Sample stimuli

How to use

from brainscore_vision import load_benchmark
benchmark = load_benchmark("Geirhos2021cueconflict-error_consistency")
score = benchmark(my_model)

Benchmark API

Code examples

Model scores

Score Legend Min Alignment Max Alignment

Rank	Model	Score
1	convnext_xxlarge:clip_laion2b_soup_ft_in1k	.943
2	vit_huge_patch14_clip_224:laion2b_ft_in12k_in1k	.938
3	vit_large_patch14_clip_224:laion2b_ft_in1k	.855
4	vit_large_patch14_clip_224:openai_ft_in1k	.840
5	vit_large_patch14_clip_224:openai_ft_in12k_in1k	.804
6	vit_large_patch14_clip_224:laion2b_ft_in12k_in1k	.784
7	vit_huge_patch14_clip_336:laion2b_ft_in12k_in1k	.780
8	vit_large_patch14_clip_336:openai_ft_in12k_in1k	.752
9	vit_large_patch14_clip_336:laion2b_ft_in1k	.748
10	convnext_large_mlp:clip_laion2b_augreg_ft_in1k_384	.740
11	convnext_large:fb_in22k_ft_in1k	.698
12	convnext_base:clip_laiona_augreg_ft_in1k_384	.686
13	vision_transformer_vit_large_patch16_224	.656
14	convnext_xlarge:fb_in22k_ft_in1k	.651
15	cvt_cvt-w24-384-in22k_finetuned-in1k_4	.633
16	fixres_resnext101_32x48d_wsl	.611
17	resnext101_32x32d_wsl	.596
18	resnext101_32x16d_wsl	.595
19	vit_base_patch16_clip_224:openai_ft_in12k_in1k	.571
20	resnext101_32x48d_wsl	.560
21	vit_base_patch16_clip_224:openai_ft_in1k	.560
22	SWSL_resnext101_32x8d	.546
23	resnext101_32x8d_wsl	.540
24	CLIP_ViT-B_32	.482
25	cvt_cvt-21-384-in22k_finetuned-in1k_4_LucyV4	.477
26	resnet50-VITO-8deg-cc	.451
27	effnetv2m_custom384	.448
28	cvt_cvt-13-384-in1k_4_LucyV4	.441
29	resnet-152_v2-tf	.405
30	resnet_152_v2	.402
31	cvt_cvt-13-384-in22k_finetuned-in1k_4_LucyV4	.394
32	resnet_101_v2	.393
33	SWSL_resnet50	.390
34	resnet_50_v2	.383
35	vit_relpos_base_patch16_clsgap_224:sw_in1k	.380
36	ViT_B_16_imagenet1k	.376
37	cvt_cvt-21-384-in1k_4_LucyV4	.372
38	CLIP_resnet50_float32	.371
39	CLIP_resnet50	.371
40	effnetb1_272x240	.357
41	pnasnet_large	.355
42	efficientnet_b0	.346
43	inception_v3	.345
44	ViT_L_32_imagenet1k	.344
45	xception-keras	.343
46	nasnet_large	.332
47	cv_18_dagger_408_pretrained	.328
48	pnasnet_large-tf	.327
49	convnext_femto_ols:d1_in1k	.325
50	convnext_tiny_imagenet_full_seed-0	.321
51	CLIP-RN50	.317
52	pnasnet_large_pytorch	.316
53	resnet101	.316
54	resnet50_finetune_cutmix_e3_robust_linf8255_e0_247x234	.316
55	focalnet_tiny_lrf_in1k	.313
56	inception_v3-tf	.311
57	ViT_B_32_imagenet1k	.309
58	resnet152_imagenet_full	.309
59	cvt_cvt-21-224-in1k_4_LucyV4	.309
60	inception_v3_pytorch	.300
61	effnetb1_cutmixrespatch_SAM_robust32_e2_manylayers_324x288	.299
62	convnext_tiny:in12k_ft_in1k	.297
63	effnetb1_cutmixpatch_SAM_robust32_avge6e8e9e10_manylayers_324x288	.294
64	xception	.292
65	resnet-50_v2-tf	.292
66	effnetb1_cutmixrespatch_SAM_robust32_e10_manylayers_324x288	.284
67	convnext_large_imagenet_full_seed-0	.284
68	vit_relpos_base_patch32_plus_rpn_256:sw_in1k	.282
69	convnext_base_imagenet_full_seed-0	.278
70	resnet50-SIN_IN	.272
71	resnet_101_v1	.271
72	vit_tiny_r_s16_p8_384:augreg_in21k_ft_in1k	.271
73	nasnet_large-tf	.266
74	resnet-152_v2_pytorch	.260
75	resnet_152_v1	.255
76	inception_v4	.254
77	resnet50_byol	.254
78	antialias-resnet152	.253
79	effnetb1_cutmixpatch_augmix_robust32_e5_324x288	.250
80	resnet50_imagenet_full	.244
81	resnet101_imagenet_full	.242
82	antialiased-rnext101_32x8d	.240
83	AdvProp_efficientnet-b7	.238
84	densenet-201	.236
85	densenet_201_pytorch	.236
86	effnetb1_cutmixpatch_augmix_robust32_avge4e7_manylayers_324x288	.235
87	custom_model_cv_18_dagger_408	.235
88	cvt_cvt-13-224-in1k_4_LucyV4	.233
89	resnet50	.233
90	densenet-201-keras	.233
91	resnet34_imagenet_full	.233
92	AT_efficientnet-b2	.232
93	AdvProp_efficientnet-b2	.232
94	resnet50-moclr8deg	.229
95	efficientnet-b7	.228
96	resnet50-SIN_IN_IN	.228
97	resnet_SIN_IN_FT_IN	.228
98	resnet50_tutorial	.226
99	resnet_50_v1	.226
100	resnet50-sup	.226
101	resnet-50-pytorch	.226
102	resnet50_random_l2_perturb	.220
103	AdvProp_efficientnet-b6	.219
104	Res2Net50_26w_4s	.218
105	densenet-121-keras	.213
106	AT_efficientnet-b4	.213
107	AdvProp_efficientnet-b4	.213
108	densenet-121	.212
109	mobilenet_v2_0_5_224	.211
110	mobilenet_v2_1_4_224	.210
111	mobilenet_v2_1-4_224_pytorch	.210
112	inception_v4-tf	.210
113	grcnn_109	.208
114	convnext_small_imagenet_full_seed-0	.206
115	voneresnet-50-non_stochastic	.204
116	ReAlnet04	.204
117	texture_shape_resnet50_trained_on_SIN	.202
118	resnet-152_v1-tf	.196
119	resnet50_moco_v2	.191
120	cornet_s	.191
121	nasnet_mobile	.190
122	swin_small_patch4_window7_224:ms_in22k_ft_in1k	.189
123	mobilenet_v2_1.4_224-tf	.189
124	resnet-101_v1-tf	.189
125	BiT-S-R152x4	.189
126	vgg_16	.188
127	ReAlnet03	.187
128	resnet18_imagenet_full	.186
129	ReAlnet09	.184
130	antialiased-r50	.182
131	konkle_alexnetgn_ipcl_ref12_supervised_ipcl_aug	.181
132	ReAlnet_incorrect	.181
133	ReAlnet01	.181
134	deit_large_imagenet_full_seed-0	.181
135	resnet-50_v1-tf	.180
136	ReAlnet10	.179
137	grcnn	.179
138	resnet50_random_linf8_perturb	.177
139	BiT-S-R101x1	.177
140	blt_vs	.177
141	CORnet-S	.175
142	resnet50_linf_4_robust	.175
143	ReAlnet08	.173
144	resnet50_linf_8_robust	.173
145	BiT-S-R50x1	.173
146	mobilenet_v2_1_0_224	.171
147	inception_v1	.167
148	densenet-169	.166
149	ReAlnet06	.166
150	mobilenet_v1_0.75_224	.165
151	texture_shape_alexnet_trained_on_SIN	.164
152	mobilenet_v2_1_0_160	.164
153	BiT-S-R50x3	.163
154	resnet-18	.163
155	resnet50-vitoimagevidnet8	.162
156	voneresnet-50-robust	.162
157	inception_v2-tf	.160
158	deit_base_imagenet_full_seed-0	.160
159	AdvProp_efficientnet-b8	.159
160	ReAlnet05	.157
161	regnet_y_400mf	.157
162	mobilenet_v1_0.5_192	.157
163	vonealexnet_gaussian_noise_std4_fixed	.157
164	BiT-S-R152x2	.156
165	mobilenet_v1_1.0_224	.156
166	resnet50_l2_3_robust	.155
167	alexnet_reduced_aliasing_early_checkpoint	.155
168	inception_v1-tf	.154
169	alexnet_training_seed_03	.153
170	mobilenet_v2_1.0_224-tf	.153
171	ReAlnet07	.153
172	mobilenet_v2_1_3_224	.153
173	alexnet_random_linf8_perturb	.153
174	ReAlnet02	.153
175	mobilenet_v2_0.5_224-tf	.153
176	vonegrcnn_52e_full	.153
177	mobilenet_v2_1.3_224-tf	.152
178	resnet50_simclr	.152
179	densenet-169-keras	.149
180	mobilenet_v1_1.0_192	.147
181	vgg-16-keras	.146
182	0.5x_resnet-18_LC_w_sh_1_iter	.146
183	squeezenet1_1	.146
184	efficientnet_b1_imagenet_full	.145
185	vonegrcnn_47e	.144
186	resnet50_robust_l2_eps1	.144
187	resnet50-SIN	.144
188	efficientnet_b2_imagenet_full	.144
189	BiT-S-R101x3	.144
190	vgg_19	.142
191	alexnet_random_l2_3_perturb	.142
192	mobilenet_v2_0.5_96	.141
193	mobilenet_v1_1.0_128	.140
194	mobilenet_v2_1.0_160-tf	.137
195	vonegrcnn_62e_nobn	.137
196	mobilenet_v1_0.75_192	.136
197	resnet-34	.136
198	mobilenet_v2_1_0_128	.135
199	AdvProp_efficientnet-b0	.135
200	alexnet_early_checkpoint	.135
201	efficientnet_b0_imagenet_full	.134
202	mobilenet_v2_0.35_128	.134
203	convnext_small_imagenet_100_seed-0	.133
204	effnetb1_cutmix_augmix_sam_e1_5avg_424x377	.133
205	mobilenet_v2_0.35_96	.133
206	konkle_alexnetgn_ipcl_ref01_primary_model	.132
207	mobilenet_v2_1.0_128-tf	.132
208	mobilenet_v1_0.25_224	.132
209	mobilenet_v2_0.75_224-tf	.132
210	AlexNet_SIN	.132
211	vonegrcnn_52e_full	.132
212	imagenet_l2_3_0	.131
213	resnet50_robust_l2_eps3	.131
214	resnet-50-robust	.131
215	mobilenet_v2_1_0_192	.130
216	alexnet_ks_torevert	.128
217	alexnet	.128
218	alexnet	.128
219	alexnet	.128
220	alexnet	.128
221	alexnet-baseline	.128
222	alexnet	.128
223	alexnet	.128
224	alexnet	.128
225	alexnet	.128
226	alexnet	.128
227	mobilenet_v2_0.5_192-tf	.128
228	resnet50_imagenet_100_seed-0	.127
229	vgg-19-keras	.127
230	vonegrcnn_62e_nobn	.127
231	mobilenet_v2_0_75_224	.125
232	tf_efficientnetv2_s_in21ft1k_robust_linf12255_400x400	.125
233	mobilenet_v2_1.0_192-tf	.125
234	shufflenet_v2_x1_0	.123
235	mobilenet_v1_0.75_128	.122
236	alexnet	.121
237	voneresnet-50	.121
238	vonegrcnn_47e	.121
239	mobilenet_v2_0.75_96	.119
240	mobilenet_v2_0.5_160	.117
241	mobilenet_v1_1.0_160	.114
242	mobilenet_v1_0.25_192	.112
243	convnext_tiny_sup	.112
244	alexnet_l2_3_robust	.110
245	squeezenet1_0	.110
246	mobilenet_v2_0.5_128	.109
247	mobilenet_v2_0_75_192	.107
248	mobilenet_v2_0_5_192	.106
249	mobilenet_v2_0.35_160	.102
250	mobilenet_v2_0.35_192	.102
251	mobilenet_v1_0.5_128	.102
252	mobilenet_v2_0.75_160-tf	.101
253	nasnet_mobile-tf	.101
254	mobilenet_v2_0_75_160	.098
255	mobilenet_v1_0.25_160	.098
256	mobilenet_v1_0.75_160	.097
257	deit_small_imagenet_full_seed-0	.096
258	mobilenet_v2_0.75_128	.096
259	convnext_small_imagenet_1_seed-0	.096
260	mobilenet_v2_0.35_224	.093
261	convnext_small_imagenet_10_seed-0	.091
262	deit_small_imagenet_10_seed-0	.090
263	mobilenet_v2_0.75_192-tf	.090
264	deit_small_imagenet_1_seed-0	.087
265	mobilenet_v1_0.5_224	.083
266	mobilevit_small	.079
267	mobilenet_v2_1.0_96	.078
268	alexnet_training_seed_04	.068
269	alexnet_training_seed_08	.068
270	alexnet_training_seed_02	.068
271	alexnet_training_seed_10	.068
272	alexnet_training_seed_09	.067
273	alexnet_training_seed_07	.066
274	resnet50-meshes-lt-100-original-pretrained	.066
275	alexnet_robust_correct	.063
276	mobilenet_v1_0.25_128	.059
277	mobilenet_v1_0.5_160	.055
278	barlow-twins-resnet50	.053
279	resnet50-meshes-lt-100-original-scratch	.052
280	alexnet_training_seed_06	.050
281	resnet50-vicreg	.049
282	resnet50-barlow	.046
283	alexnet_linf_8_robust	.045
284	imagenet_l2_10_0	.034
285	resnet50_FractalDB	.032
286	resnet50-vicregl0p9	.032
287	resnet-50_untrained	.031
288	deit_small_imagenet_100_seed-0	.028
289	artResNet18_1	.013
290	resnet50_imagenet_1_seed-0	.011
291	resnet50-simclr-vissl	.011
292	resnet50-vicregl0p75	.011
293	resnet50_imagenet_10_seed-0	.003
294	alexnet_training_seed_05	.003
295	alexnet_training_seed_01	.003
296	resnet-18-LC_w_sh_1_iter_conv_init_m	X
297	ViT_B_32	X
298	ViT_B_16	X
299	r3m_resnet34_nocrop	X
300	resnet18-contrastive_predictive	X
301	omnivore_swinT	X
302	resnet-18-LC_untrained	X
303	resnet-18_untrained	X
304	resnet-50x2_untrained	X
305	dcgan	X
306	v1-pyr-nodown	X
307	unet_entire	X
308	CORnetZ_CIFAR10_bs32_20_04	X
309	pixels	X
310	dorinet_cornet_z	X
311	my-model	X
312	resnet50	X
313	openclip	X
314	ViT_L_16_imagenet1k	X
315	ViT_L_32	X
316	resnet18-simclr	X
317	resnet18-autoencoder	X
318	resnet18-relative_position	X
319	RN50	X
320	0.5x_resnet-18_LC_w_sh_100_iter	X
321	RN50	X
322	0.5x_resnet-18	X
323	resnet50-cifar	X
324	resnet50_pretrained_with_retinal_waves	X
325	resnet50_ImageNet	X
326	resnet18-contrastive_multiview	X
327	fulltest_microblockvf_nobottleneck_freeat_nopretrain_eps2_m4	X
328	resnet-18-LC_conv_init_m	X
329	resnet-18-LC_1st_conv_m	X
330	resnet18-supervised	X
331	r3m_resnet50_nocrop	X
332	resnet-18-LC_m	X
333	r3m_resnet34	X
334	yudixie_resnet50_imagenet1kpret_0_240908	X
335	r3m_resnet50	X
336	yudixie_resnet50_imagenet1kpret_0_240312	X
337	yudixie_resnet50_cat_obj_class_all_latents_0_240908	X
338	yudixie_resnet50_object_class_0_240908	X
339	0.5x_resnet-18_LC_w_sh_10_iter	X
340	vggface	X
341	my-model	X
342	resnet18-instance_recognition	X
343	CORnet-Z	X
344	yudixie_resnet50_category_class_0_240908	X
345	resnet18-colorization	X
346	r3m_resnet18	X
347	resnet18_imagenet21kP	X
348	ConvLSTM	X
349	PredRNN	X
350	yudixie_resnet50_distance_reg_0_240908	X
351	SimVP	X
352	TAU	X
353	MIM	X
354	hmax	X
355	resnet18-deepcluster	X
356	ReAlnet01_cornet	X
357	CORnetZ_CIFAR10	X
358	0.5x_resnet-18_LC	X
359	resnet18-local_aggregation	X
360	FrankRobWobv0	X
361	yudixie_resnet18_translation_reg_0_240719	X
362	resnet-18-LC_w_sh_10_iter_conv_init_m	X
363	ReAlnet02_cornet	X
364	ReAlnet03_cornet	X
365	yudixie_resnet18_distance_reg_0_240719	X
366	yudixie_resnet18_rotation_reg_0_240719	X
367	yudixie_resnet18_imagenet1kpret_0_240719	X
368	resnet-18-LC_w_sh_100_iter_m	X
369	yudixie_resnet50_random_0_240908	X
370	resnet18_ecoset_full	X
371	resnet34_ecoset_full	X
372	resnet50_ecoset_full	X
373	resnet-18-LC_w_sh_10_iter_m	X
374	resnet101_ecoset_full	X
375	yudixie_resnet50_distance_rotation_0_240908	X
376	yudixie_resnet50_rotation_reg_0_240908	X
377	resnet152_ecoset_full	X
378	yudixie_resnet18_distance_translation_0_240719	X
379	ReAlnet04_cornet	X
380	yudixie_resnet50_distance_translation_0_240908	X
381	resnet-18-LC_w_sh_1_iter_m	X
382	yudixie_resnet18_translation_rotation_0_240719	X
383	yudixie_resnet18_distance_rotation_0_240719	X
384	yudixie_resnet18_category_class_0_240719	X
385	ReAlnet05_cornet	X
386	yudixie_resnet18_random_0_240719	X
387	yudixie_resnet18_cat_obj_class_all_latents_0_240719	X
388	resnet-18_test_m	X
389	yudixie_resnet18_distance_translation_rotation_0_240719	X
390	yudixie_resnet18_object_class_0_240719	X
391	ReAlnet06_cornet	X
392	yudixie_resnet50_translation_reg_0_240908	X
393	ReAlnet07_cornet	X
394	yudixie_resnet50_translation_rotation_0_240908	X
395	ReAlnet08_cornet	X
396	VOneCORnet-S	X
397	resnet18-depth_prediction	X
398	ReAlnet09_cornet	X
399	omnivore_swinB	X
400	BiT-M-R152x2	X
401	omnivore_swinS	X
402	ReAlnet10_cornet	X
403	BiT-M-R50x1	X
404	BiT-M-R50x3	X
405	yudixie_resnet50_distance_translation_rotation_0_240908	X
406	ViT-B/32	X
407	resnet-18-LC_w_sh_100_iter_conv_init_m	X

Benchmark bibtex

@article{geirhos2021partial,
              title={Partial success in closing the gap between human and machine vision},
              author={Geirhos, Robert and Narayanappa, Kantharaju and Mitzkus, Benjamin and Thieringer, Tizian and Bethge, Matthias and Wichmann, Felix A and Brendel, Wieland},
              journal={Advances in Neural Information Processing Systems},
              volume={34},
              year={2021},
              url={https://openreview.net/forum?id=QkljT4mrfs}
        }

Ceiling

0.33.

Note that scores are relative to this ceiling.

Data: Geirhos2021cueconflict

Metric: error_consistency