MuZero-style value-aware model losses are uncalibrated with stochastic models; the paper derives a variance-corrected CVAML loss and proves deterministic models can be value-equivalent.
A., Garg, A., and Farahmand, A.-m
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Calibrated Value-Aware Model Learning with Probabilistic Environment Models
MuZero-style value-aware model losses are uncalibrated with stochastic models; the paper derives a variance-corrected CVAML loss and proves deterministic models can be value-equivalent.