PRISM: Probing Reasoning, Instruction, and Source Memory in LLM Hallucinations

Documenti analoghi

• Sandwich Reasoning: An Answer-Reasoning-Answer Approach for Low-Latency Query Correction
  di: Zhang, Chen, et al.
  Pubblicazione: (2026)
• Is LLMs Hallucination Usable? LLM-based Negative Reasoning for Fake News Detection
  di: Zhang, Chaowei, et al.
  Pubblicazione: (2025)
• Semantic Entropy Probes: Robust and Cheap Hallucination Detection in LLMs
  di: Kossen, Jannik, et al.
  Pubblicazione: (2024)
• ICR Probe: Tracking Hidden State Dynamics for Reliable Hallucination Detection in LLMs
  di: Zhang, Zhenliang, et al.
  Pubblicazione: (2025)
• Probing LLM Hallucination from Within: Perturbation-Driven Approach via Internal Knowledge
  di: Lee, Seongmin, et al.
  Pubblicazione: (2024)