The exact ground truth for this ROM and how every interpretability method scores
against it — one of the 42 scored games. Every number is read from
site_data.json (games.frostbite).
← back to the Ground Truth ROMs · Paper 2 audit · method catalogue
Each XAI / mechanistic-interpretability method's faithfulness on this ROM, sorted by F (click a header to re-sort). The left border marks the tradition: green = causal / intervention, red = gradient / correlational, amber = other — the split the paper's headline turns on. F is all-regime; content-F and position-F are the two output regimes (n/a where this game has no record for that regime).
| Method | Family | F (all) | content-F | position-F | S | M |
|---|---|---|---|---|---|---|
| A8 · Whole-state recording | descriptive | 1.000 | n/a | n/a | 1.000 | 0.109 |
| Activation patching / causal mediation | causal / intervention | 1.000 | n/a | 1.000 | 1.000 | 0.900 |
| Causal scrubbing | causal / intervention | 1.000 | n/a | n/a | 1.000 | 1.000 |
| Interchange interventions / DAS | causal / intervention | 1.000 | n/a | n/a | 1.000 | 0.593 |
| A2 · Single-unit lesions | causal / intervention | 0.803 | n/a | n/a | 0.721 | 1.000 |
| Occlusion | causal / intervention | 0.791 | 0.724 | 0.857 | 0.500 | 1.000 |
| KernelSHAP / Shapley | gradient / correlational | 0.783 | 0.763 | 0.804 | 0.400 | 0.395 |
| LIME | gradient / correlational | 0.702 | 0.580 | 0.824 | 0.500 | 0.395 |
| ACDC — automatic circuit discovery | causal / intervention | 0.667 | n/a | n/a | 0.421 | 1.000 |
| NMF/PCA dictionaries | dim_reduction | 0.659 | n/a | n/a | -0.864 | 1.000 |
| RISE | gradient / correlational | 0.619 | 0.534 | 0.704 | 0.250 | 0.395 |
| Logit / tuned lens | causal / intervention | 0.568 | n/a | n/a | n/a | 0.137 |
| On-distribution counterfactual | causal / intervention | 0.511 | 0.826 | 0.196 | 0.550 | 1.000 |
| Extremal / meaningful perturbation | causal / intervention | 0.493 | 0.560 | 0.425 | 0.125 | 1.000 |
| A3 · Tuning curves | gradient / correlational | 0.476 | n/a | n/a | 0.346 | 1.000 |
| A4 · Pairwise correlations | gradient / correlational | 0.362 | n/a | n/a | -0.039 | 0.821 |
| A6 · Granger causality | gradient / correlational | 0.345 | n/a | n/a | 0.901 | 0.812 |
| Sparse autoencoders | dim_reduction | 0.275 | n/a | n/a | 0.059 | 1.000 |
| Linear probing + control tasks | probing | 0.275 | n/a | n/a | n/a | 1.000 |
| Grad×Input / DeepLIFT | gradient / correlational | 0.272 | 0.484 | 0.059 | 0.150 | 1.000 |
| Integrated Gradients | gradient / correlational | 0.272 | 0.484 | 0.059 | 0.150 | 1.000 |
| Vanilla gradient (saliency) | gradient / correlational | 0.272 | 0.484 | 0.059 | 0.150 | 1.000 |
| SmoothGrad | gradient / correlational | 0.272 | 0.484 | 0.059 | 0.150 | 1.000 |
| Guided Backprop | gradient / correlational | 0.242 | 0.484 | 0.000 | 0.150 | 1.000 |
| A7 · Dimensionality reduction (NMF/PCA) | dim_reduction | 0.200 | n/a | n/a | n/a | n/a |
| A5 · Local field potentials | gradient / correlational | 0.138 | n/a | n/a | n/a | n/a |
| Attribution / edge patching | gradient / correlational | 0.031 | n/a | n/a | 0.943 | 1.000 |
| A1 · Connectomics / data-flow graph | causal / intervention | 0.000 | n/a | n/a | 0.977 | n/a |
| Expected Gradients | gradient / correlational | 0.000 | 0.000 | 0.000 | 0.125 | n/a |
| Path patching / IOI circuit | causal / intervention | 0.000 | n/a | n/a | n/a | n/a |
Source: site_data.json
· generated by gen_site_data.py from the
committed §R records + leaderboard.json.