Transformer world models on Atari exhibit game-specific scaling regimes, but joint training on 26 environments produces consistent monotonic gains that improve downstream control policies to a median normalized score of 0.770.
Bigger, better, faster: Human-level
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Probing the Impact of Scale on Data-Efficient, Generalist Transformer World Models for Atari
Transformer world models on Atari exhibit game-specific scaling regimes, but joint training on 26 environments produces consistent monotonic gains that improve downstream control policies to a median normalized score of 0.770.