An author proposes a research design for a reasoning system that detects contradictions between two propositions by decomposing them into smaller components and searching for logical collisions, rather than relying on direct LLM judgment.
- The system separates probabilistic exploration from formal verification to improve efficiency under limited generation budgets.
- Proposition identification uses implication relations to map corresponding content across different logical structures.
- A collision-seeking search strategy aims to find positive and negative forms of the same subject more efficiently than undirected search.
- The initial implementation targets propositional logic, outputting CONTRADICTION, CONSISTENT, or UNKNOWN.
The author is seeking discussion, criticism, alternative approaches, and collaborators to help implement or test this hypothesis.