[论文] When to Assist and What to Suggest: Joint Gating and Memory-Grounded H...

论文概要

研究领域: NLP 作者: A. J. A. Jubayer Talukder, Mohammad Ariful Haque 发布时间: 2026-10-29 arXiv: 2610.40611

中文摘要

世界模型可以在不忠实捕捉环境动力学的情况下实现强大的控制性能，这限制了它们在机械可解释性和科学发现方面的潜力。在这里，我们介绍了 world2vec，一种无需奖励信息、直接从观察和动作中学习机械性世界模型的方法。world2vec 学习能够忠实重现真实系统长期预测动态的潜在动力学，参数量比 DreamerV3、TD-MPC2 和 LeWM 少至少一个数量级。这使得它在基于像素控制连续环境方面达到了最先进的性能，并在包含71个吸引子的基准上实现了准确的系统辨识，展示了世界模型如何适用于科学发现。

原文摘要

Proactive assistants offer short, memory-grounded hints during a conversation, deciding when to assist and what to suggest. Current systems split this between a gate and a larger generator; we evaluate both together on 28,058 decision points. Stage-wise generation scores are oracle-gate ceilings: behind a real gate, end-to-end exact match is 0.2619 against a ceiling of 0.2971, and end-to-end precision is 0.2387, leaving substantial headroom. A gate pays off through precision and compute, whose saving falls from 4.39x with an 8B generator to 1.38x with a 1B one. The memory profile drives hint quality: removing it costs 24.7 exact-match points (rescaled BERTScore 0.57 to 0.35), while retrieval and a larger generator add nothing on exact match. One 1B decoder can do both, gating from an inter...

#论文 #arXiv #NLP #小凯