Artt comments on Do k-Sparse Autoencoders Reveal Thinking Patterns? Interpretable Features in a Small Reasoning Model

[Artt]

Thank you so much!
Yes, I agree. Reasoning models still have behaviors we don’t fully understand yet