Hugging Face Daily Papers · August 20, 2026 · 3 min read

Decision-Metric Alignment in Latent World Models: Diagnostics and Action-Conditioned Objectives for MPC Planning

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

Like Read original ↗

TL;DR: Strong latent representations are not necessarily good planning metrics. This paper introduces diagnostics for measuring whether latent distances reflect real task progress and shows that action-conditioned objectives substantially improve latent-space geometry and MPC performance—even when conventional representation probes remain unchanged.</p>\n","updatedAt":"2026-08-20T03:55:06.439Z","author":{"_id":"64060b49a577649430bf6974","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/64060b49a577649430bf6974/0YhJeunF5brCysnnEpLKG.jpeg","fullname":"Jiawei Wang","name":"Jarvis1111","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":13,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.7978218793869019},"editors":["Jarvis1111"],"editorAvatarUrls":["https://cdn-avatars.huggingface.co/v1/production/uploads/64060b49a577649430bf6974/0YhJeunF5brCysnnEpLKG.jpeg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2608.18746","authors":[{"_id":"6a8679efdb13816030683f09","name":"Jiawei Wang","hidden":false},{"_id":"6a8679efdb13816030683f0a","name":"Ke Rui","hidden":false},{"_id":"6a8679efdb13816030683f0b","name":"Yushen Zuo","hidden":false},{"_id":"6a8679efdb13816030683f0c","name":"Yichun Feng","hidden":false},{"_id":"6a8679efdb13816030683f0d","name":"Minglei Li","hidden":false}],"publishedAt":"2026-08-19T00:00:00.000Z","submittedOnDailyAt":"2026-08-20T00:00:00.000Z","title":"Decision-Metric Alignment in Latent World Models: Diagnostics and Action-Conditioned Objectives for MPC Planning","submittedOnDailyBy":{"_id":"64060b49a577649430bf6974","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/64060b49a577649430bf6974/0YhJeunF5brCysnnEpLKG.jpeg","isPro":false,"fullname":"Jiawei Wang","user":"Jarvis1111","type":"user","name":"Jarvis1111"},"summary":"JEPA-style latent world models can use Euclidean distance to a goal latent as the cost for model-predictive control (MPC). Strong decoding of task variables, however, does not guarantee that this particular cost ranks candidate action sequences by real task progress. We call the latter property decision-metric alignment. We introduce Plan-Real Spearman, which measures latent--real rank agreement on random plans, and CEM-stage Spearman, which measures the same agreement as cross-entropy-method (CEM) search concentrates its proposal. We analyze sufficient conditions under which latent distance preserves real-cost rankings, identifying encoder distortion, terminal rollout error, and candidate margins as the controlling quantities. Guided by the observed empirical alignment gap, DA-LeWM augments LeWM with inverse-dynamics and demonstration-conditioned goal-action heads. Across all our experiments, DA-LeWM accelerates convergence and achieves higher online success than LeWM, while probe scores remain similar. These results show that action-conditioned objectives improve the geometry used by Euclidean-cost, CEM-based latent MPC.","upvotes":12,"discussionId":"6a8679efdb13816030683f0e","ai_summary":"Action-conditioned objectives improve latent geometry for Euclidean-cost model-predictive control by enhancing decision-metric alignment in world models.","ai_keywords":["JEPA","latent world models","model-predictive control","decision-metric alignment","Plan-Real Spearman","CEM-stage Spearman","encoder distortion","DA-LeWM","inverse-dynamics","demonstration-conditioned goal-action"],"ai_summary_model":"thinkingmachines/Inkling-Small","organization":{"_id":"6a58b6bcf45a179ffad5c858","name":"simple-world-lab","fullname":"Simple World Lab","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/6a58b4505c2ab1e637affd07/bKcJxgIPjCUkqyOOnX91d.png"}},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"6a6a92931b3822fa45f64623","avatarUrl":"/avatars/c51da7089b42b77a78ee76de22250514.svg","isPro":false,"fullname":"Jessica Brown","user":"jessica-brown","type":"user"},{"_id":"6a6a9d2c28e0925e7302d8a1","avatarUrl":"/avatars/fdd0b6d29da96655180b28642951e87e.svg","isPro":false,"fullname":"Richard Thomas","user":"zenithridge","type":"user"},{"_id":"6a6aa3dd03669d40f667b35f","avatarUrl":"/avatars/7c6aef74f66e6a05ecf5993ba0bfe73c.svg","isPro":false,"fullname":"James Gonzalez","user":"Zenith-James","type":"user"},{"_id":"6a6c846fa3e5e6b7b047d92d","avatarUrl":"/avatars/1a3d200bdd027ce628bf7a7150c91e3c.svg","isPro":false,"fullname":"John Anderson","user":"Cedar-Kai","type":"user"},{"_id":"6a6c89729155f13f6e9253d6","avatarUrl":"/avatars/f3eae096bba4b812eaa06291ea108e11.svg","isPro":false,"fullname":"Elizabeth Taylor","user":"elizabethTaylor","type":"user"},{"_id":"6a6da7e008e6705013bc9e63","avatarUrl":"/avatars/c3d8bc73df06931ba549417b8af48e6b.svg","isPro":false,"fullname":"Linda Gonzalez","user":"NimbusLens","type":"user"},{"_id":"6a6dc6d49caa2ff7db786e2a","avatarUrl":"/avatars/efdc260d9bb75fcf331f574b1d1efbe6.svg","isPro":false,"fullname":"Jessica Gonzalez","user":"Azure-Rune","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"},{"_id":"6a6dee455f41627fe6324742","avatarUrl":"/avatars/dea18e5af39e0088a42d9a6c82c7b37a.svg","isPro":false,"fullname":"Mary Jones","user":"AtlasScope","type":"user"},{"_id":"6a701cdfb9714d274c3b746e","avatarUrl":"/avatars/57c2eaf1bed1f8ac512bdbf4db7e0c4b.svg","isPro":false,"fullname":"Richard Harris","user":"Echo-Lens","type":"user"},{"_id":"6a7e7f452beeed89a7dc205b","avatarUrl":"/avatars/d3035c865343d79b20dd7485b21a911d.svg","isPro":false,"fullname":"John Martinez","user":"Pixel-John655","type":"user"},{"_id":"6a81166e570086915074c6b2","avatarUrl":"/avatars/1cd0c1f5fcbe2da4ca59cefa510e86ed.svg","isPro":false,"fullname":"zeyu xu","user":"atlasridge","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"organization":{"_id":"6a58b6bcf45a179ffad5c858","name":"simple-world-lab","fullname":"Simple World Lab","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/6a58b4505c2ab1e637affd07/bKcJxgIPjCUkqyOOnX91d.png"},"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2608/2608.18746.md","query":{}}">

Papers

arxiv:2608.18746

Decision-Metric Alignment in Latent World Models: Diagnostics and Action-Conditioned Objectives for MPC Planning

Published on Aug 19

· Submitted by

Jiawei Wang on Aug 20

Simple World Lab

Upvote

Authors:

Abstract

Action-conditioned objectives improve latent geometry for Euclidean-cost model-predictive control by enhancing decision-metric alignment in world models.

Generated by thinkingmachines/Inkling-Small

JEPA-style latent world models can use Euclidean distance to a goal latent as the cost for model-predictive control (MPC). Strong decoding of task variables, however, does not guarantee that this particular cost ranks candidate action sequences by real task progress. We call the latter property decision-metric alignment. We introduce Plan-Real Spearman, which measures latent--real rank agreement on random plans, and CEM-stage Spearman, which measures the same agreement as cross-entropy-method (CEM) search concentrates its proposal. We analyze sufficient conditions under which latent distance preserves real-cost rankings, identifying encoder distortion, terminal rollout error, and candidate margins as the controlling quantities. Guided by the observed empirical alignment gap, DA-LeWM augments LeWM with inverse-dynamics and demonstration-conditioned goal-action heads. Across all our experiments, DA-LeWM accelerates convergence and achieves higher online success than LeWM, while probe scores remain similar. These results show that action-conditioned objectives improve the geometry used by Euclidean-cost, CEM-based latent MPC.