computed_values/expanded_baselines.json

{
  "n_trained": 1008,
  "n_evaluated": 788,
  "n_tpe_trained": 504,
  "n_tpe_evaluated": 394,
  "n_random_trained": 504,
  "n_random_evaluated": 394,
  "best_overall_mAP": 0.737,
  "best_tpe_mAP": 0.6935,
  "best_random_mAP": 0.737,
  "best_non_vjepa2_mAP": 0.737,
  "best_vjepa2_mAP": 0.689,
  "vjepa2_gap": 0.048,
  "mean_mAP": 0.5919,
  "median_mAP": 0.6011,
  "backbone_ranking": {
    "dinov3_vitb16": {
      "n": 130,
      "best": 0.737,
      "mean": 0.6204
    },
    "siglip2_vit_b16": {
      "n": 121,
      "best": 0.733,
      "mean": 0.6218
    },
    "dinov3_vitl16": {
      "n": 108,
      "best": 0.733,
      "mean": 0.5848
    },
    "dinov2_vitb14": {
      "n": 109,
      "best": 0.725,
      "mean": 0.6151
    },
    "vjepa2_vitl": {
      "n": 320,
      "best": 0.689,
      "mean": 0.5634
    }
  },
  "encoder_ranking": {
    "gru_temporal": {
      "n": 182,
      "best": 0.737,
      "mean": 0.5769
    },
    "lstm_temporal": {
      "n": 103,
      "best": 0.736,
      "mean": 0.5855
    },
    "hybrid_temporal": {
      "n": 115,
      "best": 0.733,
      "mean": 0.5799
    },
    "bimamba": {
      "n": 149,
      "best": 0.733,
      "mean": 0.6032
    },
    "zipformer_temporal": {
      "n": 112,
      "best": 0.725,
      "mean": 0.5853
    },
    "retention_temporal": {
      "n": 127,
      "best": 0.71,
      "mean": 0.6217
    }
  }
}