{"experimentId":"bo_paper_aligned_iteration_20261014","default":"luna-kissing_number_d11-root_sd3","entries":[{"id":"luna-circle_packing_rectangle-root_sd3","taskId":"circle_packing_rectangle","task":"矩形装圆","label":"root_sd3 · 行为目标冻结方案 vs AQ04","runIds":{"bo":"luna_002_circle_packing_rectangle_root_sd3","oe":"luna_001_circle_packing_rectangle_aq04"},"model":"Luna · medium","description":"左右按同一四小时计费时间对齐，查看处理方案与原版 AQ04 的父节点选择。"},{"id":"luna-circle_packing_rectangle-operation_corr0","taskId":"circle_packing_rectangle","task":"矩形装圆","label":"operation_corr0 · Luna kissing 最高分 vs AQ04","runIds":{"bo":"luna_021_circle_packing_rectangle_operation_corr0","oe":"luna_001_circle_packing_rectangle_aq04"},"model":"Luna · medium","description":"左右按同一四小时计费时间对齐，查看处理方案与原版 AQ04 的父节点选择。"},{"id":"luna-circle_packing_rectangle-noise075","taskId":"circle_packing_rectangle","task":"矩形装圆","label":"noise075 · Luna kissing 第二高分 vs AQ04","runIds":{"bo":"luna_005_circle_packing_rectangle_noise075","oe":"luna_001_circle_packing_rectangle_aq04"},"model":"Luna · medium","description":"左右按同一四小时计费时间对齐，查看处理方案与原版 AQ04 的父节点选择。"},{"id":"muse-circle_packing_rectangle-root_sd3","taskId":"circle_packing_rectangle","task":"矩形装圆","label":"root_sd3 · Muse 独立验证","runIds":{"bo":"muse_025_circle_packing_rectangle_root_sd3"},"model":"Muse Spark 1.3 · high","description":"冻结方案的单次 Muse 轨迹。两题未保留完整目标行为；没有新增同模型对照。"},{"id":"luna-kissing_number_d11-root_sd3","taskId":"kissing_number_d11","task":"Kissing number · D11","label":"root_sd3 · 行为目标冻结方案 vs AQ04","runIds":{"bo":"luna_008_kissing_number_d11_root_sd3","oe":"luna_007_kissing_number_d11_aq04"},"model":"Luna · medium","description":"左右按同一四小时计费时间对齐，查看处理方案与原版 AQ04 的父节点选择。"},{"id":"luna-kissing_number_d11-operation_corr0","taskId":"kissing_number_d11","task":"Kissing number · D11","label":"operation_corr0 · Luna kissing 最高分 vs AQ04","runIds":{"bo":"luna_023_kissing_number_d11_operation_corr0","oe":"luna_007_kissing_number_d11_aq04"},"model":"Luna · medium","description":"左右按同一四小时计费时间对齐，查看处理方案与原版 AQ04 的父节点选择。"},{"id":"luna-kissing_number_d11-noise075","taskId":"kissing_number_d11","task":"Kissing number · D11","label":"noise075 · Luna kissing 第二高分 vs AQ04","runIds":{"bo":"luna_011_kissing_number_d11_noise075","oe":"luna_007_kissing_number_d11_aq04"},"model":"Luna · medium","description":"左右按同一四小时计费时间对齐，查看处理方案与原版 AQ04 的父节点选择。"},{"id":"muse-kissing_number_d11-root_sd3","taskId":"kissing_number_d11","task":"Kissing number · D11","label":"root_sd3 · Muse 独立验证","runIds":{"bo":"muse_026_kissing_number_d11_root_sd3"},"model":"Muse Spark 1.3 · high","description":"冻结方案的单次 Muse 轨迹。两题未保留完整目标行为；没有新增同模型对照。"}],"notes":["每个设置×任务仅一次，seed=42；图形是已完成实验的历史回放。","Luna 三个方案均与同模型原版 AQ04 对照；Muse 只有冻结方案的独立验证，没有新增同模型匹配对照。","root_sd3 是行为目标方案；operation_corr0 / noise075 因 Luna kissing 最终分数居前而展示，不代表矩形任务排名或统计显著。","时间使用原计费搜索时钟：包含共同初始化与恢复累计；暂停不增加预算，结果只在完成事件后出现。","采样效用来自在线实际的一次 joint TS；获选概率来自已封存后验的离线估计，未重新运行实验。","轨道是五个初始草稿谱系；融合可引用其他分支，谱系不同不等于完全独立路线。","Luna noise075 kissing 保留路径故障及同身份恢复；预算后封存的运行保留原 FAILED 状态。","已评估数只计预算内产生节点的评估；生成失败单列，不能等同于有效反馈。截止未返回事件保留，不展示预算外分数。","最优曲线保留首个有效基线及改善；谱系图保留退步与无效节点，用于查看真实选择。","Bucket 仅包含选定公开派生数据；完整原始日志与规划原文未上传。"],"bucketUrl":"https://huggingface.co/buckets/jzshared/autoresearch-runs/tree/experiments/bo_paper_aligned_iteration_20261014/report","budget":14400}
