世界模型中的幻覺是可預測且可預防的

摘要

現代生成式世界模型能渲染出愈來愈逼真且可受動作控制的未來場景，但它們經常產生幻覺：雖然生成的序列在視覺上保持流暢，卻偏離了真實動態。我們假設幻覺集中在狀態-動作空間的低覆蓋區域，而輕量級的以數據為中心的信號既能偵測幻覺，也能引導緩解措施。為驗證此假設，我們提出MMBench2，一個包含427小時、210項任務的視覺世界模型資料集，具備真實動作、獎勵及即時模擬器，並在此基礎上訓練了一個3.5億參數的世界模型。我們辨識出三種不同的幻覺模式：感知幻覺、動作邊緣化幻覺與場景發散幻覺——每一種都對應到管線的不同階段，並開發出三種能精確預測模型何時會失敗的信號。為在訓練時填補覆蓋缺口，我們開發了一種覆蓋感知取樣技術；在線上情境中，我們的幻覺預測器則作為好奇心獎勵，引導有目標的數據收集，從而形成一套數據高效的微調策略，使預訓練的世界模型能用最少僅50條真實環境軌跡，適應完全未見過的環境。整體而言，我們的研究發現世界模型中的幻覺本質上是數據覆蓋問題，而用來偵測幻覺的同一組信號也可用於緩解幻覺。我們論文的互動網頁版可在 https://www.nicklashansen.com/mmbench2 取得。

English

Modern generative world models render increasingly realistic action-controllable futures, yet they frequently hallucinate: rollouts remain visually fluent while drifting from the ground-truth dynamics. We hypothesize that hallucination concentrates in low-coverage regions of the state-action space, where lightweight data-centric signals can both detect it and guide mitigation. To test this, we introduce MMBench2, a 427-hour, 210-task dataset for visual world modeling with ground-truth actions, rewards, and live simulators, and train a 350M-parameter world model on it. We identify three distinct hallucination modes: perceptual, action-marginalized, and scene-diverging -- each anchored to a different stage of the pipeline, and develop three signals that accurately predict where the model will fail. To close coverage gaps at training time, we develop a coverage-aware sampling technique; to close them online, our hallucination predictors serve as curiosity rewards for targeted data collection, yielding a data-efficient finetuning recipe that adapts the pretrained world model to entirely unseen environments with as few as 50 real environment trajectories. Overall, our findings reveal that hallucination in world models is inherently a data coverage issue, and that the same signals used to detect it can also be used for mitigation. An interactive web version of our paper is available at https://www.nicklashansen.com/mmbench2