LessWrong (30+ Karma)

“Self-explaining SAE features” by Dmitrii Kharlapenko, neverix


Listen Later

TL;DR

  • We apply the method of SelfIE/Patchscopes to explain SAE features – we give the model a prompt like “What does X mean?”, replace the residual stream on X with the decoder direction times some scale, and have it generate an explanation. We call this self-explanation.
  • The natural alternative is auto-interp, using a larger LLM to spot patterns in max activating examples. We show that our method is effective, and comparable with Neuronpedia's auto-interp labels (with the caveat that Neuronpedia's auto-interp used the comparatively weak GPT-3.5 so this is not a fully fair comparison).
  • We aren’t confident you should use our method over auto-interp, but we think in some situations it has advantages: no max activating dataset examples are needed, and it's cheaper as you just run the model being studied (eg Gemma 2B) not a larger model like GPT-4.
    • Further, it has different errors to auto-interp, so [...]

---

Outline:

(00:05) TL;DR

(01:39) Introduction

(03:49) How to use

(03:53) Basic method

(04:48) Metrics for choosing scale and evaluating

(05:58) Self-Similarity

(07:44) Entropy

(09:33) Composite

(10:36) Other experiments

(11:18) Evaluation

(15:55) Limitations and improvements

(15:59) Recovering the activating token

(18:42) Failure detection

(19:04) Maximum Self-Similarity Thresholding

(19:44) Repeat Prompt Failure Detection

(20:08) Layer-Specific Thresholds

(20:57) Prior work

(21:55) Examples

(21:58) Gemma 2B

(22:02) Random simple features

(23:02) More complex features

(25:05) Phi-3 Mini

(25:25) Random features

(26:18) Refusal features

(26:48) Acknowledgements

(27:03) Appendix

(27:06) Entropy justification

The original text contained 1 image which was described by AI.

---

First published:

August 5th, 2024

Source:

https://www.lesswrong.com/posts/8ev6coxChSWcxCDy8/self-explaining-sae-features

---

Narrated by TYPE III AUDIO.

---

Images from the article:

Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.

...more
View all episodesView all episodes
Download on the App Store

LessWrong (30+ Karma)By LessWrong


More shows like LessWrong (30+ Karma)

View all
Making Sense with Sam Harris by Sam Harris

Making Sense with Sam Harris

26,420 Listeners

Conversations with Tyler by Mercatus Center at George Mason University

Conversations with Tyler

2,387 Listeners

The Peter Attia Drive by Peter Attia, MD

The Peter Attia Drive

7,893 Listeners

Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas by Sean Carroll | Wondery

Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas

4,126 Listeners

ManifoldOne by Steve Hsu

ManifoldOne

87 Listeners

Your Undivided Attention by Tristan Harris and Aza Raskin, The Center for Humane Technology

Your Undivided Attention

1,458 Listeners

All-In with Chamath, Jason, Sacks & Friedberg by All-In Podcast, LLC

All-In with Chamath, Jason, Sacks & Friedberg

9,040 Listeners

Machine Learning Street Talk (MLST) by Machine Learning Street Talk (MLST)

Machine Learning Street Talk (MLST)

87 Listeners

Dwarkesh Podcast by Dwarkesh Patel

Dwarkesh Podcast

390 Listeners

Hard Fork by The New York Times

Hard Fork

5,431 Listeners

The Ezra Klein Show by New York Times Opinion

The Ezra Klein Show

15,216 Listeners

Moonshots with Peter Diamandis by PHD Ventures

Moonshots with Peter Diamandis

474 Listeners

No Priors: Artificial Intelligence | Technology | Startups by Conviction

No Priors: Artificial Intelligence | Technology | Startups

121 Listeners

Latent Space: The AI Engineer Podcast by swyx + Alessio

Latent Space: The AI Engineer Podcast

75 Listeners

BG2Pod with Brad Gerstner and Bill Gurley by BG2Pod

BG2Pod with Brad Gerstner and Bill Gurley

459 Listeners