| Dataset Name | SOTA Method | Metric | Trend | ||
|---|---|---|---|---|---|
| IOI | FAP | Last-Token KL Divergence0.15 | 40 | 3mo ago | |
| IOI | EAP-IG | CPR3.491 | 16 | 1mo ago | |
| Mechanistic Interpretability Benchmark (MIB) Indirect Object Identification (IOI) (standard) | EAP-IG-inputs | CMD0 | 12 | 4mo ago | |
| Indirect Object Identification (IOI) (500 randomly-sampled examples) | Alignment0.9182 | 2 | 4mo ago | ||
| Wang-style IOI templates | Manifestation Units | Lift (routing)4.49 | 1 | 23d ago | |
| IOI evaluation episodes (held-out) | MechRL | Policy Score2.976 | 1 | 1mo ago |
