Preprint
Aug 2026
Prompt Embedding Probes (PEP): Hallucination Detection in LLMs from Hidden States
Results show that prompt-based adaptation can strengthen hidden-state probing while keeping the backbone frozen and adding only a small number of trainable parameters, as well as improving hidden-state-based detection over standard linear probes in the main in-distribution setting.
Zakhar Mrykhin, Valentin Malykh
· 2 citations