Hugging Face Daily Papers · · 5 min read

OCC-RAG: Optimal Cognitive Core for Faithful Question Answering

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

Recent progress in the development of language models has been defined by scale, with each generation absorbing more of the world's knowledge into its weights. However, many practical applications benefit more from robust reasoning than from extensive parametric knowledge. In this setting, task-specialized small language models (SLMs) offer a principled design choice. We introduce Optimal Cognitive Core (OCC), a family of SLMs built around this premise. As a variant of OCC, we present OCC-RAG, optimized for faithful question answering (QA) grounded in the provided context. This task directly aligns with the OCC design approach, requiring multi-hop reasoning over supplied passages while ignoring memorized knowledge. To train OCC-RAG, we implement a novel pipeline for synthesizing multi-context, multi-hop QA data at scale, producing a corpus of over three million examples targeting multi-hop reasoning, strict context faithfulness, and calibrated abstention. We release OCC-RAG-0.6B and OCC-RAG-1.7B, both mid-trained on this corpus. The models produce structured reasoning traces with source citations grounded in literal quotes from the context. Through OCC-RAG, we demonstrate that compact, task-specialized SLMs can match or exceed general-purpose models 2 - 6x their size across multi-hop reasoning (HotpotQA, MuSiQue, TAT-QA), faithfulness (ConFiQA), and refusal (MuSiQue-Un) benchmarks.</p>\n","updatedAt":"2026-06-03T12:50:56.831Z","author":{"_id":"622b1f6b9f6139daa8e998ce","avatarUrl":"/avatars/842719c100a5969be75d04da97333675.svg","fullname":"Vasily Konovalov","name":"Vasily","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":5,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.8938180208206177},"editors":["Vasily"],"editorAvatarUrls":["/avatars/842719c100a5969be75d04da97333675.svg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2606.00683","authors":[{"_id":"6a1e8d56808ddbc3c7d43f7a","name":"Maksim Savkin","hidden":false},{"_id":"6a1e8d56808ddbc3c7d43f7b","name":"Mikhail Goncharov","hidden":false},{"_id":"6a1e8d56808ddbc3c7d43f7c","name":"Alexander Gambashidze","hidden":false},{"_id":"6a1e8d56808ddbc3c7d43f7d","name":"Alla Chepurova","hidden":false},{"_id":"6a1e8d56808ddbc3c7d43f7e","name":"Dmitrii Tarasov","hidden":false},{"_id":"6a1e8d56808ddbc3c7d43f7f","name":"Nikita Andriianov","hidden":false},{"_id":"6a1e8d56808ddbc3c7d43f80","name":"Daria Pugacheva","hidden":false},{"_id":"6a1e8d56808ddbc3c7d43f81","name":"Vasily Konovalov","hidden":false},{"_id":"6a1e8d56808ddbc3c7d43f82","name":"Andrey Galichin","hidden":false},{"_id":"6a1e8d56808ddbc3c7d43f83","name":"Ivan Oseledets","hidden":false}],"mediaUrls":["https://cdn-uploads.huggingface.co/production/uploads/622b1f6b9f6139daa8e998ce/j62V3gfJzR-_N0dfgpeL_.png","https://cdn-uploads.huggingface.co/production/uploads/622b1f6b9f6139daa8e998ce/VlkpePvRGO9MsEBVdRPnB.png"],"publishedAt":"2026-05-30T00:00:00.000Z","submittedOnDailyAt":"2026-06-03T00:00:00.000Z","title":"OCC-RAG: Optimal Cognitive Core for Faithful Question Answering","submittedOnDailyBy":{"_id":"622b1f6b9f6139daa8e998ce","avatarUrl":"/avatars/842719c100a5969be75d04da97333675.svg","isPro":false,"fullname":"Vasily Konovalov","user":"Vasily","type":"user","name":"Vasily"},"summary":"Recent progress in the development of language models has been defined by scale, with each generation absorbing more of the world's knowledge into its weights. However, many practical applications benefit more from robust reasoning than from extensive parametric knowledge. In this setting, task-specialized small language models (SLMs) offer a principled design choice. We introduce Optimal Cognitive Core (OCC), a family of SLMs built around this premise. As a variant of OCC, we present OCC-RAG, optimized for faithful question answering (QA) grounded in the provided context. This task directly aligns with the OCC design approach, requiring multi-hop reasoning over supplied passages while ignoring memorized knowledge. To train OCC-RAG, we implement a novel pipeline for synthesizing multi-context, multi-hop QA data at scale, producing a corpus of over three million examples targeting multi-hop reasoning, strict context faithfulness, and calibrated abstention. We release OCC-RAG-0.6B and OCC-RAG-1.7B, both mid-trained on this corpus. The models produce structured reasoning traces with source citations grounded in literal quotes from the context. Through OCC-RAG, we demonstrate that compact, task-specialized SLMs can match or exceed general-purpose models 2 -- 6x their size across multi-hop reasoning (HotpotQA, MuSiQue, TAT-QA), faithfulness (ConFiQA), and refusal (MuSiQue-Un) benchmarks.","upvotes":32,"discussionId":"6a1e8d57808ddbc3c7d43f84","githubRepo":"https://github.com/optimal-cognitive-core/OCC-RAG","githubRepoAddedBy":"user","ai_summary":"Compact task-specialized language models demonstrate superior performance in multi-hop reasoning and faithfulness compared to larger general-purpose models through a novel training pipeline and structured reasoning traces.","ai_keywords":["language models","task-specialized models","multi-hop reasoning","question answering","context faithfulness","structured reasoning traces","source citations","HotpotQA","MuSiQue","TAT-QA","ConFiQA"],"ai_summary_model":"Qwen/Qwen2.5-Coder-32B-Instruct","githubStars":0,"organization":{"_id":"6a1866161a1da2e5937c6a94","name":"occ-ai","fullname":"OCC","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/661e44cf1d8ffc49b57ba07e/jibWvvUHa3KlYL-tXkek0.png"}},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"678bf132da686d5964ec445b","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/5j-1yiJ5OXiqLs8pGnPZ2.jpeg","isPro":false,"fullname":"Maksim","user":"mks-logic","type":"user"},{"_id":"661e44cf1d8ffc49b57ba07e","avatarUrl":"/avatars/3e937cc4f784b369b9f996ba82d1b81d.svg","isPro":false,"fullname":"Andrey Galichin","user":"andreuka18","type":"user"},{"_id":"6734e315c1aadce903f73aea","avatarUrl":"/avatars/95d95c49419372debc201cb63c354b86.svg","isPro":false,"fullname":"Li Pengyi","user":"LiPengyi29","type":"user"},{"_id":"654621f45cd5692b3a9d08cb","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/NJj1FJyYHNPRXb0QBXjv-.jpeg","isPro":false,"fullname":"Mikhail Seleznev","user":"myyycroft","type":"user"},{"_id":"643bf565b409fef15e07da48","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/643bf565b409fef15e07da48/2uYSZqdZTYcDK3fZ34nm7.jpeg","isPro":false,"fullname":"Mikhail Goncharov","user":"mishgon","type":"user"},{"_id":"622b1f6b9f6139daa8e998ce","avatarUrl":"/avatars/842719c100a5969be75d04da97333675.svg","isPro":false,"fullname":"Vasily Konovalov","user":"Vasily","type":"user"},{"_id":"6310ff34bc152fa3e810c186","avatarUrl":"/avatars/bfd63bcd81548283f5e496e3693bf143.svg","isPro":true,"fullname":"Elizaveta Goncharova","user":"Elizaveta","type":"user"},{"_id":"62bd6c6baaf1480f1aa2222e","avatarUrl":"/avatars/fd92ae2986d435a47eb1e382ac11d8e0.svg","isPro":false,"fullname":"Mikhail Salnikov","user":"msalnikov","type":"user"},{"_id":"634c72e6fe1bfa967d6c2b5c","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/634c72e6fe1bfa967d6c2b5c/WFWIAlWl-FsiJRyGxQTTx.jpeg","isPro":false,"fullname":"Nikita Sushko","user":"chameleon-lizard","type":"user"},{"_id":"641d9ffc5b4c7eb277d20d3f","avatarUrl":"/avatars/fe2c9d8abb8cddf1e61fff9a67c71ba3.svg","isPro":false,"fullname":"————","user":"sir-timio","type":"user"},{"_id":"665b10fb270e47e678f2ddf1","avatarUrl":"/avatars/1bc7a9211acf767f7bfca998c24315a0.svg","isPro":false,"fullname":"max","user":"maksimko123","type":"user"},{"_id":"6626214189e599f5163604f6","avatarUrl":"/avatars/f7aae7a8d90f629d9abb58a3c96193d6.svg","isPro":false,"fullname":"Viktoriia Zinkovich","user":"zinkovich","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":1,"organization":{"_id":"6a1866161a1da2e5937c6a94","name":"occ-ai","fullname":"OCC","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/661e44cf1d8ffc49b57ba07e/jibWvvUHa3KlYL-tXkek0.png"},"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2606/2606.00683.md"}">
Papers
arxiv:2606.00683

OCC-RAG: Optimal Cognitive Core for Faithful Question Answering

Published on May 30
· Submitted by
Vasily Konovalov
on Jun 3
#1 Paper of the day
Authors:
,
,
,
,
,
,
,
,
,

Abstract

Compact task-specialized language models demonstrate superior performance in multi-hop reasoning and faithfulness compared to larger general-purpose models through a novel training pipeline and structured reasoning traces.

Recent progress in the development of language models has been defined by scale, with each generation absorbing more of the world's knowledge into its weights. However, many practical applications benefit more from robust reasoning than from extensive parametric knowledge. In this setting, task-specialized small language models (SLMs) offer a principled design choice. We introduce Optimal Cognitive Core (OCC), a family of SLMs built around this premise. As a variant of OCC, we present OCC-RAG, optimized for faithful question answering (QA) grounded in the provided context. This task directly aligns with the OCC design approach, requiring multi-hop reasoning over supplied passages while ignoring memorized knowledge. To train OCC-RAG, we implement a novel pipeline for synthesizing multi-context, multi-hop QA data at scale, producing a corpus of over three million examples targeting multi-hop reasoning, strict context faithfulness, and calibrated abstention. We release OCC-RAG-0.6B and OCC-RAG-1.7B, both mid-trained on this corpus. The models produce structured reasoning traces with source citations grounded in literal quotes from the context. Through OCC-RAG, we demonstrate that compact, task-specialized SLMs can match or exceed general-purpose models 2 -- 6x their size across multi-hop reasoning (HotpotQA, MuSiQue, TAT-QA), faithfulness (ConFiQA), and refusal (MuSiQue-Un) benchmarks.

Community

Paper submitter about 1 hour ago

Recent progress in the development of language models has been defined by scale, with each generation absorbing more of the world's knowledge into its weights. However, many practical applications benefit more from robust reasoning than from extensive parametric knowledge. In this setting, task-specialized small language models (SLMs) offer a principled design choice. We introduce Optimal Cognitive Core (OCC), a family of SLMs built around this premise. As a variant of OCC, we present OCC-RAG, optimized for faithful question answering (QA) grounded in the provided context. This task directly aligns with the OCC design approach, requiring multi-hop reasoning over supplied passages while ignoring memorized knowledge. To train OCC-RAG, we implement a novel pipeline for synthesizing multi-context, multi-hop QA data at scale, producing a corpus of over three million examples targeting multi-hop reasoning, strict context faithfulness, and calibrated abstention. We release OCC-RAG-0.6B and OCC-RAG-1.7B, both mid-trained on this corpus. The models produce structured reasoning traces with source citations grounded in literal quotes from the context. Through OCC-RAG, we demonstrate that compact, task-specialized SLMs can match or exceed general-purpose models 2 - 6x their size across multi-hop reasoning (HotpotQA, MuSiQue, TAT-QA), faithfulness (ConFiQA), and refusal (MuSiQue-Un) benchmarks.

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 2606.00683
Don't have the latest CLI?
curl -LsSf https://hf.co/cli/install.sh | bash

Models citing this paper 2

Datasets citing this paper 0

No dataset linking this paper

Cite arxiv.org/abs/2606.00683 in a dataset README.md to link it from this page.

Spaces citing this paper 0

No Space linking this paper

Cite arxiv.org/abs/2606.00683 in a Space README.md to link it from this page.

Collections including this paper 0

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