AI Tools & Models HOTFeaturedAll
arXiv cs.AIScore70

DECODE: Diagnosing Interpretability Failures in Automated Interpretation Process

Summary

This work identifies where automated interpretation of LLM internals breaks down and offers a way to diagnose those failures.

Why this matters

Covers model and LLM · Relevant to this topic

Open original ↗

Source: arXiv cs.AI · arxiv.org

Topics

Related items