Work Done

Confusions

• The loss is pretty stagnant, will this improve performance for the MNIST?

Next Steps

• Train the model on EMNIST rather than MNIST to get more data to actually run the SAEs

• Message Bart Bussman for repo for meta-sae's https://www.alignmentforum.org/inbox/quDmw96SJdzJk8yS3 --> also ask about if this research direction is even logical --> package it up.

• Need to look at the impact of sae reconstructions on MNIST loss
  • Essentially tackling cross-entropy loss score
• Need to Evaluate my SAEs
  • Full Post that contains all of the metrics