A lightweight implicit resasoning method for generative recommendation</p>\n","updatedAt":"2026-06-16T05:04:10.136Z","author":{"_id":"654f058c35d81e5153c4dce0","avatarUrl":"/avatars/87480482d8191cd63cac39e7a6a00285.svg","fullname":"Jonathan Yinhan He","name":"jonathanhe123","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":1,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.7584826946258545},"editors":["jonathanhe123"],"editorAvatarUrls":["/avatars/87480482d8191cd63cac39e7a6a00285.svg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2606.14142","authors":[{"_id":"6a30d906a0d4daae428601fe","name":"Yinhan He","hidden":false},{"_id":"6a30d906a0d4daae428601ff","name":"Liam Collins","hidden":false},{"_id":"6a30d906a0d4daae42860200","name":"Bhuvesh Kumar","hidden":false},{"_id":"6a30d906a0d4daae42860201","name":"Jundong Li","hidden":false},{"_id":"6a30d906a0d4daae42860202","name":"Neil Shah","hidden":false},{"_id":"6a30d906a0d4daae42860203","name":"Donald Loveland","hidden":false}],"publishedAt":"2026-06-15T00:00:00.000Z","submittedOnDailyAt":"2026-06-16T00:00:00.000Z","title":"Implicit Reasoning for Large Language Model-based Generative Recommendation","submittedOnDailyBy":{"_id":"654f058c35d81e5153c4dce0","avatarUrl":"/avatars/87480482d8191cd63cac39e7a6a00285.svg","isPro":false,"fullname":"Jonathan Yinhan He","user":"jonathanhe123","type":"user","name":"jonathanhe123"},"summary":"Large Language Models (LLMs) are increasingly adopted as backbones for Generative Recommendation (GR), promising access to pretrained world knowledge. Yet reliably invoking this knowledge for GR remains poorly understood. A key obstacle is that LLM-based GR typically represents items with Semantic IDs (SIDs), disrupting LLMs' natural-language reasoning interface because these tokens are unseen by the LLM during pretraining. Existing approaches address this with expensive multi-stage pipelines that ground SIDs and elicit explicit rationales, but offer limited insight into when and why each stage is necessary. In this work, we systematically decompose explicit reasoning training pipelines for LLM-based GR, revealing three key limitations: weakened world-knowledge verbalization, misalignment between SID and natural-language token embedding spaces, and sensitivity to rationale quality, all of which hurt explicit reasoning performance. To circumvent these issues, we propose PauseRec, a lightweight implicit reasoning paradigm tailored for GR. PauseRec is exceptionally practical, avoiding costly reasoning trace acquisition and reasoning alignment training, leading to a multitude of benefits: (1) it outperforms standard explicit CoT methods by up to 6.22%, (2) it reduces training cost by up to 65% GPU hours, and (3) it speeds up inference by up to 71.3%. These results position PauseRec as a lightweight alternative to explicit rationale generation, enabling more effective and efficient LLM-based GR.","upvotes":0,"discussionId":"6a30d906a0d4daae42860204","ai_summary":"Large Language Models for generative recommendation face challenges with semantic IDs disrupting natural-language reasoning, prompting a lightweight implicit reasoning approach that outperforms explicit methods while reducing computational costs.","ai_keywords":["Large Language Models","Generative Recommendation","Semantic IDs","explicit reasoning","implicit reasoning","CoT methods","training cost","inference speed"],"ai_summary_model":"Qwen/Qwen2.5-Coder-32B-Instruct","organization":{"_id":"63c87c41cd6a490608ce31d1","name":"snap-research","fullname":"Snap Research","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/1674083325534-61f19829233c91cbd2f79e70.png"}},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[],"acceptLanguages":["en"],"organization":{"_id":"63c87c41cd6a490608ce31d1","name":"snap-research","fullname":"Snap Research","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/1674083325534-61f19829233c91cbd2f79e70.png"},"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2606/2606.14142.md","query":{}}">
Implicit Reasoning for Large Language Model-based Generative Recommendation
Abstract
Large Language Models for generative recommendation face challenges with semantic IDs disrupting natural-language reasoning, prompting a lightweight implicit reasoning approach that outperforms explicit methods while reducing computational costs.
Large Language Models (LLMs) are increasingly adopted as backbones for Generative Recommendation (GR), promising access to pretrained world knowledge. Yet reliably invoking this knowledge for GR remains poorly understood. A key obstacle is that LLM-based GR typically represents items with Semantic IDs (SIDs), disrupting LLMs' natural-language reasoning interface because these tokens are unseen by the LLM during pretraining. Existing approaches address this with expensive multi-stage pipelines that ground SIDs and elicit explicit rationales, but offer limited insight into when and why each stage is necessary. In this work, we systematically decompose explicit reasoning training pipelines for LLM-based GR, revealing three key limitations: weakened world-knowledge verbalization, misalignment between SID and natural-language token embedding spaces, and sensitivity to rationale quality, all of which hurt explicit reasoning performance. To circumvent these issues, we propose PauseRec, a lightweight implicit reasoning paradigm tailored for GR. PauseRec is exceptionally practical, avoiding costly reasoning trace acquisition and reasoning alignment training, leading to a multitude of benefits: (1) it outperforms standard explicit CoT methods by up to 6.22%, (2) it reduces training cost by up to 65% GPU hours, and (3) it speeds up inference by up to 71.3%. These results position PauseRec as a lightweight alternative to explicit rationale generation, enabling more effective and efficient LLM-based GR.
Community
A lightweight implicit resasoning method for generative recommendation
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/2606.14142 in a model README.md to link it from this page.
Cite arxiv.org/abs/2606.14142 in a dataset README.md to link it from this page.
Cite arxiv.org/abs/2606.14142 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.