Hugging Face Daily Papers · · 4 min read

Cross-Model Memory Transfer via Target-Side Reader Adaptation

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

We study cross-model frozen-memory transfer, showing that target-side readers can extract and align representations across model architectures and tokenizers.</p>\n","updatedAt":"2026-08-19T07:37:10.580Z","author":{"_id":"617a92e16f37340367d5d791","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/617a92e16f37340367d5d791/omgyzmaF90KBLa3YgFxhS.png","fullname":"Shaoxiong","name":"jisx","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":12,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.867068886756897},"editors":["jisx"],"editorAvatarUrls":["https://cdn-avatars.huggingface.co/v1/production/uploads/617a92e16f37340367d5d791/omgyzmaF90KBLa3YgFxhS.png"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2608.17050","authors":[{"_id":"6a855cdc536bdd3bdd48f923","name":"Mingyuan Li","hidden":false},{"_id":"6a855cdc536bdd3bdd48f924","name":"Guangsheng Yu","hidden":false},{"_id":"6a855cdc536bdd3bdd48f925","name":"Xu Wang","hidden":false},{"_id":"6a855cdc536bdd3bdd48f926","name":"Shaoxiong Ji","hidden":false}],"publishedAt":"2026-08-17T00:00:00.000Z","submittedOnDailyAt":"2026-08-19T00:00:00.000Z","title":"Cross-Model Memory Transfer via Target-Side Reader Adaptation","submittedOnDailyBy":{"_id":"617a92e16f37340367d5d791","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/617a92e16f37340367d5d791/omgyzmaF90KBLa3YgFxhS.png","isPro":false,"fullname":"Shaoxiong","user":"jisx","type":"user","name":"jisx"},"summary":"Methods for improving knowledge use in large language models typically fall into two regimes. Non-parametric retrieval offers flexible access to external knowledge, but adds retrieval latency, context overhead, and only shallow integration with the backbone. Parametric adaptation is efficient at inference time, but entangles knowledge with model weights and can be hard to update, audit, or transfer. Engram-style hashed memory occupies a middle regime: it stores learned information in an external, addressable table, yet consumes that table through a small learned reader. This raises a basic question: when such a memory is moved across backbones, what matters more, the frozen memory itself or the target-side reader? We study this question through cross-model frozen-memory extraction, in which a memory trained on a source model is frozen and attached to a different target model, with only a lightweight reader trained. Ablations show that learned memory content and correct addressing both matter, but the transferred table becomes useful only through a reader aligned to the target model. In downstream question answering tasks, a dual-layer, four-branch reader nearly closes the gap between same-model and cross-model reuse, achieving an average score of 38.8 under our controlled evaluation protocol. Moreover, when the provider reader is directly compatible with the target interface, the frozen artifact can provide substantial utility without target-side training, while optional reader adaptation yields further improvement. These results suggest that Engram can serve as a reusable external knowledge artifact, provided that the target has access to a compatible reader interface; target-side adaptation can further improve alignment when direct reader reuse is insufficient.","upvotes":2,"discussionId":"6a855cdc536bdd3bdd48f927","projectPage":"https://www.olaresearch.org/XMemTransfer/","githubRepo":"https://github.com/OLAResearch/XMemTransfer","githubRepoAddedBy":"user","ai_summary":"Cross-model reuse of frozen external memory tables depends primarily on aligning a lightweight target-side reader rather than the memory content alone, enabling reusable knowledge artifacts with optional adaptation.","ai_keywords":["Engram","hashed memory","cross-model frozen-memory extraction","reader alignment","dual-layer four-branch reader","external addressable table"],"ai_summary_model":"thinkingmachines/Inkling-Small","githubStars":0,"organization":{"_id":"6a1ffda6310029dbc8ac9adc","name":"OLAResearchX","fullname":"Omni Language AI Research","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/617a92e16f37340367d5d791/dMU_WZq7clwnvUTMragYe.png"}},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"617a92e16f37340367d5d791","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/617a92e16f37340367d5d791/omgyzmaF90KBLa3YgFxhS.png","isPro":false,"fullname":"Shaoxiong","user":"jisx","type":"user"},{"_id":"67f4d3bfa5ea8948d62539c0","avatarUrl":"/avatars/c0e600debf2339257a92595aca62c31d.svg","isPro":false,"fullname":"yuan","user":"leehenry","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"organization":{"_id":"6a1ffda6310029dbc8ac9adc","name":"OLAResearchX","fullname":"Omni Language AI Research","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/617a92e16f37340367d5d791/dMU_WZq7clwnvUTMragYe.png"},"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2608/2608.17050.md","query":{}}">
Papers
arxiv:2608.17050

Cross-Model Memory Transfer via Target-Side Reader Adaptation

Published on Aug 17
· Submitted by
Shaoxiong
on Aug 19
Authors:
,

Abstract

Cross-model reuse of frozen external memory tables depends primarily on aligning a lightweight target-side reader rather than the memory content alone, enabling reusable knowledge artifacts with optional adaptation.

Methods for improving knowledge use in large language models typically fall into two regimes. Non-parametric retrieval offers flexible access to external knowledge, but adds retrieval latency, context overhead, and only shallow integration with the backbone. Parametric adaptation is efficient at inference time, but entangles knowledge with model weights and can be hard to update, audit, or transfer. Engram-style hashed memory occupies a middle regime: it stores learned information in an external, addressable table, yet consumes that table through a small learned reader. This raises a basic question: when such a memory is moved across backbones, what matters more, the frozen memory itself or the target-side reader? We study this question through cross-model frozen-memory extraction, in which a memory trained on a source model is frozen and attached to a different target model, with only a lightweight reader trained. Ablations show that learned memory content and correct addressing both matter, but the transferred table becomes useful only through a reader aligned to the target model. In downstream question answering tasks, a dual-layer, four-branch reader nearly closes the gap between same-model and cross-model reuse, achieving an average score of 38.8 under our controlled evaluation protocol. Moreover, when the provider reader is directly compatible with the target interface, the frozen artifact can provide substantial utility without target-side training, while optional reader adaptation yields further improvement. These results suggest that Engram can serve as a reusable external knowledge artifact, provided that the target has access to a compatible reader interface; target-side adaptation can further improve alignment when direct reader reuse is insufficient.

Community

Paper submitter about 2 hours ago

We study cross-model frozen-memory transfer, showing that target-side readers can extract and align representations across model architectures and tokenizers.

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.
Tap or paste here to upload images

· Sign up or log in to comment

Get this paper in your agent:

hf papers read 2608.17050
Don't have the latest CLI?
curl -LsSf https://hf.co/cli/install.sh | bash

Models citing this paper

No model linking this paper

Cite arxiv.org/abs/2608.17050 in a model README.md to link it from this page.

Datasets citing this paper

No dataset linking this paper

Cite arxiv.org/abs/2608.17050 in a dataset README.md to link it from this page.

Spaces citing this paper

No Space linking this paper

Cite arxiv.org/abs/2608.17050 in a Space README.md to link it from this page.

Collections including this paper

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.

More from Hugging Face Daily Papers