### gpt2 · full · geo
   idiom config:    {'model': 'gpt2', 'reduction': 'geometric_mean', 'medial_only': False, 'dtype': 'float32', 'dataset': '/home/prada/PID_evaluation/data/dataset.tsv', 'num_idioms': 18, 'syn_reg_eps': 0.01}
   nonidiom config: {'model': 'gpt2', 'reduction': 'geometric_mean', 'medial_only': False, 'dtype': 'float32', 'dataset': '/home/prada/PID_evaluation/data/nonidioms_dataset.tsv', 'num_idioms': 18, 'syn_reg_eps': 0.01}

== idioms :: ratio_u_idiom ==  (N=18 phrases)
     mean     median                    95% CI
   1.1097     1.0917  [  1.0864,  1.1366]

== non-idioms :: ratio_u_idiom ==  (N=18 phrases)
     mean     median                    95% CI
   1.0471     1.0485  [  1.0372,  1.0574]
  cross-dataset ratio_u_idiom: idioms - nonidioms  Δ=+0.0626  CI=[+0.0369,+0.0913]  *

== idioms :: ratio_s_idiom ==  (N=5 phrases)  (13 non-finite dropped)
     mean     median                    95% CI
   1.1940     1.2121  [  1.1431,  1.2297]

== non-idioms :: ratio_s_idiom ==  (N=1 phrases)  (17 non-finite dropped)
     mean     median                    95% CI
   1.3821     1.3821  [  1.3821,  1.3821]
  cross-dataset ratio_s_idiom: idioms - nonidioms  Δ=-0.1881  CI=[-0.2389,-0.1524]  *