$K$-MSHC: Unmasking Minimally Sufficient Head Circuits in Large Language Models with Experiments on Syntactic Classification Tasks

Documenti analoghi

• AssertBench: A Benchmark for Evaluating Self-Assertion in Large Language Models
  di: Lee, Jaeho, et al.
  Pubblicazione: (2025)
• Necessary and Sufficient Watermark for Large Language Models
  di: Takezawa, Yuki, et al.
  Pubblicazione: (2023)
• Latent Performance Profiling of Large Language Models
  di: Chakraborty, Tanmoy, et al.
  Pubblicazione: (2026)
• Unmasking Hallucinations: A Causal Graph-Attention Perspective on Factual Reliability in Large Language Models
  di: kurra, Sailesh kiran, et al.
  Pubblicazione: (2026)
• SAP: Syntactic Attention Pruning for Transformer-based Language Models
  di: Lee, Tzu-Yun, et al.
  Pubblicazione: (2025)