PACS samples LLM-generated proofs as observations of distinct individual commonsense beliefs and aggregates them to predict majority judgments, outperforming chain-of-thought and prior neurosymbolic baselines on benchmarks.
LRM s are not thinking straight: Unreliability of thinking trajectories
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
-
Abductive Reasoning with Probabilistic Commonsense
PACS samples LLM-generated proofs as observations of distinct individual commonsense beliefs and aggregates them to predict majority judgments, outperforming chain-of-thought and prior neurosymbolic baselines on benchmarks.