Unifying Model-Free Efficiency and Model-Based Representations via Latent Dynamics Conference: Proc. of the 25th International Conference on Autonomous Agents and Multiagent Systems (AAMAS 2026); May 25 – 29, 2026; Paphos, Cyprus

Jashaswimalya Acharjee, Balaraman Ravindran · 2026

Read Unifying Model-Free Efficiency and Model-Based Representations via Latent Dynamics Conference: Proc. of the 25th International Conference on Autonomous Agents and Multiagent Systems (AAMAS 2026); May 25 – 29, 2026; Paphos, Cyprus by Jashaswimalya Acharjee, Balaraman Ravindran (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-2602-12643.txt.

Description

Reinforcement learning methods split between model-free speed and model-based representational strength, each with its own costs. Jashaswimalya Acharjee led this work with one co-author, Balaraman Ravindran, giving Unified Latent Dynamics, a new method that unifies model-free speed with model-based strengths, without planning cost.

State-action pairs are placed into a latent space in which the true value function is nearly linear. That supports a single set of tuning knobs across varied domains, from continuous control with low-dimensional and pixel inputs to high-dimensional Atari games. Under mild rules, the fixed point of embedding-based temporal-difference updates matches that of a matching linear model-based value expansion. Tests show strong scores across domains with a single setup.

Catalog record

CREATORJashaswimalya Acharjee, Balaraman Ravindran
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE54 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2602-12643

Browse the full library · About the Tower of Babel