Beyond Data Filtering: Knowledge Localization for Capability Removal in LLMs

Documenti analoghi

• The Hot Mess of AI: How Does Misalignment Scale With Model Intelligence and Task Complexity?
  di: Hägele, Alexander, et al.
  Pubblicazione: (2026)
• Gradient Routing: Masking Gradients to Localize Computation in Neural Networks
  di: Cloud, Alex, et al.
  Pubblicazione: (2024)
• Lost in Time: Clock and Calendar Understanding Challenges in Multimodal LLMs
  di: Saxena, Rohit, et al.
  Pubblicazione: (2025)
• Inverse Scaling in Test-Time Compute
  di: Gema, Aryo Pradipta, et al.
  Pubblicazione: (2025)
• Scalpel vs. Hammer: GRPO Amplifies Existing Capabilities, SFT Replaces Them
  di: Rajani, Neel, et al.
  Pubblicazione: (2025)