CE-Bench: Towards a Reliable Contrastive Evaluation Benchmark of Interpretability of Sparse Autoencoders

Published in BlackboxNLP at EMNLP 2025. *=equal contribution, 2025

Alex Gulko*, Yusen Peng*, Sachin Kumar

Citation

Alex Gulko*, Yusen Peng*, Sachin Kumar. (2025). CE-Bench: Towards a Reliable Contrastive Evaluation Benchmark of Interpretability of Sparse Autoencoders. BlackboxNLP at EMNLP 2025. *=equal contribution.