BehR-based training of world models improves long-term behavioral alignment with real text environments in WebShop and TextWorld while preserving single-step prediction quality.
Title resolution pending
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
-
Beyond State Consistency: Behavior Consistency in Text-Based World Models
BehR-based training of world models improves long-term behavioral alignment with real text environments in WebShop and TextWorld while preserving single-step prediction quality.