Enhancing Automated Interpretability with Output-Centric Feature Descriptions
Paper
•
2501.08319
•
Published
•
10
Atticus Geiger
atticusg
AI & ML interests
NLP, explainable AI, causality
Recent Activity
upvoted
a
paper
9 days ago
Enhancing Automated Interpretability with Output-Centric Feature
Descriptions
updated
a dataset
about 1 month ago
atticusg/ravel
authored
a paper
4 months ago
Recurrent Neural Networks Learn to Store and Generate Sequences using
Non-Linear Representations
Organizations
atticusg's activity
upvoted
a
paper
9 days ago
updated
a
dataset
about 1 month ago
authored
a
paper
4 months ago
authored
12
papers
8 months ago
DynaSent: A Dynamic Benchmark for Sentiment Analysis
Paper
•
2012.15349
•
Published
ScoNe: Benchmarking Negation Reasoning in Language Models With Fine-Tuning and In-Context Learning
Paper
•
2305.19426
•
Published
CEBaB: Estimating the Causal Effects of Real-World Concepts on NLP Model Behavior
Paper
•
2205.14140
•
Published
Rigorously Assessing Natural Language Explanations of Neurons
Paper
•
2309.10312
•
Published
Linear Representations of Sentiment in Large Language Models
Paper
•
2310.15154
•
Published
Neural Natural Language Inference Models Partially Embed Theories of Lexical Entailment and Negation
Paper
•
2004.14623
•
Published
A Reply to Makelov et al. (2023)'s "Interpretability Illusion" Arguments
Paper
•
2401.12631
•
Published
ReFT: Representation Finetuning for Language Models
Paper
•
2404.03592
•
Published
•
92
pyvene: A Library for Understanding and Improving PyTorch Models via Interventions
Paper
•
2403.07809
•
Published
•
1
Causal Abstraction for Faithful Model Interpretation
Paper
•
2301.04709
•
Published
Finding Alignments Between Interpretable Causal Variables and Distributed Neural Representations
Paper
•
2303.02536
•
Published
•
1
RAVEL: Evaluating Interpretability Methods on Disentangling Language Model Representations
Paper
•
2402.17700
•
Published
•
2