arXiv 2610.26844 · cs.AI · 2026-10-19
AnyGoal: On World Model Generalization Beyond Expert Rollouts
Heejeong Nam, Divyat Mahajan, Quentin Le Lidec, Randall Balestriero, David Klindt
galilai-group
在最新榜 #290 本站收录于 2026-10-22
一句话要点
AnyGoal 基准测试世界模型能否组合已见过的状态和动作因素，发现现有模型在这类重组上普遍失败。
AnyGoal benchmarks whether world models generalize to recombinations of familiar state and action factors; current models consistently fail.
阅读 arXiv 原文 PDF Hugging Face 讨论 代码仓库 项目页
主题与学科
同主题：评测与数据
- PremiseCheck: Efficient Fact-Checking of Human and AI Peer Reviews ▲727
- OR-License: Evaluating LLM Agents as Real-World Operations Research Practitioners ▲291
- RealCompanion: Benchmarking Human Understanding from Reasoning over Longitudinal Real-World Conversations ▲287
- GameHorizon Suite: Multi-Horizon Data and Evaluation in Gameplay ▲73
- More Choices, Fewer Decisions: Ordinal-Scale Bias in JEV-like Direct-Decision Models ▲65
数据说明
元数据来自 arXiv API（已核实）；点赞来自 Hugging Face 每日论文；引用来自 Semantic Scholar。最近刷新 2026-10-22。
要点由本站根据摘要自写，不转载摘要原文，请阅读原文。