Information-Theoretic Text Hallucination Reduction for Video-grounded Dialogue

31037

Knowledge Triples