{ "optimizer": "PairAdam", "stsb_dev_spearman_before": 0.4551, "stsb_dev_spearman_after": 0.7179, "stsb_test_spearman_before": 0.2849, "stsb_test_spearman_after": 0.6242, "stsb_test_pearson_before": 0.2981, "stsb_test_pearson_after": 0.633, "hessian_pair_adam_ablation": { "stsb_dev_spearman_after": 0.7278, "stsb_dev_pearson_after": 0.6917, "stsb_test_spearman_after": 0.6288, "stsb_test_pearson_after": 0.6316 }, "adamw_ablation": { "stsb_dev_spearman_after": 0.7289, "stsb_test_spearman_after": 0.6317, "stsb_test_pearson_after": 0.6372 }, "parameters": 29305497, "trainable_parameters": 9891865, "optimizer_ablations_8way": { "polar_natural_adam": { "stsb_dev_spearman_after": 0.7262, "stsb_dev_pearson_after": 0.702, "stsb_test_spearman_after": 0.6471, "stsb_test_pearson_after": 0.6494 }, "cyclide_adam": { "s1_dev_spearman": 0.6255, "stsb_dev_spearman_after": 0.7368, "stsb_dev_pearson_after": 0.7179, "stsb_test_spearman_after": 0.6287, "stsb_test_pearson_after": 0.6476, "rocm_elapsed_sec": 380.42 }, "hessian_pair_adam_1x_coupling": { "stsb_dev_spearman_after": 0.7242, "stsb_dev_pearson_after": 0.6876, "stsb_test_spearman_after": 0.6337, "stsb_test_pearson_after": 0.6369 }, "rotor_adam": { "stsb_dev_spearman_after": 0.7322, "stsb_dev_pearson_after": 0.7099, "stsb_test_spearman_after": 0.6329, "stsb_test_pearson_after": 0.6309 }, "toroidal_lion": { "stsb_dev_spearman_after": 0.7289, "stsb_dev_pearson_after": 0.697, "stsb_test_spearman_after": 0.6319, "stsb_test_pearson_after": 0.6341 }, "adamw": { "stsb_dev_spearman_after": 0.7289, "stsb_test_spearman_after": 0.6317, "stsb_test_pearson_after": 0.6372 }, "hessian_pair_adam_3x_coupling": { "stsb_dev_spearman_after": 0.7278, "stsb_dev_pearson_after": 0.6917, "stsb_test_spearman_after": 0.6288, "stsb_test_pearson_after": 0.6316 }, "pair_adam_3x_coupling": { "stsb_dev_spearman_after": 0.7179, "stsb_test_spearman_after": 0.6242, "stsb_test_pearson_after": 0.633 }, "toroidal_muon": { "stsb_dev_spearman_after": 0.7361, "stsb_dev_pearson_after": 0.7105, "stsb_test_spearman_after": 0.6197, "stsb_test_pearson_after": 0.6304 } }, "optimizer_ablations_9way": { "polar_natural_adam": { "stsb_dev_spearman_after": 0.7262, "stsb_dev_pearson_after": 0.702, "stsb_test_spearman_after": 0.6471, "stsb_test_pearson_after": 0.6494 }, "cyclide_adam": { "s1_dev_spearman": 0.6255, "stsb_dev_spearman_after": 0.7368, "stsb_dev_pearson_after": 0.7179, "stsb_test_spearman_after": 0.6287, "stsb_test_pearson_after": 0.6476, "rocm_elapsed_sec": 380.42 }, "hessian_pair_adam_1x_coupling": { "stsb_dev_spearman_after": 0.7242, "stsb_dev_pearson_after": 0.6876, "stsb_test_spearman_after": 0.6337, "stsb_test_pearson_after": 0.6369 }, "rotor_adam": { "stsb_dev_spearman_after": 0.7322, "stsb_dev_pearson_after": 0.7099, "stsb_test_spearman_after": 0.6329, "stsb_test_pearson_after": 0.6309 }, "toroidal_lion": { "stsb_dev_spearman_after": 0.7289, "stsb_dev_pearson_after": 0.697, "stsb_test_spearman_after": 0.6319, "stsb_test_pearson_after": 0.6341 }, "adamw": { "stsb_dev_spearman_after": 0.7289, "stsb_test_spearman_after": 0.6317, "stsb_test_pearson_after": 0.6372 }, "hessian_pair_adam_3x_coupling": { "stsb_dev_spearman_after": 0.7278, "stsb_dev_pearson_after": 0.6917, "stsb_test_spearman_after": 0.6288, "stsb_test_pearson_after": 0.6316 }, "pair_adam_3x_coupling": { "stsb_dev_spearman_after": 0.7179, "stsb_test_spearman_after": 0.6242, "stsb_test_pearson_after": 0.633 }, "toroidal_muon": { "stsb_dev_spearman_after": 0.7361, "stsb_dev_pearson_after": 0.7105, "stsb_test_spearman_after": 0.6197, "stsb_test_pearson_after": 0.6304 } }, "optimizer_ablations_16way": { "symplectic_adam": { "s1_dev_spearman": 0.6198, "stsb_dev_spearman_after": 0.736, "stsb_dev_pearson_after": 0.7087, "stsb_test_spearman_after": 0.6516, "stsb_test_pearson_after": 0.6564, "rocm_elapsed_sec": 243.28 }, "huygens_symplectic_adam": { "s1_dev_spearman": 0.6282, "stsb_dev_spearman_after": 0.7368, "stsb_dev_pearson_after": 0.709, "stsb_test_spearman_after": 0.6509, "stsb_test_pearson_after": 0.6548, "rocm_elapsed_sec": 236.41 }, "polar_natural_adam": { "stsb_dev_spearman_after": 0.7262, "stsb_dev_pearson_after": 0.702, "stsb_test_spearman_after": 0.6471, "stsb_test_pearson_after": 0.6494 }, "gyric_adam": { "s1_dev_spearman": 0.6221, "stsb_dev_spearman_after": 0.7304, "stsb_dev_pearson_after": 0.7023, "stsb_test_spearman_after": 0.6455, "stsb_test_pearson_after": 0.6503, "rocm_elapsed_sec": 419.85 }, "huygens_adam": { "s1_dev_spearman": 0.6383, "stsb_dev_spearman_after": 0.7295, "stsb_dev_pearson_after": 0.7035, "stsb_test_spearman_after": 0.642, "stsb_test_pearson_after": 0.6469, "rocm_elapsed_sec": 331.8 }, "hessian_pair_adam_1x_coupling": { "stsb_dev_spearman_after": 0.7242, "stsb_dev_pearson_after": 0.6876, "stsb_test_spearman_after": 0.6337, "stsb_test_pearson_after": 0.6369 }, "beltrami_adam": { "s1_dev_spearman": 0.6337, "stsb_dev_spearman_after": 0.7363, "stsb_dev_pearson_after": 0.7068, "stsb_test_spearman_after": 0.6333, "stsb_test_pearson_after": 0.6328, "rocm_elapsed_sec": 321.44 }, "hannay_adam": { "s1_dev_spearman": 0.6115, "stsb_dev_spearman_after": 0.7257, "stsb_dev_pearson_after": 0.699, "stsb_test_spearman_after": 0.6333, "stsb_test_pearson_after": 0.633, "rocm_elapsed_sec": 239.35 }, "rotor_adam": { "stsb_dev_spearman_after": 0.7322, "stsb_dev_pearson_after": 0.7099, "stsb_test_spearman_after": 0.6329, "stsb_test_pearson_after": 0.6309 }, "toroidal_lion": { "stsb_dev_spearman_after": 0.7289, "stsb_dev_pearson_after": 0.697, "stsb_test_spearman_after": 0.6319, "stsb_test_pearson_after": 0.6341 }, "adamw": { "stsb_dev_spearman_after": 0.7289, "stsb_test_spearman_after": 0.6317, "stsb_test_pearson_after": 0.6372 }, "poncelet_adam": { "s1_dev_spearman": 0.6262, "stsb_dev_spearman_after": 0.7275, "stsb_dev_pearson_after": 0.6987, "stsb_test_spearman_after": 0.6315, "stsb_test_pearson_after": 0.6374, "rocm_elapsed_sec": 267.34 }, "hessian_pair_adam_3x_coupling": { "stsb_dev_spearman_after": 0.7278, "stsb_dev_pearson_after": 0.6917, "stsb_test_spearman_after": 0.6288, "stsb_test_pearson_after": 0.6316 }, "cyclide_adam": { "s1_dev_spearman": 0.6255, "stsb_dev_spearman_after": 0.7368, "stsb_dev_pearson_after": 0.7179, "stsb_test_spearman_after": 0.6287, "stsb_test_pearson_after": 0.6476, "rocm_elapsed_sec": 380.42 }, "rindler_adam": { "s1_dev_spearman": 0.6208, "stsb_dev_spearman_after": 0.7318, "stsb_dev_pearson_after": 0.7069, "stsb_test_spearman_after": 0.6264, "stsb_test_pearson_after": 0.6353, "rocm_elapsed_sec": 271.68 }, "pair_adam_3x_coupling": { "stsb_dev_spearman_after": 0.7179, "stsb_test_spearman_after": 0.6242, "stsb_test_pearson_after": 0.633 }, "toroidal_muon": { "stsb_dev_spearman_after": 0.7361, "stsb_dev_pearson_after": 0.7105, "stsb_test_spearman_after": 0.6197, "stsb_test_pearson_after": 0.6304 } }, "optimizer_ablations_17way": { "symplectic_adam": { "s1_dev_spearman": 0.6198, "stsb_dev_spearman_after": 0.736, "stsb_dev_pearson_after": 0.7087, "stsb_test_spearman_after": 0.6516, "stsb_test_pearson_after": 0.6564, "rocm_elapsed_sec": 243.28 }, "huygens_symplectic_adam": { "s1_dev_spearman": 0.6282, "stsb_dev_spearman_after": 0.7368, "stsb_dev_pearson_after": 0.709, "stsb_test_spearman_after": 0.6509, "stsb_test_pearson_after": 0.6548, "rocm_elapsed_sec": 236.41 }, "polar_natural_adam": { "stsb_dev_spearman_after": 0.7262, "stsb_dev_pearson_after": 0.702, "stsb_test_spearman_after": 0.6471, "stsb_test_pearson_after": 0.6494 }, "gyric_adam": { "s1_dev_spearman": 0.6221, "stsb_dev_spearman_after": 0.7304, "stsb_dev_pearson_after": 0.7023, "stsb_test_spearman_after": 0.6455, "stsb_test_pearson_after": 0.6503, "rocm_elapsed_sec": 419.85 }, "huygens_adam": { "s1_dev_spearman": 0.6383, "stsb_dev_spearman_after": 0.7295, "stsb_dev_pearson_after": 0.7035, "stsb_test_spearman_after": 0.642, "stsb_test_pearson_after": 0.6469, "rocm_elapsed_sec": 331.8 }, "hessian_pair_adam_1x_coupling": { "stsb_dev_spearman_after": 0.7242, "stsb_dev_pearson_after": 0.6876, "stsb_test_spearman_after": 0.6337, "stsb_test_pearson_after": 0.6369 }, "beltrami_adam": { "s1_dev_spearman": 0.6337, "stsb_dev_spearman_after": 0.7363, "stsb_dev_pearson_after": 0.7068, "stsb_test_spearman_after": 0.6333, "stsb_test_pearson_after": 0.6328, "rocm_elapsed_sec": 321.44 }, "hannay_adam": { "s1_dev_spearman": 0.6115, "stsb_dev_spearman_after": 0.7257, "stsb_dev_pearson_after": 0.699, "stsb_test_spearman_after": 0.6333, "stsb_test_pearson_after": 0.633, "rocm_elapsed_sec": 239.35 }, "rotor_adam": { "stsb_dev_spearman_after": 0.7322, "stsb_dev_pearson_after": 0.7099, "stsb_test_spearman_after": 0.6329, "stsb_test_pearson_after": 0.6309 }, "toroidal_lion": { "stsb_dev_spearman_after": 0.7289, "stsb_dev_pearson_after": 0.697, "stsb_test_spearman_after": 0.6319, "stsb_test_pearson_after": 0.6341 }, "adamw": { "stsb_dev_spearman_after": 0.7289, "stsb_test_spearman_after": 0.6317, "stsb_test_pearson_after": 0.6372 }, "poncelet_adam": { "s1_dev_spearman": 0.6262, "stsb_dev_spearman_after": 0.7275, "stsb_dev_pearson_after": 0.6987, "stsb_test_spearman_after": 0.6315, "stsb_test_pearson_after": 0.6374, "rocm_elapsed_sec": 267.34 }, "hessian_pair_adam_3x_coupling": { "stsb_dev_spearman_after": 0.7278, "stsb_dev_pearson_after": 0.6917, "stsb_test_spearman_after": 0.6288, "stsb_test_pearson_after": 0.6316 }, "cyclide_adam": { "s1_dev_spearman": 0.6255, "stsb_dev_spearman_after": 0.7368, "stsb_dev_pearson_after": 0.7179, "stsb_test_spearman_after": 0.6287, "stsb_test_pearson_after": 0.6476, "rocm_elapsed_sec": 380.42 }, "rindler_adam": { "s1_dev_spearman": 0.6208, "stsb_dev_spearman_after": 0.7318, "stsb_dev_pearson_after": 0.7069, "stsb_test_spearman_after": 0.6264, "stsb_test_pearson_after": 0.6353, "rocm_elapsed_sec": 271.68 }, "pair_adam_3x_coupling": { "stsb_dev_spearman_after": 0.7179, "stsb_test_spearman_after": 0.6242, "stsb_test_pearson_after": 0.633 }, "toroidal_muon": { "stsb_dev_spearman_after": 0.7361, "stsb_dev_pearson_after": 0.7105, "stsb_test_spearman_after": 0.6197, "stsb_test_pearson_after": 0.6304 } } }