Alert button
Picture for Ellie Pavlick

Ellie Pavlick

Alert button

Deep Neural Networks Can Learn Generalizable Same-Different Visual Relations

Oct 14, 2023
Alexa R. Tartaglini, Sheridan Feucht, Michael A. Lepori, Wai Keen Vong, Charles Lovering, Brenden M. Lake, Ellie Pavlick

Figure 1 for Deep Neural Networks Can Learn Generalizable Same-Different Visual Relations
Figure 2 for Deep Neural Networks Can Learn Generalizable Same-Different Visual Relations
Figure 3 for Deep Neural Networks Can Learn Generalizable Same-Different Visual Relations
Figure 4 for Deep Neural Networks Can Learn Generalizable Same-Different Visual Relations
Viaarxiv icon

Circuit Component Reuse Across Tasks in Transformer Language Models

Oct 12, 2023
Jack Merullo, Carsten Eickhoff, Ellie Pavlick

Figure 1 for Circuit Component Reuse Across Tasks in Transformer Language Models
Figure 2 for Circuit Component Reuse Across Tasks in Transformer Language Models
Figure 3 for Circuit Component Reuse Across Tasks in Transformer Language Models
Figure 4 for Circuit Component Reuse Across Tasks in Transformer Language Models
Viaarxiv icon

NeuroSurgeon: A Toolkit for Subnetwork Analysis

Sep 01, 2023
Michael A. Lepori, Ellie Pavlick, Thomas Serre

Viaarxiv icon

Language Models Implement Simple Word2Vec-style Vector Arithmetic

May 25, 2023
Jack Merullo, Carsten Eickhoff, Ellie Pavlick

Figure 1 for Language Models Implement Simple Word2Vec-style Vector Arithmetic
Figure 2 for Language Models Implement Simple Word2Vec-style Vector Arithmetic
Figure 3 for Language Models Implement Simple Word2Vec-style Vector Arithmetic
Figure 4 for Language Models Implement Simple Word2Vec-style Vector Arithmetic
Viaarxiv icon

Testing Causal Models of Word Meaning in GPT-3 and -4

May 24, 2023
Sam Musker, Ellie Pavlick

Figure 1 for Testing Causal Models of Word Meaning in GPT-3 and -4
Figure 2 for Testing Causal Models of Word Meaning in GPT-3 and -4
Figure 3 for Testing Causal Models of Word Meaning in GPT-3 and -4
Figure 4 for Testing Causal Models of Word Meaning in GPT-3 and -4
Viaarxiv icon

Simfluence: Modeling the Influence of Individual Training Examples by Simulating Training Runs

Mar 14, 2023
Kelvin Guu, Albert Webson, Ellie Pavlick, Lucas Dixon, Ian Tenney, Tolga Bolukbasi

Figure 1 for Simfluence: Modeling the Influence of Individual Training Examples by Simulating Training Runs
Figure 2 for Simfluence: Modeling the Influence of Individual Training Examples by Simulating Training Runs
Figure 3 for Simfluence: Modeling the Influence of Individual Training Examples by Simulating Training Runs
Figure 4 for Simfluence: Modeling the Influence of Individual Training Examples by Simulating Training Runs
Viaarxiv icon

Break It Down: Evidence for Structural Compositionality in Neural Networks

Jan 26, 2023
Michael A. Lepori, Thomas Serre, Ellie Pavlick

Figure 1 for Break It Down: Evidence for Structural Compositionality in Neural Networks
Figure 2 for Break It Down: Evidence for Structural Compositionality in Neural Networks
Figure 3 for Break It Down: Evidence for Structural Compositionality in Neural Networks
Figure 4 for Break It Down: Evidence for Structural Compositionality in Neural Networks
Viaarxiv icon

Are Language Models Worse than Humans at Following Prompts? It's Complicated

Jan 17, 2023
Albert Webson, Alyssa Marie Loo, Qinan Yu, Ellie Pavlick

Figure 1 for Are Language Models Worse than Humans at Following Prompts? It's Complicated
Figure 2 for Are Language Models Worse than Humans at Following Prompts? It's Complicated
Figure 3 for Are Language Models Worse than Humans at Following Prompts? It's Complicated
Figure 4 for Are Language Models Worse than Humans at Following Prompts? It's Complicated
Viaarxiv icon

Does CLIP Bind Concepts? Probing Compositionality in Large Image Models

Dec 20, 2022
Martha Lewis, Qinan Yu, Jack Merullo, Ellie Pavlick

Figure 1 for Does CLIP Bind Concepts? Probing Compositionality in Large Image Models
Figure 2 for Does CLIP Bind Concepts? Probing Compositionality in Large Image Models
Figure 3 for Does CLIP Bind Concepts? Probing Compositionality in Large Image Models
Figure 4 for Does CLIP Bind Concepts? Probing Compositionality in Large Image Models
Viaarxiv icon

Evaluation Beyond Task Performance: Analyzing Concepts in AlphaZero in Hex

Nov 26, 2022
Charles Lovering, Jessica Zosa Forde, George Konidaris, Ellie Pavlick, Michael L. Littman

Figure 1 for Evaluation Beyond Task Performance: Analyzing Concepts in AlphaZero in Hex
Figure 2 for Evaluation Beyond Task Performance: Analyzing Concepts in AlphaZero in Hex
Figure 3 for Evaluation Beyond Task Performance: Analyzing Concepts in AlphaZero in Hex
Figure 4 for Evaluation Beyond Task Performance: Analyzing Concepts in AlphaZero in Hex
Viaarxiv icon