Trained SetConCA sparse dictionary checkpoint for the paper.
1model.safetensors or setconca_latest.pt
2config.json
3metrics.json
4eval_table_fullbank.csv
5eval_table_geometry.csv
6eval_table_lovo.csv
Research on sparse representation learning, grouped activation-set geometry, and mechanistic interpretability.
The learned sparse coordinates are not guaranteed to be monosemantic and are not validated as causal intervention targets.
