Ground truth
The screenshot is the exact analysis frame the ground truth is computed on. Labels are imported (AtariARI / OCAtari) then verified causally on the bit-exact machine.
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.seaquest).
← back to the Ground Truth ROMs · Paper 2 audit · method catalogue
The screenshot is the exact analysis frame the ground truth is computed on. Labels are imported (AtariARI / OCAtari) then verified causally on the bit-exact machine.
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 |
|---|---|---|---|---|---|---|
| A2 · Single-unit lesions | causal / intervention | 1.000 | n/a | n/a | 0.800 | 1.000 |
| A8 · Whole-state recording | descriptive | 1.000 | n/a | n/a | 1.000 | 0.117 |
| Activation patching / causal mediation | causal / intervention | 1.000 | n/a | 1.000 | 1.000 | 1.000 |
| Causal scrubbing | causal / intervention | 1.000 | n/a | n/a | 1.000 | 0.786 |
| Interchange interventions / DAS | causal / intervention | 1.000 | n/a | n/a | 1.000 | 0.970 |
| ACDC — automatic circuit discovery | causal / intervention | 0.880 | n/a | n/a | 0.217 | 1.000 |
| Logit / tuned lens | causal / intervention | 0.812 | n/a | n/a | n/a | 0.310 |
| Grad×Input / DeepLIFT | gradient / correlational | 0.762 | 0.813 | 0.710 | 0.958 | 1.000 |
| A7 · Dimensionality reduction (NMF/PCA) | dim_reduction | 0.600 | n/a | n/a | n/a | n/a |
| On-distribution counterfactual | causal / intervention | 0.387 | 0.775 | 0.000 | 0.958 | 1.000 |
| Vanilla gradient (saliency) | gradient / correlational | 0.364 | 0.017 | 0.710 | 0.479 | 1.000 |
| SmoothGrad | gradient / correlational | 0.364 | 0.017 | 0.710 | 0.479 | 1.000 |
| Occlusion | causal / intervention | 0.357 | 0.000 | 0.714 | 0.479 | 1.000 |
| LIME | gradient / correlational | 0.357 | 0.000 | 0.714 | 0.479 | 0.413 |
| KernelSHAP / Shapley | gradient / correlational | 0.357 | 0.000 | 0.714 | 0.479 | 0.432 |
| Integrated Gradients | gradient / correlational | 0.355 | 0.000 | 0.710 | 0.479 | 1.000 |
| RISE | gradient / correlational | 0.346 | 0.000 | 0.693 | 0.479 | 0.413 |
| A5 · Local field potentials | gradient / correlational | 0.305 | n/a | n/a | n/a | n/a |
| A4 · Pairwise correlations | gradient / correlational | 0.304 | n/a | n/a | 0.033 | 0.429 |
| A1 · Connectomics / data-flow graph | causal / intervention | 0.250 | n/a | n/a | 0.984 | 1.000 |
| Path patching / IOI circuit | causal / intervention | 0.200 | n/a | n/a | n/a | 1.000 |
| Attribution / edge patching | gradient / correlational | 0.178 | n/a | n/a | 0.952 | 1.000 |
| Extremal / meaningful perturbation | causal / intervention | 0.145 | 0.000 | 0.289 | 0.479 | 1.000 |
| A3 · Tuning curves | gradient / correlational | 0.078 | n/a | n/a | 0.311 | 1.000 |
| A6 · Granger causality | gradient / correlational | 0.061 | n/a | n/a | 0.947 | 0.737 |
| Guided Backprop | gradient / correlational | 0.009 | 0.017 | 0.000 | 0.479 | 1.000 |
| NMF/PCA dictionaries | dim_reduction | 0.000 | n/a | n/a | 0.652 | 1.000 |
| Expected Gradients | gradient / correlational | 0.000 | 0.000 | 0.000 | 0.479 | n/a |
| Sparse autoencoders | dim_reduction | 0.000 | n/a | n/a | 0.706 | 1.000 |
Source: site_data.json
· generated by gen_site_data.py from the
committed §R records + leaderboard.json.