自然言語の曖昧なタスクにおける文脈的不確実性の能動的解決
Actively Resolving Contextual Uncertainty for Underspecified Tasks in Natural Language
LLMとオンライン構築した言語埋め込みマップを組み合わせ、閉ループ環境相互作用を通じて曖昧な指示の文脈的不確実性を能動的に解消するフレームワークCLUEを提案し、実機Spotで検証した。
詳しい要約
1. どんなもの?
2. 先行研究と比べてどこがすごい?
3. 技術・手法の肝は?
4. どうやって有効だと検証した?
5. 議論はある?
6. 次に読むべき論文は?
※ AIが要旨から生成した要約です。正確性は原文をご確認ください。
著者: Zachary Ravichandran, Jonathan Diller, Fernando Cladera, Varun Murali, George J. Pappas, Vijay Kumar
分類: cs.RO, cs.AI
原文アブストラクト
Foundation models provide robots with the ability to interpret natural language and reason about environmental context, yet most language-conditioned policies assume that goals are well-specified and that task-relevant information is provided upfront via a prior map. Operating in unfamiliar environments with underspecified tasks entails high contextual uncertainty: the robot must jointly infer what constitutes task success, what constitutes relevant information, and where (or whether) that information exists. We address these limitations via CLUE (Closed-Loop contextual Uncertainty rEsolution), a framework for actively resolving contextual uncertainty given underspecified tasks in natural language. CLUE uses an LLM-derived policy to hypothesize task-relevant concepts and potential plans. It then uses a language-embedded map, which is constructed online, to ground these hypotheses into actions. The policy sequentially evaluates hypotheses via closed-loop environment interaction and refines its plans as it gathers new information. We deploy CLUE on a Boston Dynamics Spot across three real indoor and outdoor environments spanning 15 tasks that require object disambiguation, functional inference, and occlusion reasoning. CLUE achieves a success rate within 7 percentage points of an oracle policy and outperforms an LLM-enabled planner without closed-loop feedback by a 4x margin. Supporting experiments demonstrate that simply building and then querying a language-enriched map is insufficient to resolve complex contextual planning tasks; these approaches achieve roughly one third the success rate of CLUE while requiring over 10x more VLM tokens. We provide additional information at https://zacravichandran.github.io/CLUE.