NLP

Truthfulness prediction of LLM’s responses using linguistic features.

Generating sensible universal adversarial triggers that are closer to natural phrases.