We're sunsetting PodQuest on 2025-07-28. Thank you for your support!
Export Podcast Subscriptions
back
“What We Learned Trying to Diff Base and Chat Models (And Why It Matters)” by Clément Dumas, Julian Minder, Neel Nanda
19:37
Share
2025/6/30
LessWrong (30+ Karma)
AI Chapters
Transcribe
Chapters
What’s the TL;DR of Model Diffing?
Why is Model Diffing Important?
The Problem: Are Model-Specific Latents Really Unique?
The Fix: How Does BatchTopK Crosscoders Work?
Do We Really Need Crosscoders?
How Effective Are diff-SAEs in Capturing Behavioral Differences?
What’s the Conclusion of This Research?
What Other Techniques Are in the Model Diffing Toolkit?
What’s the Closed Form Solution for Latent Scaling?
Shownotes
Transcript
No transcript made for this episode yet, you may request it for free.