Pong — Paper 2 per-game results

Every method's faithfulness on Pong

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	0.836	1.000
A8 · Whole-state recording	descriptive	1.000	n/a	n/a	1.000	0.031
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	0.667
Interchange interventions / DAS	causal / intervention	1.000	n/a	n/a	1.000	1.000
Attribution / edge patching	gradient / correlational	0.876	n/a	n/a	0.973	1.000
On-distribution counterfactual	causal / intervention	0.641	0.696	0.586	0.389	0.875
Occlusion	causal / intervention	0.641	0.696	0.586	0.389	0.875
A4 · Pairwise correlations	gradient / correlational	0.633	n/a	n/a	-0.170	0.667
KernelSHAP / Shapley	gradient / correlational	0.625	0.690	0.559	0.389	0.281
LIME	gradient / correlational	0.620	0.686	0.554	0.389	0.281
A7 · Dimensionality reduction (NMF/PCA)	dim_reduction	0.600	n/a	n/a	n/a	n/a
Grad×Input / DeepLIFT	gradient / correlational	0.550	0.822	0.278	0.500	1.000
RISE	gradient / correlational	0.534	0.610	0.458	0.389	0.281
Logit / tuned lens	causal / intervention	0.438	n/a	n/a	n/a	0.054
Extremal / meaningful perturbation	causal / intervention	0.402	0.469	0.335	0.389	0.875
Expected Gradients	gradient / correlational	0.371	0.464	0.278	0.444	1.000
Integrated Gradients	gradient / correlational	0.371	0.464	0.278	0.444	1.000
Vanilla gradient (saliency)	gradient / correlational	0.371	0.464	0.278	0.444	1.000
SmoothGrad	gradient / correlational	0.371	0.464	0.278	0.444	1.000
NMF/PCA dictionaries	dim_reduction	0.300	n/a	n/a	0.932	1.000
A5 · Local field potentials	gradient / correlational	0.248	n/a	n/a	n/a	n/a
Guided Backprop	gradient / correlational	0.232	0.464	0.000	0.389	1.000
ACDC — automatic circuit discovery	causal / intervention	0.154	n/a	n/a	0.625	1.000
Sparse autoencoders	dim_reduction	0.041	n/a	n/a	-0.458	1.000
A1 · Connectomics / data-flow graph	causal / intervention	0.000	n/a	n/a	0.839	n/a
A3 · Tuning curves	gradient / correlational	0.000	n/a	n/a	-0.412	1.000
A6 · Granger causality	gradient / correlational	0.000	n/a	n/a	0.821	1.000
Linear probing + control tasks	probing	0.000	n/a	n/a	n/a	0.600
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.

Ground truth

Every method's faithfulness on Pong