claim
Mode-seeking decoding methods appear to reduce hallucinations in language models, particularly in knowledge-grounded settings.

Authors

Sources

Referenced by nodes (2)