Towards Monosemanticity: Decomposing Language Models With Dictionary Learning
סדרה בארכיון ("עדכון לא פעיל" status)
When? This feed was archived on February 21, 2025 21:08 (
Why? עדכון לא פעיל status. השרתים שלנו לא הצליחו לאחזר פודקאסט חוקי לזמן ממושך.
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 424744796 series 3498845
Using a sparse autoencoder, we extract a large number of interpretable features from a one-layer transformer.
Mechanistic interpretability seeks to understand neural networks by breaking them into components that are more easily understood than the whole. By understanding the function of each component, and how they interact, we hope to be able to reason about the behavior of the entire network. The first step in that program is to identify the correct components to analyze.
Unfortunately, the most natural computational unit of the neural network – the neuron itself – turns out not to be a natural unit for human understanding. This is because many neurons are polysemantic: they respond to mixtures of seemingly unrelated inputs. In the vision model Inception v1, a single neuron responds to faces of cats and fronts of cars . In a small language model we discuss in this paper, a single neuron responds to a mixture of academic citations, English dialogue, HTTP requests, and Korean text. Polysemanticity makes it difficult to reason about the behavior of the network in terms of the activity of individual neurons.
Source:
https://transformer-circuits.pub/2023/monosemantic-features/index.html
Narrated for AI Safety Fundamentals by Perrin Walker
A podcast by BlueDot Impact.
Learn more on the AI Safety Fundamentals website.
פרקים
1. Towards Monosemanticity: Decomposing Language Models With Dictionary Learning (00:00:00)
2. Summary of Results (00:05:50)
85 פרקים