Learn from What We HAVE: H istory- A ware VE rifier that Reasons about Past Interactions Online
Yishu Li, Xinyi Mao, Ying Yuan, Kyutae Sim, Ben Eisner, David Held · 2025
Read Learn from What We HAVE: H istory- A ware VE rifier that Reasons about Past Interactions Online by Yishu Li, Xinyi Mao, Ying Yuan, Kyutae Sim, Ben Eisner, David Held (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2509-00271.txt.
Description
Yishu Li, Xinyi Mao, Ying Yuan, Kyutae Sim, Ben Eisner, and David Held present HAVE in 2025, a history-aware verifier that helps robots pick actions under ambiguity. Robots often face objects whose handling outcome is unclear until touched. Generative models can propose actions, but even with action history they tend to choose poorly in ambiguous cases.
HAVE splits action generation from verification. A diffusion-based generator proposes many candidate actions with no conditions. The verifier then picks the most promising one by reasoning over past interactions with the object. Theory shows a verifier lifts expected action quality. Tests in simulated and real settings back the gain, with analysis of where history helps most.
Catalog record
| CREATOR | Yishu Li, Xinyi Mao, Ying Yuan, Kyutae Sim, Ben Eisner, David Held |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 82 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2509-00271