Introduces a pointwise total-variation robust Plackett-Luce loss for listwise preference optimization that exactly decomposes into nominal loss plus a worst-case correction obtained by reverse-sorting implicit scores.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AI 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Distributionally Robust Listwise Preference Optimization
Introduces a pointwise total-variation robust Plackett-Luce loss for listwise preference optimization that exactly decomposes into nominal loss plus a worst-case correction obtained by reverse-sorting implicit scores.