Hugging Face Daily Papers · · 7 min read

CritICL: Inference-Time Weak-to-Strong Generalization from Small Language Model Failure Modes

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

Recent advances in inference-time scaling have significantly improved the reasoning performance of large language models (LLMs). However, these methods typically rely on repeated generation or external verification. To address this limitation, we introduce CritICL, a novel inference-time framework that improves reasoning while maintaining high efficiency. Our key insight is that LLM failure modes exhibit structured patterns across model scales within the same family. Instead of treating failures as undesirable outputs, CritICL leverages them as a source of guidance. Specifically, we utilize failure modes derived from weaker models and incorporate them into inference through critique-based in-context examples. We propose two variants: CritICL-dynamic, which adaptively predicts input-specific failure modes and retrieves critiques, and CritICL-static, which uses a global failure mode profile to provide stable guidance. Experimental results show that CritICL consistently outperforms standard in-context learning and achieves performance competitive with or superior to test-time scaling methods, while requiring significantly fewer generations and lower token cost.</p>\n","updatedAt":"2026-08-28T21:49:01.884Z","author":{"_id":"652abf5360e706730596e8f4","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/zRFJ4FjZJZGkz2wH8PPmU.jpeg","fullname":"Yinghui He","name":"yinghuihe","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":1,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.9150237441062927},"editors":["yinghuihe"],"editorAvatarUrls":["https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/zRFJ4FjZJZGkz2wH8PPmU.jpeg"],"reactions":[],"isReport":false}},{"id":"6a9232af047ffcbf7aa825c3","author":{"_id":"63d3e0e8ff1384ce6c5dd17d","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/1674830754237-63d3e0e8ff1384ce6c5dd17d.jpeg","fullname":"Librarian Bot (Bot)","name":"librarian-bot","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":378,"isUserFollowing":false},"createdAt":"2026-08-29T01:15:27.000Z","type":"comment","data":{"edited":false,"hidden":false,"latest":{"raw":"This is an automated message from the [Librarian Bot](https://huggingface.co/librarian-bots). I found the following papers similar to this paper. \n\nThe following papers were recommended by the Semantic Scholar API \n\n* [Learning from Synthetic Data without Model Collapse in Iterative Instruction Tuning](https://huggingface.co/papers/2607.17043) (2026)\n* [ThinkRetrieve: Retrieval-Augmented Reasoning Traces for Test-Time Scaling](https://huggingface.co/papers/2608.10928) (2026)\n* [Refining Over Resampling: Test-Time Self-Correction for LLM Reasoning](https://huggingface.co/papers/2608.05643) (2026)\n* [ReflectRL: Learning from Golden Negative Trajectories via Reflective-to-Direct Reasoning](https://huggingface.co/papers/2608.03972) (2026)\n* [StrategyBench: Evaluating Explicit Strategy Induction in Large Language Models](https://huggingface.co/papers/2608.23475) (2026)\n* [AI4AI at Test-Time: Strong-to-Weak Capability Transfer via Harnesses](https://huggingface.co/papers/2608.12307) (2026)\n* [SPARK: Susceptibility-Guided Profiling and Steering of Latent Reasoning States in Large Language Models](https://huggingface.co/papers/2607.10296) (2026)\n\n\n Please give a thumbs up to this comment if you found it helpful!\n\n If you want recommendations for any Paper on Hugging Face checkout [this](https://huggingface.co/spaces/librarian-bots/recommend_similar_papers) Space\n\n You can directly ask Librarian Bot for paper recommendations by tagging it in a comment: `@librarian-bot recommend`","html":"<p>This is an automated message from the <a href=\"https://huggingface.co/librarian-bots\">Librarian Bot</a>. I found the following papers similar to this paper. </p>\n<p>The following papers were recommended by the Semantic Scholar API </p>\n<ul>\n<li><a href=\"https://huggingface.co/papers/2607.17043\">Learning from Synthetic Data without Model Collapse in Iterative Instruction Tuning</a> (2026)</li>\n<li><a href=\"https://huggingface.co/papers/2608.10928\">ThinkRetrieve: Retrieval-Augmented Reasoning Traces for Test-Time Scaling</a> (2026)</li>\n<li><a href=\"https://huggingface.co/papers/2608.05643\">Refining Over Resampling: Test-Time Self-Correction for LLM Reasoning</a> (2026)</li>\n<li><a href=\"https://huggingface.co/papers/2608.03972\">ReflectRL: Learning from Golden Negative Trajectories via Reflective-to-Direct Reasoning</a> (2026)</li>\n<li><a href=\"https://huggingface.co/papers/2608.23475\">StrategyBench: Evaluating Explicit Strategy Induction in Large Language Models</a> (2026)</li>\n<li><a href=\"https://huggingface.co/papers/2608.12307\">AI4AI at Test-Time: Strong-to-Weak Capability Transfer via Harnesses</a> (2026)</li>\n<li><a href=\"https://huggingface.co/papers/2607.10296\">SPARK: Susceptibility-Guided Profiling and Steering of Latent Reasoning States in Large Language Models</a> (2026)</li>\n</ul>\n<p> Please give a thumbs up to this comment if you found it helpful!</p>\n<p> If you want recommendations for any Paper on Hugging Face checkout <a href=\"https://huggingface.co/spaces/librarian-bots/recommend_similar_papers\">this</a> Space</p>\n<p> You can directly ask Librarian Bot for paper recommendations by tagging it in a comment: <code>@librarian-bot recommend</code></p>\n","updatedAt":"2026-08-29T01:15:27.168Z","author":{"_id":"63d3e0e8ff1384ce6c5dd17d","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/1674830754237-63d3e0e8ff1384ce6c5dd17d.jpeg","fullname":"Librarian Bot (Bot)","name":"librarian-bot","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":378,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.7523523569107056},"editors":["librarian-bot"],"editorAvatarUrls":["https://cdn-avatars.huggingface.co/v1/production/uploads/1674830754237-63d3e0e8ff1384ce6c5dd17d.jpeg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2608.27455","authors":[{"_id":"6a920236073195fee51564dd","name":"Yufan Wu","hidden":false},{"_id":"6a920236073195fee51564de","name":"Yinghui He","hidden":false},{"_id":"6a920236073195fee51564df","name":"Zhengyi Hu","hidden":false},{"_id":"6a920236073195fee51564e0","name":"Lang Wei","hidden":false},{"_id":"6a920236073195fee51564e1","name":"Ruichen Li","hidden":false},{"_id":"6a920236073195fee51564e2","name":"Qifan Yang","hidden":false},{"_id":"6a920236073195fee51564e3","name":"Ting Zhu","hidden":false}],"publishedAt":"2026-08-27T00:00:00.000Z","submittedOnDailyAt":"2026-08-28T00:00:00.000Z","title":"CritICL: Inference-Time Weak-to-Strong Generalization from Small Language Model Failure Modes","submittedOnDailyBy":{"_id":"652abf5360e706730596e8f4","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/zRFJ4FjZJZGkz2wH8PPmU.jpeg","isPro":false,"fullname":"Yinghui He","user":"yinghuihe","type":"user","name":"yinghuihe"},"summary":"Recent advances in inference-time scaling have significantly improved the reasoning performance of large language models (LLMs). However, these methods typically rely on repeated generation or external verification. To address this limitation, we introduce CritICL, a novel inference-time framework that improves reasoning while maintaining high efficiency. Our key insight is that LLM failure modes exhibit structured patterns across model scales within the same family. Instead of treating failures as undesirable outputs, CritICL leverages them as a source of guidance. Specifically, we utilize failure modes derived from weaker models and incorporate them into inference through critique-based in-context examples. We propose two variants: CritICL-dynamic, which adaptively predicts input-specific failure modes and retrieves critiques, and CritICL-static, which uses a global failure mode profile to provide stable guidance. Experimental results show that CritICL consistently outperforms standard in-context learning and achieves performance competitive with or superior to test-time scaling methods, while requiring significantly fewer generations and lower token cost. Code available at: https://github.com/umwyf/CRITICL","upvotes":4,"discussionId":"6a920237073195fee51564e4","ai_summary":"CritICL improves LLM reasoning at inference time by using structured failure patterns from weaker models as critique-based guidance, reducing generation and token costs.","ai_keywords":["inference-time scaling","large language models","failure modes","critique-based in-context learning","CritICL-dynamic","CritICL-static"],"ai_summary_model":"thinkingmachines/Inkling-Small"},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"652abf5360e706730596e8f4","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/zRFJ4FjZJZGkz2wH8PPmU.jpeg","isPro":false,"fullname":"Yinghui He","user":"yinghuihe","type":"user"},{"_id":"653af56d75acbdef0a6bd807","avatarUrl":"/avatars/80dbb1f559b2b13dcabc58048a700031.svg","isPro":false,"fullname":"Yufan Wu","user":"umwyf","type":"user"},{"_id":"65d0411170354febcb82f7a3","avatarUrl":"/avatars/d7e99cb4094dc419ce9e4dfec35bbba2.svg","isPro":false,"fullname":"Dustin Shi","user":"NPCv7","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"}],"acceptLanguages":["en"],"dailyPaperRank":0,"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2608/2608.27455.md","query":{}}">
Papers
arxiv:2608.27455

CritICL: Inference-Time Weak-to-Strong Generalization from Small Language Model Failure Modes

Published on Aug 27
· Submitted by
Yinghui He
on Aug 28
Authors:
,

Abstract

CritICL improves LLM reasoning at inference time by using structured failure patterns from weaker models as critique-based guidance, reducing generation and token costs.

Recent advances in inference-time scaling have significantly improved the reasoning performance of large language models (LLMs). However, these methods typically rely on repeated generation or external verification. To address this limitation, we introduce CritICL, a novel inference-time framework that improves reasoning while maintaining high efficiency. Our key insight is that LLM failure modes exhibit structured patterns across model scales within the same family. Instead of treating failures as undesirable outputs, CritICL leverages them as a source of guidance. Specifically, we utilize failure modes derived from weaker models and incorporate them into inference through critique-based in-context examples. We propose two variants: CritICL-dynamic, which adaptively predicts input-specific failure modes and retrieves critiques, and CritICL-static, which uses a global failure mode profile to provide stable guidance. Experimental results show that CritICL consistently outperforms standard in-context learning and achieves performance competitive with or superior to test-time scaling methods, while requiring significantly fewer generations and lower token cost. Code available at: https://github.com/umwyf/CRITICL

Community

Paper submitter about 4 hours ago

Recent advances in inference-time scaling have significantly improved the reasoning performance of large language models (LLMs). However, these methods typically rely on repeated generation or external verification. To address this limitation, we introduce CritICL, a novel inference-time framework that improves reasoning while maintaining high efficiency. Our key insight is that LLM failure modes exhibit structured patterns across model scales within the same family. Instead of treating failures as undesirable outputs, CritICL leverages them as a source of guidance. Specifically, we utilize failure modes derived from weaker models and incorporate them into inference through critique-based in-context examples. We propose two variants: CritICL-dynamic, which adaptively predicts input-specific failure modes and retrieves critiques, and CritICL-static, which uses a global failure mode profile to provide stable guidance. Experimental results show that CritICL consistently outperforms standard in-context learning and achieves performance competitive with or superior to test-time scaling methods, while requiring significantly fewer generations and lower token cost.

This is an automated message from the Librarian Bot. I found the following papers similar to this paper.

The following papers were recommended by the Semantic Scholar API

Please give a thumbs up to this comment if you found it helpful!

If you want recommendations for any Paper on Hugging Face checkout this Space

You can directly ask Librarian Bot for paper recommendations by tagging it in a comment: @librarian-bot recommend

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 2608.27455
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/2608.27455 in a model README.md to link it from this page.

Datasets citing this paper

No dataset linking this paper

Cite arxiv.org/abs/2608.27455 in a dataset README.md to link it from this page.

Spaces citing this paper

No Space linking this paper

Cite arxiv.org/abs/2608.27455 in a Space README.md to link it from this page.

Collections including this paper

No Collection including this paper

Add this paper to a collection 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.

More from Hugging Face Daily Papers