Datasets:
model_id string | split string | window_id string | policy_name string | ranking_score float64 | probability_epsilon_optimal_proxy float64 | predicted_regret float64 | true_reward float64 | true_regret float64 | selected int64 |
|---|---|---|---|---|---|---|---|---|---|
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0016 | fifo | 0.000002 | 0.004838 | 0.996871 | 0.074405 | 0.279762 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0016 | edf | 0.000003 | 0.004838 | 0.99687 | 0.077381 | 0.276786 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0016 | shortest_output_first | 0.001461 | 0.004875 | 0.995412 | 0.178571 | 0.175595 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0016 | shortest_prompt_first | 0.000139 | 0.004841 | 0.996734 | 0.25 | 0.104167 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0016 | greedy_token_fill | 0 | 0.004838 | 0.996873 | 0.074405 | 0.279762 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0016 | least_loaded | 0 | 0.004838 | 0.996873 | 0.074405 | 0.279762 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0016 | multi_bin_batching | 0.000053 | 0.004839 | 0.99682 | 0.083333 | 0.270833 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0016 | random_feasible | 0.000002 | 0.004838 | 0.996871 | 0.157738 | 0.196429 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0016 | first_fit | 0 | 0.004838 | 0.996873 | 0.074405 | 0.279762 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0016 | best_fit | 0 | 0.004838 | 0.996873 | 0.074405 | 0.279762 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0016 | orca_style | 0.000008 | 0.004838 | 0.996865 | 0.214286 | 0.139881 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0016 | vllm_style_token_budget | 0.000041 | 0.004839 | 0.996832 | 0.178571 | 0.175595 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0016 | sarathi_style | 0.000012 | 0.004838 | 0.99686 | 0.125 | 0.229167 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0016 | splitfuse_style | 0 | 0.004838 | 0.996873 | 0.130952 | 0.223214 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0016 | slo_slack_score | 0 | 0.004838 | 0.996873 | 0.214286 | 0.139881 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0016 | weighted_shortest_processing | 0.000028 | 0.004839 | 0.996845 | 0.330357 | 0.02381 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0016 | least_laxity_first | 0.000034 | 0.004839 | 0.996839 | 0.160714 | 0.193452 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0016 | estimated_service_time_first | 0.000068 | 0.00484 | 0.996805 | 0.282738 | 0.071429 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0016 | admission_control | 0.000001 | 0.004838 | 0.996872 | 0.116071 | 0.238095 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0016 | scorpio_style_slo_guard | 0.000029 | 0.004839 | 0.996844 | 0.008929 | 0.345238 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0016 | sola_style_state_aware | 0.996873 | 0.874138 | 0 | 0.354167 | 0 | 1 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0016 | slai_style_phase_aware | 0.000757 | 0.004857 | 0.996116 | 0.354167 | 0 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0016 | flow_control_stability | 0 | 0.004838 | 0.996873 | 0.116071 | 0.238095 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0016 | kv_constrained_online | 0.000366 | 0.004847 | 0.996507 | 0.309524 | 0.044643 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0016 | adaptive_chunked_prefill | 0 | 0.004838 | 0.996873 | 0.092262 | 0.261905 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0016 | aging_priority | 0.000004 | 0.004838 | 0.996869 | 0.327381 | 0.026786 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0016 | weighted_fair_share | 0.000118 | 0.004841 | 0.996754 | 0.28869 | 0.065476 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0017 | fifo | 0.000001 | 0.004839 | 0.997558 | 0.394813 | 0.216138 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0017 | edf | 0.000003 | 0.004839 | 0.997557 | 0.412104 | 0.198847 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0017 | shortest_output_first | 0.997559 | 0.874138 | 0 | 0.610951 | 0 | 1 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0017 | shortest_prompt_first | 0.000058 | 0.00484 | 0.997501 | 0.556196 | 0.054755 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0017 | greedy_token_fill | 0 | 0.004838 | 0.997559 | 0.394813 | 0.216138 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0017 | least_loaded | 0 | 0.004838 | 0.997559 | 0.394813 | 0.216138 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0017 | multi_bin_batching | 0.000112 | 0.004841 | 0.997447 | 0.507205 | 0.103746 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0017 | random_feasible | 0.000001 | 0.004839 | 0.997558 | 0.458213 | 0.152738 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0017 | first_fit | 0 | 0.004838 | 0.997559 | 0.394813 | 0.216138 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0017 | best_fit | 0 | 0.004838 | 0.997559 | 0.394813 | 0.216138 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0017 | orca_style | 0.000003 | 0.004839 | 0.997556 | 0.435159 | 0.175793 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0017 | vllm_style_token_budget | 0.000051 | 0.00484 | 0.997509 | 0.610951 | 0 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0017 | sarathi_style | 0.000007 | 0.004839 | 0.997553 | 0.391931 | 0.21902 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0017 | splitfuse_style | 0 | 0.004838 | 0.997559 | 0.391931 | 0.21902 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0017 | slo_slack_score | 0 | 0.004838 | 0.997559 | 0.435159 | 0.175793 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0017 | weighted_shortest_processing | 0.000034 | 0.004839 | 0.997525 | 0.567723 | 0.043228 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0017 | least_laxity_first | 0.000008 | 0.004839 | 0.997551 | 0.412104 | 0.198847 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0017 | estimated_service_time_first | 0.000028 | 0.004839 | 0.997532 | 0.556196 | 0.054755 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0017 | admission_control | 0.000006 | 0.004839 | 0.997554 | 0.414986 | 0.195965 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0017 | scorpio_style_slo_guard | 0.000031 | 0.004839 | 0.997528 | 0.002882 | 0.608069 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0017 | sola_style_state_aware | 0.000545 | 0.004852 | 0.997014 | 0.579251 | 0.0317 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0017 | slai_style_phase_aware | 0.000058 | 0.00484 | 0.997502 | 0.56196 | 0.048991 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0017 | flow_control_stability | 0 | 0.004838 | 0.997559 | 0.414986 | 0.195965 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0017 | kv_constrained_online | 0.000516 | 0.004851 | 0.997044 | 0.567723 | 0.043228 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0017 | adaptive_chunked_prefill | 0 | 0.004838 | 0.997559 | 0.414986 | 0.195965 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0017 | aging_priority | 0.000856 | 0.00486 | 0.996703 | 0.567723 | 0.043228 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0017 | weighted_fair_share | 0.000122 | 0.004842 | 0.997437 | 0.590778 | 0.020173 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0018 | fifo | 0.000001 | 0.004838 | 0.997037 | 0.643713 | 0.086826 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0018 | edf | 0.000001 | 0.004838 | 0.997036 | 0.616766 | 0.113772 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0018 | shortest_output_first | 0.000017 | 0.004838 | 0.997021 | 0.649701 | 0.080838 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0018 | shortest_prompt_first | 0.000003 | 0.004838 | 0.997035 | 0.673653 | 0.056886 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0018 | greedy_token_fill | 0 | 0.004838 | 0.997038 | 0.643713 | 0.086826 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0018 | least_loaded | 0 | 0.004838 | 0.997038 | 0.643713 | 0.086826 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0018 | multi_bin_batching | 0.000342 | 0.004847 | 0.996696 | 0.649701 | 0.080838 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0018 | random_feasible | 0.000003 | 0.004838 | 0.997035 | 0.676647 | 0.053892 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0018 | first_fit | 0 | 0.004838 | 0.997038 | 0.643713 | 0.086826 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0018 | best_fit | 0 | 0.004838 | 0.997038 | 0.643713 | 0.086826 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0018 | orca_style | 0.000006 | 0.004838 | 0.997031 | 0.652695 | 0.077844 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0018 | vllm_style_token_budget | 0.000113 | 0.004841 | 0.996925 | 0.649701 | 0.080838 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0018 | sarathi_style | 0.000001 | 0.004838 | 0.997036 | 0.601796 | 0.128743 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0018 | splitfuse_style | 0 | 0.004838 | 0.997038 | 0.601796 | 0.128743 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0018 | slo_slack_score | 0 | 0.004838 | 0.997038 | 0.652695 | 0.077844 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0018 | weighted_shortest_processing | 0.000841 | 0.004859 | 0.996196 | 0.706587 | 0.023952 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0018 | least_laxity_first | 0.000008 | 0.004838 | 0.99703 | 0.637725 | 0.092814 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0018 | estimated_service_time_first | 0.000021 | 0.004839 | 0.997016 | 0.670659 | 0.05988 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0018 | admission_control | 0.000011 | 0.004838 | 0.997027 | 0.649701 | 0.080838 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0018 | scorpio_style_slo_guard | 0.000023 | 0.004839 | 0.997015 | 0.008982 | 0.721557 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0018 | sola_style_state_aware | 0.000256 | 0.004844 | 0.996782 | 0.727545 | 0.002994 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0018 | slai_style_phase_aware | 0.001211 | 0.004869 | 0.995827 | 0.718563 | 0.011976 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0018 | flow_control_stability | 0 | 0.004838 | 0.997038 | 0.649701 | 0.080838 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0018 | kv_constrained_online | 0.000034 | 0.004839 | 0.997004 | 0.706587 | 0.023952 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0018 | adaptive_chunked_prefill | 0 | 0.004838 | 0.997038 | 0.643713 | 0.086826 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0018 | aging_priority | 0.000071 | 0.00484 | 0.996967 | 0.706587 | 0.023952 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0018 | weighted_fair_share | 0.997038 | 0.874138 | 0 | 0.730539 | 0 | 1 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0019 | fifo | 0 | 0.00484 | 0.998865 | 0.340426 | 0.258359 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0019 | edf | 0.000023 | 0.00484 | 0.998843 | 0.361702 | 0.237082 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0019 | shortest_output_first | 0.000569 | 0.004854 | 0.998297 | 0.495441 | 0.103343 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0019 | shortest_prompt_first | 0.000003 | 0.00484 | 0.998863 | 0.480243 | 0.118541 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0019 | greedy_token_fill | 0 | 0.00484 | 0.998865 | 0.340426 | 0.258359 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0019 | least_loaded | 0 | 0.00484 | 0.998865 | 0.340426 | 0.258359 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0019 | multi_bin_batching | 0.00001 | 0.00484 | 0.998856 | 0.325228 | 0.273556 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0019 | random_feasible | 0 | 0.00484 | 0.998865 | 0.486322 | 0.112462 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0019 | first_fit | 0 | 0.00484 | 0.998865 | 0.340426 | 0.258359 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0019 | best_fit | 0 | 0.00484 | 0.998865 | 0.340426 | 0.258359 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0019 | orca_style | 0.000037 | 0.004841 | 0.998829 | 0.37386 | 0.224924 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0019 | vllm_style_token_budget | 0.000013 | 0.00484 | 0.998853 | 0.513678 | 0.085106 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0019 | sarathi_style | 0.000026 | 0.00484 | 0.998839 | 0.258359 | 0.340426 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0019 | splitfuse_style | 0 | 0.00484 | 0.998865 | 0.243161 | 0.355623 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0019 | slo_slack_score | 0 | 0.00484 | 0.998865 | 0.37386 | 0.224924 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0019 | weighted_shortest_processing | 0.000005 | 0.00484 | 0.998861 | 0.568389 | 0.030395 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0019 | least_laxity_first | 0.000026 | 0.00484 | 0.998839 | 0.410334 | 0.18845 | 0 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0019 | estimated_service_time_first | 0.998865 | 0.874138 | 0 | 0.598784 | 0 | 1 |
finalist_direct_multiclass_classification_hist_gradient_boosting_s101 | VALIDATION | azure_2023_code__representative__b1__w0019 | admission_control | 0 | 0.00484 | 0.998865 | 0.340426 | 0.258359 | 0 |
LLM-Serving Selector Regret
LLM-Serving Selector Regret is a metrics-only research dataset for studying learned policy selection in LLM-serving schedulers. It contains derived selector/oracle/regret objects generated by Soroush Vahidi's research workflow, not raw request traces.
Creator / Provider
Dataset creator/provider: Soroush Vahidi.
The released selector/regret and policy-suitability metrics were generated by Soroush Vahidi's research workflow. Underlying third-party workload traces remain subject to their original providers and licenses; this dataset does not claim ownership of those upstream traces.
Dataset Structure
This standalone dataset has two configs:
| Config | Rows | Columns | Row meaning |
|---|---|---|---|
selector_regret |
3,852 | 160 | One workload/window with context features, 27 candidate-policy reward values, oracle fields, split metadata, and regret-related decision fields. |
policy_suitability |
95,364 | 10 | One (model_id, split, window_id, policy_name) record for the validation-selected finalist selector. |
selector_v2_registry |
640 | 51 | One (window, policy) pair with 48 numerical simulation outcome metrics, 49 window features, and split metadata — training-oriented selector data for the 8-policy Option B scope. |
This dataset is intentionally separate from SoroushVahidi/llm-serving-scheduler-baselines. That dataset contains fixed scheduler outcome rows. This dataset contains selector/oracle/regret and suitability objects that require additional intermediate features and learned-selector outputs and are not reconstructible from the scheduler-baselines release alone.
selector_regret Semantics
A selector_regret row represents one workload/window in a 27-policy full-information selector benchmark. Column groups are:
- workload/window identifiers:
window_id, source fields, temporal fields, and split fields; - context and workload features: columns prefixed with
feat_plus workload parameter columns; - policy reward vector: one numeric reward column for each candidate policy, including
fifo,edf,weighted_shortest_processing,scorpio_style_slo_guard,kv_constrained_online, and the rest of the 27-policy V2 library; - oracle decision fields:
oracle_v2_anwg,best_policy,second_best_margin, andmeaningful_margin; - V1 comparison fields:
oracle_v1_anwgandv2_minus_v1_gain; - epsilon-optimality fields:
epsilon_optimal_count_0.001,epsilon_optimal_count_0.005, andepsilon_optimal_count_0.01.
The objective is arrival_normalized_weighted_goodput; higher is better. For a window w and policy p, with reward R(w,p), the V2 oracle reward is:
oracle_v2_anwg(w) = max_p R(w,p).
The per-policy true regret used by the suitability table is absolute, non-normalized reward loss:
true_regret(w,p) = oracle_v2_anwg(w) - R(w,p).
This value is non-negative up to numerical tolerance. Ties are handled by the underlying tabular maximum/argmax behavior used during generation; epsilon-optimality counts expose near-tie structure at thresholds 0.001, 0.005, and 0.01. The release validation confirms every benchmark window has all 27 policy reward values available.
policy_suitability Semantics
A policy_suitability row represents one candidate policy evaluated for one non-TRAIN workload/window by the validation-selected finalist selector model.
Columns:
model_id: selected finalist selector model identifier.split: evaluation split.window_id: workload/window key that joins toselector_regret.window_id.policy_name: candidate policy being scored.ranking_score: model score used to rank policies for the window.probability_epsilon_optimal_proxy: softmax-style proxy over policy scores for epsilon-optimality inspection; it is a diagnostic score, not a calibrated probability guarantee.predicted_regret: non-negative model-implied regret proxy relative to the highest model score in that window.true_reward: realizedarrival_normalized_weighted_goodputfor this policy/window.true_regret:oracle_v2_anwg - true_rewardfrom the selector-regret table.selected: 1 if the finalist selector selected this policy for the window, else 0.
policy_suitability intentionally excludes TRAIN windows. It covers 3,532 non-TRAIN windows x 27 policies = 95,364 rows.
selector_v2_registry Semantics
selector_v2_registry is the selector-v2 calibration pilot — a curated, training-oriented subset derived from the same simulation infrastructure and trace families as selector_regret, but structured differently for training a learned policy selector model.
Structural relationship to selector_regret: Both share the primary objective (arrival_normalized_weighted_goodput) and the same upstream workload families (BurstGPT, AzureLLM 2023). However:
selector_regretuses a wide format (1 row per window, 27 policy reward columns) for benchmarking.selector_v2_registryuses a long format (1 row per window×policy pair) for training.
They are related but non-redundant. Registry rows are not reconstructible from selector_regret (different window sets, different policy subsets), and selector_regret rows are not reconstructible from the registry.
selector_v2_registry row definition
A selector_v2_registry row represents one simulation outcome when a single scheduling policy is applied to a single workload window.
- Window: A contiguous time slice of a workload trace (synthetic or real-trace), defined by an arrival range and row index in the source JSONL.
- Policy: One of 8 non-faithful scheduling policies (Option B scope).
- Outcome: 48 numerical metrics computed by the simulation engine.
Columns
Identifiers
window_idx— Sequential window identifier (0-based)group_key— Grouping key for split-atomicitysplit— TRAIN / VALIDATION / ID_TEST / OOD_TESTsplit_group_key— Split-group identifier for leakage-safe groupingpolicy_name— Scheduling policy evaluated
Window features (49 feat_ columns)* Context features extracted from each window: arrival rates, burstiness coefficients, queue lengths, priority distributions, resource configuration (GPU count, KV capacity, sequence capacity, token budget), saturation load estimates, tightness fractions, slack metrics, and disaggregated/multi-instance topology features where applicable.
Simulation outcome metrics (48 metric_ columns)*
Numerical outcomes from policy evaluation: arrival_normalized_weighted_goodput, latencies (mean, median, p50, p95, p99), throughput (request, token, SLO success), SLO attainment/violation rates, queue lengths, GPU utilization, admission/rejection rates, and event counts.
Some metrics are null for certain workload types:
- Disaggregated topology metrics (prefill/decode queue utilization, bridge queue metrics) are null for monolithic workloads.
- Some latency/throughput/SLO metrics are null for a small subset of synthetic workloads (up to 4.1% of rows).
Labels / targets
primary_objective_classification— Discernibility classification per window: STRONGLY_DISCRIMINATIVE, NEAR_TIE, or ALL_COMPLETE_OR_EFFECTIVELY_TIEDprimary_objective_best_policy— The best-performing policy of the 8 for this window (by ANWG)primary_objective_max_min_spread— ANWG gap between best and worst of the 8 policies
Policy set (Option B scope)
fifoedfscorpio_style_slo_guardadmission_controlweighted_shortest_processingestimated_service_time_firstbest_fitmulti_bin_batching
The following are explicitly excluded from this config: vllm_faithful, vllm_chunked_prefill_faithful, sarathi_faithful, distserve_faithful, tetriinfer_paper_reimplementation, llumnix_faithful.
Calibration
SLO multiplier = 2.0, policy-independent reference model. Trace-level OOD reservation: last 15% of rows (by arrival time) in each source trace are reserved exclusively for OOD_TEST, disjoint from historical pool by row-index construction.
Intended use
This config is intended for training learned selector models to choose which of the 8 policies to apply per workload window, given window-level features. It complements selector_regret (which benchmarks selector models across 30 policies) and policy_suitability (which evaluates a finalist selector against non-TRAIN windows).
Split distribution
| Split | Windows | Policies/Window | Rows |
|---|---|---|---|
| TRAIN | 31 | 8 | 248 |
| VALIDATION | 18 | 8 | 144 |
| ID_TEST | 13 | 8 | 104 |
| OOD_TEST | 18 | 8 | 144 |
| Total | 80 | 8 | 640 |
Known issue: 19 cross-split row-range overlap pairs across 27 distinct windows in TRAIN/VALIDATION/ID_TEST pools. OOD_TEST is structurally clean (reserved by row index).
Provenance
Generated from llm-serving-heuristic-evolution at commit b698e49df7954ad6fb8e3026d27b091e971248bd ("Reconcile Selector v2 split-leakage fix with integration branch's c8aee12").
Source experiment: selector_v2_calibrated_pilot_20260720T163235Z. Validation manifests passed for dataset construction, quality gates (7/7), and leakage audit.
Provenance
Generated from llm-serving-heuristic-evolution at commit e8bd759b6cdaa8a05096b0ceeb1c7684cfa07302.
The source experiment was v2_selector_regret_benchmark_20260722T134925Z. Validation manifests passed for dataset construction, leakage audit, model combination, finalist combination, and final report generation. The selected finalist was finalist_direct_multiclass_classification_hist_gradient_boosting_s101.
See metadata/provenance.json and metadata/validation_report.json for machine-readable provenance and validation checks.
Upstream Workload/Trace Attribution
This release contains Soroush-generated derived metrics, features, selector outputs, oracle fields, and regret/suitability records. It does not redistribute raw upstream trace rows, raw prompts, completions, provider payloads, or operational logs.
Rows may be derived from or identify these upstream workload families:
- BurstGPT: public workload trace repository, CC BY 4.0. Source: https://github.com/HPMLL/BurstGPT. If you use rows derived from BurstGPT source families, also cite the BurstGPT dataset/paper as requested by its maintainers.
- Azure LLM Inference Trace 2023: public Azure trace sample, CC BY Attribution License. Source: https://github.com/Azure/AzurePublicDataset/blob/master/AzureLLMInferenceDataset2023.md. If you use rows derived from
azure_2023_*source families, also cite the Azure trace and the Splitwise paper as requested by the dataset provider. - Synthetic policy frontier: generated by Soroush Vahidi's research workflow and covered by this dataset's license for the released metrics.
This dataset does not claim ownership of upstream trace data. No statement in this dataset should be read as claiming ownership of upstream BurstGPT or Azure trace data.
License
License for the released Soroush-generated metrics and derived selector outputs: Creative Commons Attribution 4.0 International (CC BY 4.0).
This license requires attribution. It applies to this dataset's released derived metrics and metadata. It does not replace the licenses or attribution requirements of upstream workload/trace sources.
Attribution
If you use, redistribute, adapt, benchmark with, or build derived artifacts from this dataset, please attribute the dataset to Soroush Vahidi and cite the dataset using the citation below. Upstream workload/trace sources should also be cited as specified in the provenance section.
Citation
@dataset{vahidi_llm_serving_selector_regret_2026,
author = {Soroush Vahidi},
title = {LLM-Serving Selector Regret},
year = {2026},
publisher = {Hugging Face},
version = {1.0},
url = {https://huggingface.co/datasets/SoroushVahidi/llm-serving-selector-regret}
}
Limitations
This is a decision-quality/regret benchmark, not a claim that learned top-1 policy selection is solved. Project documentation records that the selected model produces useful suitability/ranking signals but does not fully capture held-out OOD V1-to-V2 oracle gain. Users should analyze split-specific regret and OOD behavior rather than relying only on aggregate averages.
- Downloads last month
- 49