{
  "count": 105,
  "default_weights": [
    "0.6",
    "0.2",
    "0.2"
  ],
  "winner": 879,
  "top10": [
    {
      "rank": 1,
      "id": 879,
      "name": "FusionGatedFIRNet",
      "internal_name": "delta_net_psfr",
      "index_lower": "88.32495660757220696938013443",
      "index_upper": "97.29275578790141896938013443",
      "feasible_rank_best": 1,
      "feasible_rank_worst": 9,
      "weight_grid_best": 1,
      "weight_grid_worst": 5,
      "accuracy8_percent": "40.1037500",
      "accuracy7_percent": "44.02285714285714285714285714",
      "final_training_loss": "3.4048",
      "parameters_m": "466.90",
      "raw_fitness_NOT_ADDED": "2.5516100409835394",
      "accuracy_percentile": "0.9807692307692307692307692308",
      "inverse_loss_percentile": "0.9223300970873786407766990291",
      "normalized_loss_lower": "0.9223300970873786407766990291",
      "normalized_loss_upper": "0.9223300970873786407766990291",
      "judge_lower_bound": "0.5516100409835394",
      "judge_upper_bound": "1",
      "judge_status": "Unpublished; bounded, not measured",
      "matched_wiki_ppl": "NA",
      "matched_lambada_ppl": "NA",
      "final_1_3B_15B_wiki_ppl_NOT_WEIGHTED": "15.77",
      "final_1_3B_15B_lambada_ppl_NOT_WEIGHTED": "12.34",
      "final_1_3B_15B_accuracy14_NOT_WEIGHTED": "52.01",
      "coverage": "8/8 accuracy; loss measured; 0/2 held-out PPL; unpublished judge component",
      "arc_challenge_percent": "23.2900",
      "arc_easy_percent": "46.8900",
      "boolq_percent": "61.5600",
      "hellaswag_percent": "28.800",
      "piqa_percent": "60.500",
      "social_iqa_percent": "35.8200",
      "winogrande_percent": "51.300",
      "lambada_openai_percent": "12.6700"
    },
    {
      "rank": 2,
      "id": 908,
      "name": "AdaptiveFusionNet",
      "internal_name": "delta_net_aft",
      "index_lower": "87.25148555298254428678117999",
      "index_upper": "94.01418969380134428678117999",
      "feasible_rank_best": 1,
      "feasible_rank_worst": 12,
      "weight_grid_best": 1,
      "weight_grid_worst": 4,
      "accuracy8_percent": "40.0087500",
      "accuracy7_percent": "44.08285714285714285714285714",
      "final_training_loss": "3.4061",
      "parameters_m": "615.54",
      "raw_fitness_NOT_ADDED": "2.66186479295906",
      "accuracy_percentile": "0.9423076923076923076923076923",
      "inverse_loss_percentile": "0.8737864077669902912621359223",
      "normalized_loss_lower": "0.8737864077669902912621359223",
      "normalized_loss_upper": "0.8737864077669902912621359223",
      "judge_lower_bound": "0.66186479295906",
      "judge_upper_bound": "1",
      "judge_status": "Unpublished; bounded, not measured",
      "matched_wiki_ppl": "NA",
      "matched_lambada_ppl": "NA",
      "final_1_3B_15B_wiki_ppl_NOT_WEIGHTED": "NA",
      "final_1_3B_15B_lambada_ppl_NOT_WEIGHTED": "NA",
      "final_1_3B_15B_accuracy14_NOT_WEIGHTED": "NA",
      "coverage": "8/8 accuracy; loss measured; 0/2 held-out PPL; unpublished judge component",
      "arc_challenge_percent": "23.4600",
      "arc_easy_percent": "48.400",
      "boolq_percent": "60.2400",
      "hellaswag_percent": "28.7500",
      "piqa_percent": "61.2100",
      "social_iqa_percent": "35.0600",
      "winogrande_percent": "51.4600",
      "lambada_openai_percent": "11.4900"
    },
    {
      "rank": 3,
      "id": 814,
      "name": "GateDivergeTransformer",
      "internal_name": "delta_net_htfr",
      "index_lower": "86.51475041078754387154592980",
      "index_upper": "94.19156086631814787154592980",
      "feasible_rank_best": 1,
      "feasible_rank_worst": 13,
      "weight_grid_best": 2,
      "weight_grid_worst": 6,
      "accuracy8_percent": "40.06500",
      "accuracy7_percent": "44.1000",
      "final_training_loss": "3.4103",
      "parameters_m": "817.77",
      "raw_fitness_NOT_ADDED": "2.6161594772234698",
      "accuracy_percentile": "0.9711538461538461538461538462",
      "inverse_loss_percentile": "0.7961165048543689320388349515",
      "normalized_loss_lower": "0.7961165048543689320388349515",
      "normalized_loss_upper": "0.7961165048543689320388349515",
      "judge_lower_bound": "0.6161594772234698",
      "judge_upper_bound": "1",
      "judge_status": "Unpublished; bounded, not measured",
      "matched_wiki_ppl": "NA",
      "matched_lambada_ppl": "NA",
      "final_1_3B_15B_wiki_ppl_NOT_WEIGHTED": "NA",
      "final_1_3B_15B_lambada_ppl_NOT_WEIGHTED": "NA",
      "final_1_3B_15B_accuracy14_NOT_WEIGHTED": "NA",
      "coverage": "8/8 accuracy; loss measured; 0/2 held-out PPL; unpublished judge component",
      "arc_challenge_percent": "24.4900",
      "arc_easy_percent": "47.4700",
      "boolq_percent": "59.300",
      "hellaswag_percent": "28.4300",
      "piqa_percent": "60.8300",
      "social_iqa_percent": "35.9300",
      "winogrande_percent": "52.2500",
      "lambada_openai_percent": "11.8200"
    },
    {
      "rank": 4,
      "id": 1730,
      "name": "AdaptiveEntropyRouter",
      "internal_name": "delta_net_afef",
      "index_lower": "86.27248771066310056758775205",
      "index_upper": "93.04891710231516056758775205",
      "feasible_rank_best": 1,
      "feasible_rank_worst": 13,
      "weight_grid_best": 2,
      "weight_grid_worst": 6,
      "accuracy8_percent": "39.9787500",
      "accuracy7_percent": "44.31285714285714285714285714",
      "final_training_loss": "3.4066",
      "parameters_m": "468.48",
      "raw_fitness_NOT_ADDED": "2.661178530417397",
      "accuracy_percentile": "0.9326923076923076923076923077",
      "inverse_loss_percentile": "0.8543689320388349514563106796",
      "normalized_loss_lower": "0.8543689320388349514563106796",
      "normalized_loss_upper": "0.8543689320388349514563106796",
      "judge_lower_bound": "0.661178530417397",
      "judge_upper_bound": "1",
      "judge_status": "Unpublished; bounded, not measured",
      "matched_wiki_ppl": "NA",
      "matched_lambada_ppl": "NA",
      "final_1_3B_15B_wiki_ppl_NOT_WEIGHTED": "NA",
      "final_1_3B_15B_lambada_ppl_NOT_WEIGHTED": "NA",
      "final_1_3B_15B_accuracy14_NOT_WEIGHTED": "NA",
      "coverage": "8/8 accuracy; loss measured; 0/2 held-out PPL; unpublished judge component",
      "arc_challenge_percent": "23.7200",
      "arc_easy_percent": "47.6400",
      "boolq_percent": "61.8300",
      "hellaswag_percent": "28.7800",
      "piqa_percent": "60.8800",
      "social_iqa_percent": "35.0100",
      "winogrande_percent": "52.3300",
      "lambada_openai_percent": "9.6400"
    },
    {
      "rank": 5,
      "id": 1475,
      "name": "TempMixAnnealRouter",
      "internal_name": "delta_net_len_hgate_mixanneal",
      "index_lower": "86.18992095831574352501867065",
      "index_upper": "92.88274831964152352501867065",
      "feasible_rank_best": 1,
      "feasible_rank_worst": 13,
      "weight_grid_best": 1,
      "weight_grid_worst": 8,
      "accuracy8_percent": "39.937500",
      "accuracy7_percent": "44.01285714285714285714285714",
      "final_training_loss": "3.4014",
      "parameters_m": "817.01",
      "raw_fitness_NOT_ADDED": "2.665358631933711",
      "accuracy_percentile": "0.8846153846153846153846153846",
      "inverse_loss_percentile": "0.9902912621359223300970873786",
      "normalized_loss_lower": "0.9902912621359223300970873786",
      "normalized_loss_upper": "0.9902912621359223300970873786",
      "judge_lower_bound": "0.665358631933711",
      "judge_upper_bound": "1",
      "judge_status": "Unpublished; bounded, not measured",
      "matched_wiki_ppl": "NA",
      "matched_lambada_ppl": "NA",
      "final_1_3B_15B_wiki_ppl_NOT_WEIGHTED": "NA",
      "final_1_3B_15B_lambada_ppl_NOT_WEIGHTED": "NA",
      "final_1_3B_15B_accuracy14_NOT_WEIGHTED": "NA",
      "coverage": "8/8 accuracy; loss measured; 0/2 held-out PPL; unpublished judge component",
      "arc_challenge_percent": "23.7200",
      "arc_easy_percent": "47.8500",
      "boolq_percent": "60.0600",
      "hellaswag_percent": "28.6700",
      "piqa_percent": "60.3400",
      "social_iqa_percent": "35.6700",
      "winogrande_percent": "51.7800",
      "lambada_openai_percent": "11.4100"
    },
    {
      "rank": 6,
      "id": 559,
      "name": "OutputAwareMultiScaleRouter",
      "internal_name": "delta_net_mor",
      "index_lower": "85.35827047852929961165048544",
      "index_upper": "98.83495145631067961165048544",
      "feasible_rank_best": 1,
      "feasible_rank_worst": 13,
      "weight_grid_best": 2,
      "weight_grid_worst": 8,
      "accuracy8_percent": "40.3012500",
      "accuracy7_percent": "44.57571428571428571428571429",
      "final_training_loss": "3.4046",
      "parameters_m": "464.64",
      "raw_fitness_NOT_ADDED": "2.326165951110931",
      "accuracy_percentile": "1",
      "inverse_loss_percentile": "0.9417475728155339805825242718",
      "normalized_loss_lower": "0.9417475728155339805825242718",
      "normalized_loss_upper": "0.9417475728155339805825242718",
      "judge_lower_bound": "0.326165951110931",
      "judge_upper_bound": "1",
      "judge_status": "Unpublished; bounded, not measured",
      "matched_wiki_ppl": "NA",
      "matched_lambada_ppl": "NA",
      "final_1_3B_15B_wiki_ppl_NOT_WEIGHTED": "NA",
      "final_1_3B_15B_lambada_ppl_NOT_WEIGHTED": "NA",
      "final_1_3B_15B_accuracy14_NOT_WEIGHTED": "NA",
      "coverage": "8/8 accuracy; loss measured; 0/2 held-out PPL; unpublished judge component",
      "arc_challenge_percent": "26.1100",
      "arc_easy_percent": "48.7400",
      "boolq_percent": "61.500",
      "hellaswag_percent": "28.6700",
      "piqa_percent": "59.8500",
      "social_iqa_percent": "35.3100",
      "winogrande_percent": "51.8500",
      "lambada_openai_percent": "10.3800"
    },
    {
      "rank": 7,
      "id": 1354,
      "name": "FusionGate-MS",
      "internal_name": "delta_net_omsgf",
      "index_lower": "82.41315671273591340403286034",
      "index_upper": "95.17363704256908140403286034",
      "feasible_rank_best": 1,
      "feasible_rank_worst": 14,
      "weight_grid_best": 7,
      "weight_grid_worst": 9,
      "accuracy8_percent": "40.03500",
      "accuracy7_percent": "44.1800",
      "final_training_loss": "3.4058",
      "parameters_m": "443.87",
      "raw_fitness_NOT_ADDED": "2.3619759835083416",
      "accuracy_percentile": "0.9519230769230769230769230769",
      "inverse_loss_percentile": "0.9029126213592233009708737864",
      "normalized_loss_lower": "0.9029126213592233009708737864",
      "normalized_loss_upper": "0.9029126213592233009708737864",
      "judge_lower_bound": "0.3619759835083416",
      "judge_upper_bound": "1",
      "judge_status": "Unpublished; bounded, not measured",
      "matched_wiki_ppl": "NA",
      "matched_lambada_ppl": "NA",
      "final_1_3B_15B_wiki_ppl_NOT_WEIGHTED": "NA",
      "final_1_3B_15B_lambada_ppl_NOT_WEIGHTED": "NA",
      "final_1_3B_15B_accuracy14_NOT_WEIGHTED": "NA",
      "coverage": "8/8 accuracy; loss measured; 0/2 held-out PPL; unpublished judge component",
      "arc_challenge_percent": "24.400",
      "arc_easy_percent": "48.1100",
      "boolq_percent": "59.8200",
      "hellaswag_percent": "29.0900",
      "piqa_percent": "60.7200",
      "social_iqa_percent": "35.8200",
      "winogrande_percent": "51.300",
      "lambada_openai_percent": "11.0200"
    },
    {
      "rank": 8,
      "id": 458,
      "name": "DualScaleMemoryRouter",
      "internal_name": "delta_net_dlgm",
      "index_lower": "82.00003962068411698282300224",
      "index_upper": "92.87154592979835698282300224",
      "feasible_rank_best": 1,
      "feasible_rank_worst": 15,
      "weight_grid_best": 6,
      "weight_grid_worst": 9,
      "accuracy8_percent": "39.947500",
      "accuracy7_percent": "44.13285714285714285714285714",
      "final_training_loss": "3.4047",
      "parameters_m": "464.64",
      "raw_fitness_NOT_ADDED": "2.456424684544288",
      "accuracy_percentile": "0.9038461538461538461538461538",
      "inverse_loss_percentile": "0.9320388349514563106796116505",
      "normalized_loss_lower": "0.9320388349514563106796116505",
      "normalized_loss_upper": "0.9320388349514563106796116505",
      "judge_lower_bound": "0.456424684544288",
      "judge_upper_bound": "1",
      "judge_status": "Unpublished; bounded, not measured",
      "matched_wiki_ppl": "NA",
      "matched_lambada_ppl": "NA",
      "final_1_3B_15B_wiki_ppl_NOT_WEIGHTED": "NA",
      "final_1_3B_15B_lambada_ppl_NOT_WEIGHTED": "NA",
      "final_1_3B_15B_accuracy14_NOT_WEIGHTED": "NA",
      "coverage": "8/8 accuracy; loss measured; 0/2 held-out PPL; unpublished judge component",
      "arc_challenge_percent": "25.0900",
      "arc_easy_percent": "48.0200",
      "boolq_percent": "58.4700",
      "hellaswag_percent": "28.8800",
      "piqa_percent": "60.7200",
      "social_iqa_percent": "35.2600",
      "winogrande_percent": "52.4900",
      "lambada_openai_percent": "10.6500"
    },
    {
      "rank": 9,
      "id": 1311,
      "name": "HybridFlowNet",
      "internal_name": "delta_net_cagf_rc_pf_hybrid",
      "index_lower": "81.49469997556212426138909634",
      "index_upper": "90.97647498132935026138909634",
      "feasible_rank_best": 1,
      "feasible_rank_worst": 17,
      "weight_grid_best": 5,
      "weight_grid_worst": 9,
      "accuracy8_percent": "40.1912500",
      "accuracy7_percent": "44.24714285714285714285714286",
      "final_training_loss": "3.4238",
      "parameters_m": "439.03",
      "raw_fitness_NOT_ADDED": "2.5259112497116387",
      "accuracy_percentile": "0.9903846153846153846153846154",
      "inverse_loss_percentile": "0.5776699029126213592233009709",
      "normalized_loss_lower": "0.5776699029126213592233009709",
      "normalized_loss_upper": "0.5776699029126213592233009709",
      "judge_lower_bound": "0.5259112497116387",
      "judge_upper_bound": "1",
      "judge_status": "Unpublished; bounded, not measured",
      "matched_wiki_ppl": "NA",
      "matched_lambada_ppl": "NA",
      "final_1_3B_15B_wiki_ppl_NOT_WEIGHTED": "NA",
      "final_1_3B_15B_lambada_ppl_NOT_WEIGHTED": "NA",
      "final_1_3B_15B_accuracy14_NOT_WEIGHTED": "NA",
      "coverage": "8/8 accuracy; loss measured; 0/2 held-out PPL; unpublished judge component",
      "arc_challenge_percent": "25.0900",
      "arc_easy_percent": "47.9800",
      "boolq_percent": "61.9300",
      "hellaswag_percent": "28.2800",
      "piqa_percent": "60.9400",
      "social_iqa_percent": "34.600",
      "winogrande_percent": "50.9100",
      "lambada_openai_percent": "11.800"
    },
    {
      "rank": 10,
      "id": 537,
      "name": "FusionGateNet_v3",
      "internal_name": "delta_net_mshmfv2",
      "index_lower": "74.78687075377542781179985063",
      "index_upper": "86.89693801344286781179985063",
      "feasible_rank_best": 3,
      "feasible_rank_worst": 30,
      "weight_grid_best": 10,
      "weight_grid_worst": 17,
      "accuracy8_percent": "39.8187500",
      "accuracy7_percent": "43.89428571428571428571428571",
      "final_training_loss": "3.4064",
      "parameters_m": "490.52",
      "raw_fitness_NOT_ADDED": "2.394496637016628",
      "accuracy_percentile": "0.8269230769230769230769230769",
      "inverse_loss_percentile": "0.8640776699029126213592233010",
      "normalized_loss_lower": "0.8640776699029126213592233010",
      "normalized_loss_upper": "0.8640776699029126213592233010",
      "judge_lower_bound": "0.394496637016628",
      "judge_upper_bound": "1",
      "judge_status": "Unpublished; bounded, not measured",
      "matched_wiki_ppl": "NA",
      "matched_lambada_ppl": "NA",
      "final_1_3B_15B_wiki_ppl_NOT_WEIGHTED": "NA",
      "final_1_3B_15B_lambada_ppl_NOT_WEIGHTED": "NA",
      "final_1_3B_15B_accuracy14_NOT_WEIGHTED": "NA",
      "coverage": "8/8 accuracy; loss measured; 0/2 held-out PPL; unpublished judge component",
      "arc_challenge_percent": "24.6600",
      "arc_easy_percent": "48.4400",
      "boolq_percent": "56.0200",
      "hellaswag_percent": "28.7700",
      "piqa_percent": "60.9400",
      "social_iqa_percent": "36.1800",
      "winogrande_percent": "52.2500",
      "lambada_openai_percent": "11.2900"
    }
  ],
  "sensitivity": [
    {
      "name": "Default 60/20/20",
      "weights": [
        "0.6",
        "0.2",
        "0.2"
      ],
      "normalization": [
        "an",
        "ln"
      ],
      "winner": 879,
      "winner_name": "FusionGatedFIRNet",
      "score": "88.32495660757220696938013443",
      "top5": [
        879,
        908,
        814,
        1730,
        1475
      ]
    },
    {
      "name": "Equal thirds",
      "weights": [
        "0.3333333333333333333333333333",
        "0.3333333333333333333333333333",
        "0.3333333333333333333333333333"
      ],
      "normalization": [
        "an",
        "ln"
      ],
      "winner": 1475,
      "winner_name": "TempMixAnnealRouter",
      "score": "84.67550928950059818272342543",
      "top5": [
        1475,
        908,
        879,
        1730,
        814
      ]
    },
    {
      "name": "Accuracy-heavy 80/10/10",
      "weights": [
        "0.8",
        "0.1",
        "0.1"
      ],
      "normalization": [
        "an",
        "ln"
      ],
      "winner": 879,
      "winner_name": "FusionGatedFIRNet",
      "score": "93.20093984224764194622852875",
      "top5": [
        879,
        559,
        814,
        908,
        1311
      ]
    },
    {
      "name": "Judge-heavy 40/20/40",
      "weights": [
        "0.4",
        "0.2",
        "0.4"
      ],
      "normalization": [
        "an",
        "ln"
      ],
      "winner": 1475,
      "winner_name": "TempMixAnnealRouter",
      "score": "81.80478590468227121732636295",
      "top5": [
        1475,
        908,
        1730,
        879,
        814
      ]
    },
    {
      "name": "Min-max normalization",
      "weights": [
        "0.6",
        "0.2",
        "0.2"
      ],
      "normalization": [
        "amm",
        "lmm"
      ],
      "winner": 559,
      "winner_name": "OutputAwareMultiScaleRouter",
      "score": "85.06435245686907592705167173",
      "top5": [
        559,
        879,
        908,
        814,
        1475
      ]
    },
    {
      "name": "Per-task percentile normalization",
      "weights": [
        "0.6",
        "0.2",
        "0.2"
      ],
      "normalization": [
        "apt",
        "ln"
      ],
      "winner": 1475,
      "winner_name": "TempMixAnnealRouter",
      "score": "72.52405557370035890963405526",
      "top5": [
        1475,
        908,
        537,
        814,
        458
      ]
    }
  ],
  "grid_count": 11,
  "grid_winner_counts": {
    "1475": 2,
    "908": 3,
    "879": 6
  },
  "missing_ppl_count": 105,
  "unpublished_judge_count": 105,
  "missing_final_loss_ids": [
    562
  ],
  "commit": "3113c51b04a1d4a7520a3b0238c6a08d3194d58f",
  "source_blob_sha": "e403531e663cf24c800002720d293cccf77f94df",
  "input_sha256": "dbc6be1a32e34639bc320f72b6126daf7d4228a6fdecbbe887b8b9d807cd0b53"
}
