Every trial of every study, from its trace. Green rows are valid outcomes (exact success for the pending-effect task).
| run | trials | valid | agent / controller |
|---|---|---|---|
| c1_qwen38_nothink_100 | 1500 | 891 | Qwen/Qwen3.8-27B |
| c2_qwen38_nothink_100 | 500 | 399 | Qwen/Qwen3.8-27B |
| c3_qwen38_nothink_100 | 400 | 199 | Qwen/Qwen3.8-27B |
| sweep | 84 | 70 | scripted_oracle |
| wi002_qwen38_nothink | 160 | 68 | Qwen/Qwen3.8-27B |
| wi003 | 95 | 50 | scripted_oracle |
| wi004_qwen38_nothink | 150 | 90 | Qwen/Qwen3.8-27B |
| wi004_qwen38_think | 40 | 12 | Qwen/Qwen3.8-27B |
| wi005_notification_p4 | 160 | 116 | Qwen/Qwen3.8-27B |
| wi005_plain_p6 | 160 | 130 | Qwen/Qwen3.8-27B |
| wi006_qwen38_nothink | 140 | 117 | Qwen/Qwen3.8-27B |
| wi007_qwen3_vl_8b_instruct_prl | 10 | 1 | Qwen/Qwen3-VL-8B-Instruct |
| wi007_qwen3_vl_8b_instruct_wi004 | 10 | 1 | Qwen/Qwen3-VL-8B-Instruct |
| wi007_qwen3_vl_8b_instruct_wi006 | 40 | 3 | Qwen/Qwen3-VL-8B-Instruct |
| wi007_ui_tars_1_5_7b_json_prl | 10 | 6 | ByteDance-Seed/UI-TARS-1.5-7B |
| wi007_ui_tars_1_5_7b_json_wi004 | 10 | 0 | ByteDance-Seed/UI-TARS-1.5-7B |
| wi007_ui_tars_1_5_7b_native64_prl | 10 | 0 | ByteDance-Seed/UI-TARS-1.5-7B |
| wi007_ui_tars_1_5_7b_native_prl | 10 | 3 | ByteDance-Seed/UI-TARS-1.5-7B |
| wi007_ui_tars_1_5_7b_native_wi004 | 50 | 29 | ByteDance-Seed/UI-TARS-1.5-7B |
| wi007_ui_tars_1_5_7b_native_wi006 | 40 | 17 | ByteDance-Seed/UI-TARS-1.5-7B |
| wi009_qwen38_nothink | 50 | 37 | Qwen/Qwen3.8-27B |