))
Interpretability in the Wild: A Circuit for Indirect Object Identification in GPT-2 Small
Gearchiveerde serie ("Inactieve feed" status)
When?
This feed was archived on February 21, 2025 21:08 (
Why? Inactieve feed status. Onze servers konden geen geldige podcast feed ononderbroken ophalen.
What now? You might be able to find a more up-to-date version using the search function. This series will no longer be checked for updates. If you believe this to be in error, please check if the publisher's feed link below is valid and contact support to request the feed be restored or if you have any other concerns about this.
Manage episode 424744795 series 3498845
Inhoud geleverd door BlueDot Impact. Alle podcastinhoud, inclusief afleveringen, afbeeldingen en podcastbeschrijvingen, wordt rechtstreeks geüpload en geleverd door BlueDot Impact of hun podcastplatformpartner. Als u denkt dat iemand uw auteursrechtelijk beschermde werk zonder uw toestemming gebruikt, kunt u het hier beschreven proces https://nl.player.fm/legal volgen.
Research in mechanistic interpretability seeks to explain behaviors of machine learning (ML) models in terms of their internal components. However, most previous work either focuses on simple behaviors in small models or describes complicated behaviors in larger models with broad strokes. In this work, we bridge this gap by presenting an explanation for how GPT-2 small performs a natural language task called indirect object identification (IOI). Our explanation encompasses 26 attention heads grouped into 7 main classes, which we discovered using a combination of interpretability approaches relying on causal interventions. To our knowledge, this investigation is the largest end-to-end attempt at reverse-engineering a natural behavior "in the wild" in a language model. We evaluate the reliability of our explanation using three quantitative criteria–faithfulness, completeness, and minimality. Though these criteria support our explanation, they also point to remaining gaps in our understanding. Our work provides evidence that a mechanistic understanding of large ML models is feasible, pointing toward opportunities to scale our understanding to both larger models and more complex tasks. Code for all experiments is available at https://github.com/redwoodresearch/Easy-Transformer.
Source:
https://arxiv.org/pdf/2211.00593.pdf
Narrated for AI Safety Fundamentals by Perrin Walker
A podcast by BlueDot Impact.
Learn more on the AI Safety Fundamentals website.
Hoofdstukken
1. Interpretability in the Wild: A Circuit for Indirect Object Identification in GPT-2 Small (00:00:00)
2. ABSTRACT (00:00:17)
3. 1 INTRODUCTION (00:01:34)
4. 2 BACKGROUND (00:05:58)
5. 2.1 CIRCUITS AND KNOCKOUTS (00:10:41)
6. 3 DISCOVERING THE CIRCUIT (00:14:06)
7. 3.1 WHICH HEADS DIRECTLY AFFECT THE OUTPUT? (NAME MOVER HEADS) (00:19:08)
85 afleveringen
Delen
