History
The original AxBench evaluation substantially underestimated the steering performance of sparse autoencoder features.
Claim entered the graph
Nothing has happened to this claim yet beyond its creation: no assessments, contributions, or decisions are on record.