Lowest Span Confidence: A Zero-Shot Metric for Efficient and Black-Box Hallucination Detection in LLMs

Documenti analoghi

• Inducing Overthink: Hierarchical Genetic Algorithm-based DoS Attack on Black-Box Large Language Reasoning Models
  di: Wang, Shuqiang, et al.
  Pubblicazione: (2026)
• ZeroShotOpt: Towards Zero-Shot Pretrained Models for Efficient Black-Box Optimization
  di: Meindl, Jamison, et al.
  Pubblicazione: (2025)
• FactSelfCheck: Fact-Level Black-Box Hallucination Detection for LLMs
  di: Sawczyn, Albert, et al.
  Pubblicazione: (2025)
• Robust Reward Modeling for Large Language Models via Causal Decomposition
  di: Lu, Yunsheng, et al.
  Pubblicazione: (2026)
• Understanding and Mitigating Overrefusal in LLMs from an Unveiling Perspective of Safety Decision Boundary
  di: Pan, Licheng, et al.
  Pubblicazione: (2025)