com/MiG-NJU/EvoEmbedding/tree/main\" rel=\"nofollow\">https://github.com/MiG-NJU/EvoEmbedding/tree/main</a> dataset: <a href=\"https://huggingface.co/datasets/MiG-NJU/EvoTrain-180K\">https://huggingface.co/datasets/MiG-NJU/EvoTrain-180K</a> model: <a href=\"https://huggingface.co/MiG-NJU/EvoEmbedding-4B\">https://huggingface.co/MiG-NJU/EvoEmbedding-4B</a>\n","updatedAt":"2026-06-23T02:34:19.882Z","author":{"_id":"69b14cdeeed94a88011ce095","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/1LRe2njtqTGIsBTkOSIOT.jpeg","fullname":"Nie","name":"ClareNie","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":2,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.3729773461818695},"editors":["ClareNie"],"editorAvatarUrls":["https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/1LRe2njtqTGIsBTkOSIOT.jpeg"],"reactions":[],"isReport":false}},{"id":"6a39f12bc5f3b2d7d041c334","author":{"_id":"653b268cd1041ca9188954da","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/653b268cd1041ca9188954da/5Qy0fyQdRCIUUAzV6fg1k.png","fullname":"Chaoyou Fu","name":"BradyFU","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":8,"isUserFollowing":false},"createdAt":"2026-06-23T02:36:27.000Z","type":"comment","data":{"edited":false,"hidden":false,"latest":{"raw":"Paper: https://arxiv.org/pdf/2606.21649\n\nProject Page: https://clare-nie.github.io/EvoEmbedding/\n\nGitHub: https://github.com/MiG-NJU/EvoEmbedding","html":"Paper: <a href=\"https://arxiv.org/pdf/2606.21649\" rel=\"nofollow\">https://arxiv.org/pdf/2606.21649</a>\nProject Page: <a href=\"https://clare-nie.github.io/EvoEmbedding/\" rel=\"nofollow\">https://clare-nie.github.io/EvoEmbedding/</a>\nGitHub: <a href=\"https://github.com/MiG-NJU/EvoEmbedding\" rel=\"nofollow\">https://github.com/MiG-NJU/EvoEmbedding</a>\n","updatedAt":"2026-06-23T02:36:27.946Z","author":{"_id":"653b268cd1041ca9188954da","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/653b268cd1041ca9188954da/5Qy0fyQdRCIUUAzV6fg1k.png","fullname":"Chaoyou Fu","name":"BradyFU","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":8,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.33908864855766296},"editors":["BradyFU"],"editorAvatarUrls":["https://cdn-avatars.huggingface.co/v1/production/uploads/653b268cd1041ca9188954da/5Qy0fyQdRCIUUAzV6fg1k.png"],"reactions":[{"reaction":"🚀","users":["BradyFU"],"count":1}],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2606.21649","authors":[{"_id":"6a39e797fdcd3514343bb4c2","user":{"_id":"69b14cdeeed94a88011ce095","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/1LRe2njtqTGIsBTkOSIOT.jpeg","isPro":false,"fullname":"Nie","user":"ClareNie","type":"user","name":"ClareNie"},"name":"Chang Nie","status":"claimed_verified","statusLastChangedAt":"2026-06-23T13:56:58.268Z","hidden":false},{"_id":"6a39e797fdcd3514343bb4c3","name":"Chaoyou Fu","hidden":false},{"_id":"6a39e797fdcd3514343bb4c4","name":"Junlan Feng","hidden":false},{"_id":"6a39e797fdcd3514343bb4c5","name":"Caifeng Shan","hidden":false}],"publishedAt":"2026-06-19T00:00:00.000Z","submittedOnDailyAt":"2026-06-23T00:00:00.000Z","title":"EvoEmbedding: Evolvable Representations for Long-Context Retrieval and Agentic Memory","submittedOnDailyBy":{"_id":"653b268cd1041ca9188954da","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/653b268cd1041ca9188954da/5Qy0fyQdRCIUUAzV6fg1k.png","isPro":false,"fullname":"Chaoyou Fu","user":"BradyFU","type":"user","name":"BradyFU"},"summary":"Existing embedding models are inherently static: they encode text segments in isolation, ignoring their surrounding context and temporal order. This paper introduces EvoEmbedding, a novel embedding model that generates evolvable representations for retrieval. It is tailored for long-context scenarios, where information is dynamic, sequential, and requires continuous state tracking. Our design is simple: EvoEmbedding maintains a continuously updated latent memory as it sequentially processes inputs, and uses it alongside the raw content to jointly generate evolvable embeddings. Consequently, for the same query, our model adapts its representation to retrieve distinct targets based on the evolving context, going beyond static semantic search. To equip the model with this capability, we construct EvoTrain-180K, a diverse dataset for the joint optimization of latent memory and retrieval. Furthermore, we introduce a memory queue to prevent representation collapse during recurrent encoding, alongside segment-batching techniques that tackle significant length variance and accelerate training by 3.8times. Extensive experiments show that our model not only outperforms larger-scale specialists (e.g., Qwen3-Embedding-8B and KaLM-Embedding-Gemma3-12B) across a range of long-context retrieval benchmarks, but also generalizes well to downstream tasks (e.g., personalization) with contexts 10times longer than its training window. Notably, EvoEmbedding seamlessly integrates into agentic workflows to boost performance. For instance, a naive RAG pipeline equipped with our model surpasses dedicated agentic memory systems. Project Page: https://clare-nie.github.io/EvoEmbedding.","upvotes":24,"discussionId":"6a39e798fdcd3514343bb4c6","projectPage":"https://clare-nie.github.io/EvoEmbedding/","githubRepo":"https://github.com/MiG-NJU/EvoEmbedding","githubRepoAddedBy":"user","ai_summary":"EvoEmbedding is a dynamic embedding model that generates adaptive representations by maintaining a continuously updated latent memory, enabling improved retrieval performance in long-context scenarios.","ai_keywords":["evolvable representations","latent memory","sequential processing","joint generation","representation collapse","segment-batching","long-context retrieval","agentic workflows","RAG pipeline"],"ai_summary_model":"Qwen/Qwen2.5-Coder-32B-Instruct","githubStars":21,"organization":{"_id":"638f70e8f1256a80d4288555","name":"nanjinguniv","fullname":"Nanjing University","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/638f706ef1256a80d42880f9/6M6-JzwJGiLxjIJzvCflf.png"}},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"653b268cd1041ca9188954da","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/653b268cd1041ca9188954da/5Qy0fyQdRCIUUAzV6fg1k.png","isPro":false,"fullname":"Chaoyou Fu","user":"BradyFU","type":"user"},{"_id":"69b14cdeeed94a88011ce095","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/1LRe2njtqTGIsBTkOSIOT.jpeg","isPro":false,"fullname":"Nie","user":"ClareNie","type":"user"},{"_id":"68db9b2dadc11a1b70408123","avatarUrl":"/avatars/428b4ba13ee57c598a786d9421e68a10.svg","isPro":true,"fullname":"Ruoliu Yang","user":"RuoliuYang","type":"user"},{"_id":"662ddbcee93bb7380474e0ca","avatarUrl":"/avatars/51f26787cfbe4f24272c716b62dc4569.svg","isPro":false,"fullname":"dong1314","user":"charliedong","type":"user"},{"_id":"67e766ddd3f12bae210d76a4","avatarUrl":"/avatars/19cc4d99cf9b603f9e56be142b7e0ddb.svg","isPro":false,"fullname":"Chu Wu","user":"chuwuthw","type":"user"},{"_id":"6a18f8b1af364fc647a6e3e2","avatarUrl":"/avatars/b14adb25ecb9bbe2f0dbef7633a04981.svg","isPro":false,"fullname":"Qingyuan Cao","user":"qingyuannju","type":"user"},{"_id":"64a3df5878edd17f96569747","avatarUrl":"/avatars/e5bdc0d348a9f09b4da05b663e3b710f.svg","isPro":false,"fullname":"lllj","user":"lijiang","type":"user"},{"_id":"646cd947da8e99940b6e55cf","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/646cd947da8e99940b6e55cf/9c0P0WppFqNW9pdo8LgOS.jpeg","isPro":false,"fullname":"Shengyuan Ding","user":"ChrisDing1105","type":"user"},{"_id":"6463554dd2044cd1d7c6e0bf","avatarUrl":"/avatars/d7653623117268c545a7063fec69664b.svg","isPro":false,"fullname":"Bingzheng Wei","user":"Bingzheng","type":"user"},{"_id":"6555766238df282b0f0454bd","avatarUrl":"/avatars/eab271dee045e0aaf8a65fc3fe9acafe.svg","isPro":false,"fullname":"Shaoqi Dong","user":"ShaoqiDong","type":"user"},{"_id":"658c0b0574e79b9a8e9de89a","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/658c0b0574e79b9a8e9de89a/8HeMEOT5cLEzauXGlrqF_.jpeg","isPro":false,"fullname":"Xinping Zhao","user":"Yuki131","type":"user"},{"_id":"676c04f44464f476aaa53d1c","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/k488J1893F3JGwEMvaeuh.png","isPro":false,"fullname":"Chong Xia","user":"xiac24","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"organization":{"_id":"638f70e8f1256a80d4288555","name":"nanjinguniv","fullname":"Nanjing University","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/638f706ef1256a80d42880f9/6M6-JzwJGiLxjIJzvCflf.png"},"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2606/2606.21649.md","query":{}}">

Papers

arxiv:2606.21649

EvoEmbedding: Evolvable Representations for Long-Context Retrieval and Agentic Memory

Published on Jun 19

· Submitted by

Chaoyou Fu on Jun 23

Nanjing University

Upvote

Authors:

Chang Nie ,

Abstract

EvoEmbedding is a dynamic embedding model that generates adaptive representations by maintaining a continuously updated latent memory, enabling improved retrieval performance in long-context scenarios.

Generated by Qwen/Qwen2.5-Coder-32B-Instruct

Existing embedding models are inherently static: they encode text segments in isolation, ignoring their surrounding context and temporal order. This paper introduces EvoEmbedding, a novel embedding model that generates evolvable representations for retrieval. It is tailored for long-context scenarios, where information is dynamic, sequential, and requires continuous state tracking. Our design is simple: EvoEmbedding maintains a continuously updated latent memory as it sequentially processes inputs, and uses it alongside the raw content to jointly generate evolvable embeddings. Consequently, for the same query, our model adapts its representation to retrieve distinct targets based on the evolving context, going beyond static semantic search. To equip the model with this capability, we construct EvoTrain-180K, a diverse dataset for the joint optimization of latent memory and retrieval. Furthermore, we introduce a memory queue to prevent representation collapse during recurrent encoding, alongside segment-batching techniques that tackle significant length variance and accelerate training by 3.8times. Extensive experiments show that our model not only outperforms larger-scale specialists (e.g., Qwen3-Embedding-8B and KaLM-Embedding-Gemma3-12B) across a range of long-context retrieval benchmarks, but also generalizes well to downstream tasks (e.g., personalization) with contexts 10times longer than its training window. Notably, EvoEmbedding seamlessly integrates into agentic workflows to boost performance. For instance, a naive RAG pipeline equipped with our model surpasses dedicated agentic memory systems. Project Page: https://clare-nie.github.io/EvoEmbedding.