Fine-tuning a 3B LLM on randomly symbolized reasoning questions reduces spurious-correlation failures and improves OOD accuracy on CLadder and PrOntoQA.
Large language models as commonsense knowledge for large-scale task planning
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
-
Mitigating Spurious Correlations in LLMs via Causality-Aware Post-Training
Fine-tuning a 3B LLM on randomly symbolized reasoning questions reduces spurious-correlation failures and improves OOD accuracy on CLadder and PrOntoQA.