REAR decomposes the reward into question and preference components, rescales their balance, and expresses the result as a linear combination of token log-probabilities for efficient integration with best-of-N and tree search.
Drift : Decoding-time personalized alignments with implicit user preferences
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
REAR: Test-time Preference Realignment through Reward Decomposition
REAR decomposes the reward into question and preference components, rescales their balance, and expresses the result as a linear combination of token log-probabilities for efficient integration with best-of-N and tree search.