Search Inspired Exploration in Reinforcement Learning

Georgios Sotirchos, Zlatan Ajanović, Jens Kober · 2026

Read Search Inspired Exploration in Reinforcement Learning by Georgios Sotirchos, Zlatan Ajanović, Jens Kober (2026) 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-2602-00460.txt.

Description

Georgios Sotirchos and team propose Search-Inspired Exploration in Reinforcement Learning, a method guiding exploration via sub-goals based on learning progress. Exploration with sparse rewards remains fundamental. Curriculum learning and Go-Explore use hand crafted heuristics, while curiosity methods risk weak policies. Each episode, the method picks a sub-goal from the frontier, the edge of known state space, before the agent explores toward its objective. Sub-goal selection provides pairs neither overly familiar nor quite novel. That expands the frontier systematically, letting agents reach any state.

Catalog record

CREATORGeorgios Sotirchos, Zlatan Ajanović, Jens Kober
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE61221
LICENSEhttp://creativecommons.org/publicdomain/zero/1.0/
RIGHTSCC0
SOURCEarXiv

Source: https://arxiv.org/abs/2602.00460

Browse the full library · About the Tower of Babel