DARC is a retraining-free inference-time method that frames response selection as distributionally robust optimization to reduce disagreement and tail risk under heterogeneous preferences.
Clear and easy to read. It explains what the ATF actually does. A bit robotic with the ’In addition...’ transitions, but it feels more trustworthy than the first one
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
-
DARC: Disagreement-Aware Alignment via Risk-Constrained Decoding
DARC is a retraining-free inference-time method that frames response selection as distributionally robust optimization to reduce disagreement and tail risk under heterogeneous preferences.