dataset
active
dataset:lmsys-chat-1mLMSYS-CHAT-1M
Chat dataset (n=18,777 sampled) used to measure how much persona space PCs explain overall activation variance and to calibrate steering norms
Neighborhood — ranked by edge-count
Papers (7)
paper
Frameworks (2)
framework
- Turn-Averaged SAEcitesPaper's main proposal: SAE trained on the mean activation across a turn.
- Architectural variant partitioning the dictionary into A-exclusive, B-exclusive and shared sets.
Findings (1)
finding
- LMSYS-Chat-1M SAE produced the most effective features for chat applications on Llama-3-8B-InstructaboutResult of extensive hyperparameter and dataset experimentation.
Artifacts (1)
artifact
- ReLU SAE trained on residual stream layer 19 with MSE+L1 loss.