LCV explicitly reconstructs missing cross-source facts via LLM for each sentence-context pair and uses them as textual relations in heterograph reasoning, improving macro-F1 by 2.56 and 2.84 points over the strongest baseline on English and Chinese splits of a bilingual misinformation benchmark.
Evaluating evidence attribution in generated fact checking explanations
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
-
Latent Causal Void: Explicit Missing-Context Reconstruction for Misinformation Detection
LCV explicitly reconstructs missing cross-source facts via LLM for each sentence-context pair and uses them as textual relations in heterograph reasoning, improving macro-F1 by 2.56 and 2.84 points over the strongest baseline on English and Chinese splits of a bilingual misinformation benchmark.