A language model trained on reasoning tasks learns a response distribution that mixes several distinct solution strategies, but these strategies remain implicit and entangled within the model’s response distribution, with no explicit variable for identifying or controlling the strategy used during generation. </p>\n<p>We study how to decompose this distribution into a structured, strategy-conditioned representation in which a latent variable controls the generation strategy. The key challenge is that the generator already represents the target response distribution at initialization without using the latent, leading to a severe form of posterior collapse in standard variational inference. </p>\n<p>We address this with a model-directed reconstruction objective that focuses learning on strategy-relevant response variation. On a controlled benchmark, our method recovers latents aligned with reference strategies and consistent across inputs while preserving the base model’s response distribution.</p>\n","updatedAt":"2026-07-29T19:16:53.941Z","author":{"_id":"637abd9114076b808c6a2f4f","avatarUrl":"/avatars/7fa8b275b4347dfeab73384e2aa72ea4.svg","fullname":"Awni Altabaa","name":"awni00","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":1,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.8848363757133484},"editors":["awni00"],"editorAvatarUrls":["/avatars/7fa8b275b4347dfeab73384e2aa72ea4.svg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2607.17674","authors":[{"_id":"6a67c35673f69d5af2bec4f1","name":"Awni Altabaa","hidden":false},{"_id":"6a67c35673f69d5af2bec4f2","name":"John Lafferty","hidden":false}],"publishedAt":"2026-07-20T00:00:00.000Z","submittedOnDailyAt":"2026-07-29T00:00:00.000Z","title":"Uncovering Latent Reasoning Strategies in Language Models","submittedOnDailyBy":{"_id":"637abd9114076b808c6a2f4f","avatarUrl":"/avatars/7fa8b275b4347dfeab73384e2aa72ea4.svg","isPro":false,"fullname":"Awni Altabaa","user":"awni00","type":"user","name":"awni00"},"summary":"A language model p_θ(y mid x) trained on reasoning tasks learns to solve problems via multiple distinct strategies, yet these strategies are implicit and entangled within the model's response distribution. We study the problem of decomposing the response distribution of a given pretrained language model into a structured, strategy-conditioned representation. Specifically, we learn a latent-variable factorization p_θ(y mid x) leadsto (r_ϕ(z mid x), g_ϕ(y mid x,z)), where a router r maps each input to a distribution over latent strategies z and a generator g produces the response conditioned on that strategy. A key challenge is that the generator, initialized from the base model, already represents p_θ(y mid x) without using z. Standard variational inference therefore gives the model no incentive to route information through z and can yield a severe form of posterior collapse. To address this, we propose a variational objective that measures fractional information gain relative to the base model's response loss and concentrates reconstruction pressure on tokens with high base model surprisal, encouraging z to encode strategy-relevant response variation. We introduce a benchmark of multi-strategy algorithmic tasks and show that this objective recovers latent codes aligned with distinct reference strategies while preserving the base model's response distribution.","upvotes":1,"discussionId":"6a67c35773f69d5af2bec4f3","githubRepo":"https://github.com/Awni00/latent-strategies-in-lms","githubRepoAddedBy":"user","githubStars":1,"organization":{"_id":"6756014dd83c390221a3815c","name":"YaleUniversity","fullname":"Yale University","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/6755ff00d3ff7f20aad244d2/xk0r-87S0AVfUA1XX8CG3.png"}},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_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,"organization":{"_id":"6756014dd83c390221a3815c","name":"YaleUniversity","fullname":"Yale University","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/6755ff00d3ff7f20aad244d2/xk0r-87S0AVfUA1XX8CG3.png"},"query":{}}">
Uncovering Latent Reasoning Strategies in Language Models
Abstract
A language model p_θ(y mid x) trained on reasoning tasks learns to solve problems via multiple distinct strategies, yet these strategies are implicit and entangled within the model's response distribution. We study the problem of decomposing the response distribution of a given pretrained language model into a structured, strategy-conditioned representation. Specifically, we learn a latent-variable factorization p_θ(y mid x) leadsto (r_ϕ(z mid x), g_ϕ(y mid x,z)), where a router r maps each input to a distribution over latent strategies z and a generator g produces the response conditioned on that strategy. A key challenge is that the generator, initialized from the base model, already represents p_θ(y mid x) without using z. Standard variational inference therefore gives the model no incentive to route information through z and can yield a severe form of posterior collapse. To address this, we propose a variational objective that measures fractional information gain relative to the base model's response loss and concentrates reconstruction pressure on tokens with high base model surprisal, encouraging z to encode strategy-relevant response variation. We introduce a benchmark of multi-strategy algorithmic tasks and show that this objective recovers latent codes aligned with distinct reference strategies while preserving the base model's response distribution.
Community
A language model trained on reasoning tasks learns a response distribution that mixes several distinct solution strategies, but these strategies remain implicit and entangled within the model’s response distribution, with no explicit variable for identifying or controlling the strategy used during generation.
We study how to decompose this distribution into a structured, strategy-conditioned representation in which a latent variable controls the generation strategy. The key challenge is that the generator already represents the target response distribution at initialization without using the latent, leading to a severe form of posterior collapse in standard variational inference.
We address this with a model-directed reconstruction objective that focuses learning on strategy-relevant response variation. On a controlled benchmark, our method recovers latents aligned with reference strategies and consistent across inputs while preserving the base model’s response distribution.
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/2607.17674 in a model README.md to link it from this page.
Cite arxiv.org/abs/2607.17674 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.