Christopher Potts

8

Papers

114

Total Citations

Papers (8)

AxBench: Steering LLMs? Even Simple Baselines Outperform Sparse Autoencoders

MrT5: Dynamic Token Merging for Efficient Byte-level Language Models

Dynaboard: An Evaluation-As-A-Service Platform for Holistic Next-Generation Benchmarking

Decrypting Cryptic Crosswords: Semantically Complex Wordplay Puzzles as a Target for NLP

Baleen: Robust Multi-Hop Reasoning at Scale via Condensed Retrieval

CEBaB: Estimating the Causal Effects of Real-World Concepts on NLP Model Behavior

Interpretability at Scale: Identifying Causal Mechanisms in Alpaca

Causal Abstractions of Neural Networks