ConflictQA benchmark shows LLMs fail to resolve conflicts between text and KG evidence and often default to one source, motivating the XoT explanation-based reasoning method.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Exploring Knowledge Conflicts for Faithful LLM Reasoning: Benchmark and Method
ConflictQA benchmark shows LLMs fail to resolve conflicts between text and KG evidence and often default to one source, motivating the XoT explanation-based reasoning method.