Epistemically-guided forward-backward exploration
Núria Armengol Urpí, Marin Vlastelica, Georg Martius, Stelian Coros · 2025
Read Epistemically-guided forward-backward exploration by Núria Armengol Urpí, Marin Vlastelica, Georg Martius, Stelian Coros (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-2507-05477.txt.
Description
Nuria Armengol Urpi and three co-authors teach exploring agents to learn from their own doubt, in a 2025 arXiv paper. Some learning agents must be ready for any future task without knowing the reward in advance. A method called forward-backward representation packs that readiness into a factored form. Yet it has leaned on outside explorers to gather its experience, a mismatch the authors aim to fix.
They design scouting rules that grow straight out of the representation itself. The rules steer the agent where its knowledge of the packed form is shakiest, shrinking that doubt as it goes. In tests, this homegrown scouting cut the data needed to learn well by a wide margin against borrowed exploration methods. Code is shared.
Catalog record
| CREATOR | Núria Armengol Urpí, Marin Vlastelica, Georg Martius, Stelian Coros |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 44 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2507-05477