{ "schema": "public_SCALE50_round3_completion_and_remaining_budget_v1", "as_of_beijing": "2026-11-02 16:10:21", "budget_anchor_utc": "2026-11-02T08:10:21.268705+00:00", "training": { "completed_main_rounds": 21, "total_main_rounds": 21, "all_main_training_complete": true, "auxiliary_additional_updates_still_planned": 3, "final_common_test_complete": false }, "completed_round": { "arm": "WDT_SCALE50", "round": 3, "status": "COMPLETE", "seed_questions": 50, "review_questions": 25, "new_questions": 9, "questions": 84, "answers": 672, "answers_per_question": 8, "effective_questions": 40, "effective_new_questions": 4, "all_succeeded_questions": 33, "all_failed_questions": 11, "new_main_learning_updates": 1, "final_model_saved": true, "repair_enabled": false, "new_question_breakdown": { "deep_down": { "all_failed": 0, "all_succeeded": 1, "effective_questions": 0, "minimum_actions_delta_range": [ -10, -10 ], "questions": 1, "required_chain_depth_delta_range": [ -1, -1 ] }, "deep_up": { "all_failed": 0, "all_succeeded": 0, "effective_questions": 2, "minimum_actions_delta_range": [ 1, 4 ], "questions": 2, "required_chain_depth_delta_range": [ 1, 1 ] }, "wide": { "all_failed": 1, "all_succeeded": 3, "effective_questions": 2, "minimum_actions_delta_range": [ -3, 1 ], "questions": 6, "required_chain_depth_delta_range": [ -1, 1 ] } }, "seed_observations": { "questions": 50, "previous_round": 2, "current_round": 3, "previous_successes": 279, "previous_answers": 400, "current_successes": 279, "current_answers": 400, "improved_questions": 12, "regressed_questions": 17, "unchanged_questions": 21, "recovered_from_all_failed": 0, "regressed_to_all_failed": 2 }, "common_old_task_observations": { "questions": 75, "previous_round": 2, "current_round": 3, "previous_successes": 402, "previous_answers": 600, "current_successes": 407, "current_answers": 600, "improved_questions": 20, "regressed_questions": 22, "unchanged_questions": 33, "recovered_from_all_failed": 1, "regressed_to_all_failed": 4 }, "observation_interpretation": "第二轮与第三轮作答的比较；第三轮答卷来自第二轮最终模型。第三轮训练后的最终表现由共同测试衡量。", "start_utc": "2026-11-02T06:22:35.695521+00:00", "end_utc": "2026-11-02T08:10:21.268705+00:00", "start_beijing": "2026-11-02 14:22:35", "end_beijing": "2026-11-02 16:10:21", "full_elapsed_seconds": 6465.573184, "stage_sum_seconds": 6465.546611000001, "interstage_seconds": 0.026573, "recorded_stage_seconds": { "plan-round": 4874.832532, "train-round": 1590.714079 }, "pure_parameter_training_seconds": null, "elapsed_scope": "整轮用时包含出题、作答、学习、保存及环节交接；后半段也不是纯模型调整时间。" }, "remaining_work": { "main_final_test_answers": 720, "auxiliary_final_test_answers": 180, "auxiliary_additional_updates": 3, "final_return_and_stop_included": true, "counts_are_at_budget_anchor": true }, "later_test_observation": { "as_of_beijing": "2026-11-02 16:11", "basis": "parent-session handoff", "arm": "N", "phase": "test_started", "main_test_answers_saved": 0, "used_to_measure_test_throughput": false, "deducted_from_remaining_budget": false }, "budget_summary": { "as_of_beijing": "2026-11-02 16:10:21", "central_remaining": "约 4 小时 20 分", "central_complete": "2026-11-02 20:30 左右", "planning_range": "约 2 小时 20 分–11 小时 20 分", "faster_complete": "2026-11-02 18:30 左右", "slower_complete": "2026-11-03 03:30 左右", "gpu_cost_central_usd": 8, "gpu_cost_range_usd": [ 4, 20 ], "scope": "720 份主要测试，小对照追加 3 次学习及 180 份测试，结果回收与结束实验。", "rate_note": "沿用每小时 1.79 美元的已有记录，只计后续显卡费用。" }, "scenarios": [ { "case": "faster", "remaining_seconds": 8095.524122810128, "remaining_hours": 2.248756700780591, "display_duration": "约 2 小时 20 分", "estimated_complete_beijing_calculated": "2026-11-02T18:25:16.792828+08:00", "estimated_complete_beijing_display": "2026-11-02 18:30 左右", "gpu_usd_at_recorded_rate": 4.025274494397258, "display_gpu_usd": 4, "components_seconds": { "eight_model_common30_test_720_answers": 5298.001070400001, "fixed_batch_additional_three_updates_and_initial_probabilities": 1173.0227848101267, "fixed_batch_common30_test_180_answers": 1324.5002676000001, "final_return_and_stop_allowance": 300 } }, { "case": "central", "remaining_seconds": 15405.801203733652, "remaining_hours": 4.279389223259348, "display_duration": "约 4 小时 20 分", "estimated_complete_beijing_calculated": "2026-11-02T20:27:07.069909+08:00", "estimated_complete_beijing_display": "2026-11-02 20:30 左右", "gpu_usd_at_recorded_rate": 7.660106709634233, "display_gpu_usd": 8, "components_seconds": { "eight_model_common30_test_720_answers": 10671.618178176795, "fixed_batch_additional_three_updates_and_initial_probabilities": 1466.2784810126584, "fixed_batch_common30_test_180_answers": 2667.904544544199, "final_return_and_stop_allowance": 600 } }, { "case": "slower", "remaining_seconds": 40456.01261716947, "remaining_hours": 11.237781282547076, "display_duration": "约 11 小时 20 分", "estimated_complete_beijing_calculated": "2026-11-03T03:24:37.281322+08:00", "estimated_complete_beijing_display": "2026-11-03 03:30 左右", "gpu_usd_at_recorded_rate": 20.115628495759264, "display_gpu_usd": 20, "components_seconds": { "eight_model_common30_test_720_answers": 28813.36908107735, "fixed_batch_additional_three_updates_and_initial_probabilities": 2639.3012658227854, "fixed_batch_common30_test_180_answers": 7203.342270269338, "final_return_and_stop_allowance": 1800 } } ], "budget_basis": "只扣除已完成的扩大题量组第三轮；测试和小对照保留原速度假设，尚未用开始但没有答卷的记录计算速度。", "cost": { "gpu_hourly_usd_recorded": 1.79, "rate_observed_utc": "2026-11-01T14:07:02.455461+00:00", "rate_rechecked": false, "teacher_charges_included": false, "storage_charges_included": false, "current_accumulated_bill_known": false }, "source_record": { "record_id": "SCALE50_R3_COMPLETED_20261102_CAPTURE01", "capture_sha256": "af87ede171f84c1fcba3e26909a05feecb355986ea23b12819c1cffe12dcb4b9" } }