### qwen3-8b · medial · joint
   idiom config:    {'model': 'Qwen/Qwen3-8B', 'reduction': 'joint', 'medial_only': True, 'dtype': 'bfloat16', 'dataset': '/home/prada/PID_evaluation/data/dataset.tsv', 'num_idioms': 18, 'syn_reg_eps': 0.01}
   nonidiom config: {'model': 'Qwen/Qwen3-8B', 'reduction': 'joint', 'medial_only': True, 'dtype': 'bfloat16', '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
   1.1527     1.1232  [  1.1200,  1.1878]

== non-idioms :: ratio_u_idiom ==  (N=18 phrases)
     mean     median                    95% CI
   1.0388     1.0312  [  1.0265,  1.0524]
  cross-dataset ratio_u_idiom: idioms - nonidioms  Δ=+0.1139  CI=[+0.0786,+0.1521]  *

== idioms :: ratio_s_idiom ==  (N=9 phrases)  (9 non-finite dropped)
     mean     median                    95% CI
   1.0016     1.0001  [  1.0001,  1.0038]

== non-idioms :: ratio_s_idiom ==  (N=1 phrases)  (17 non-finite dropped)
     mean     median                    95% CI
   1.0024     1.0024  [  1.0024,  1.0024]
  cross-dataset ratio_s_idiom: idioms - nonidioms  Δ=-0.0008  CI=[-0.0023,+0.0014]