Designing Role Vectors to Improve LLM Inference Behaviour

Documenti analoghi

• Disce aut Deficere: Evaluating LLMs Proficiency on the INVALSI Italian Benchmark
  di: Mercorio, Fabio, et al.
  Pubblicazione: (2024)
• XAI meets LLMs: A Survey of the Relation between Explainable AI and Large Language Models
  di: Cambria, Erik, et al.
  Pubblicazione: (2024)
• SAFE: A Sparse Autoencoder-Based Framework for Robust Query Enrichment and Hallucination Mitigation in LLMs
  di: Abdaljalil, Samir, et al.
  Pubblicazione: (2025)
• Propensity Inference: Environmental Contributors to LLM Behaviour
  di: Järviniemi, Olli, et al.
  Pubblicazione: (2026)
• Vectors Are Not Neutral: Sensitive-Information Inference from Exported LLM Representations in Summarization
  di: Liu, Weixin, et al.
  Pubblicazione: (2026)