Across synthetic spurious-correlation tasks, DPO and KTO are more robust in math reasoning while SFT is stronger in document QA, and higher spuriousness does not uniformly degrade performance.
Length-controlled alpacaeval: A simple way to debias automatic evaluators
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Assessing Robustness to Spurious Correlations in Post-Training Language Models
Across synthetic spurious-correlation tasks, DPO and KTO are more robust in math reasoning while SFT is stronger in document QA, and higher spuriousness does not uniformly degrade performance.