AI Tools & Models HOTFeaturedAll
arXiv cs.AIScore65

AttentionSwitch: Jailbreak Defense for LLMs via Joint Learning of Sparse Attention-Head Selection and Scaling

Summary

This work defends LLMs against jailbreak attacks by jointly learning which attention heads to select and how to scale them.

Why this matters

Covers model and LLM · Relevant to this topic

Open original ↗

Source: arXiv cs.AI · arxiv.org

Topics

Related items