SDL-2 sealed leaderboard: mined results
Board worlds per arm: [32].
Justified: Supported at K100. Material outcome: the nominated exact route meets the requirement in the world, within the allowance, on the feasible worlds. Evidence rule: the v1.2 scripted rule, a diagnostic, never Justified.
| Rank | Agent | Both (95% CI) | Material outcome, feasible worlds (95% CI) | Justified (95% CI) | Outcome, all worlds (95% CI) | Evidence rule | Nominate / restrict / none | Wrong restrictions | Median cost |
|---|---|---|---|---|---|---|---|---|---|
| 1 | GPT-6 Astra | 16/32 = 0.5 [0.3189, 0.6811] | 8/16 = 0.5 [0.2465, 0.7535] | 18/32 = 0.5625 [0.3766, 0.7364] | 23/32 = 0.7188 [0.5325, 0.8625] | 0.7188 | 11 / 21 / 0 | 6/16 | $8.02 |
| 2 | Sonnet 5.5 | 15/32 = 0.4688 [0.2909, 0.6526] | 9/16 = 0.5625 [0.2988, 0.8025] | 16/32 = 0.5 [0.3189, 0.6811] | 20/32 = 0.625 [0.4369, 0.789] | 0.5 | 20 / 12 / 0 | 1/16 | $1.52 |
| 3 | Gemini 4 Argon | 15/32 = 0.4688 [0.2909, 0.6526] | 10/16 = 0.625 [0.3543, 0.848] | 19/32 = 0.5938 [0.4064, 0.763] | 22/32 = 0.6875 [0.4999, 0.8388] | 0.7812 | 16 / 16 / 0 | 4/16 | $12.48 |
| 4 | Opus 5.5 | 14/32 = 0.4375 [0.2636, 0.6234] | 9/16 = 0.5625 [0.2988, 0.8025] | 17/32 = 0.5312 [0.3474, 0.7091] | 25/32 = 0.7812 [0.6003, 0.9072] | 0.6562 | 10 / 22 / 0 | 6/16 | $22.46 |
| 5 | DeepSeek 1017 (base) | 9/32 = 0.2812 [0.1375, 0.4675] | 1/16 = 0.0625 [0.0016, 0.3023] | 12/32 = 0.375 [0.211, 0.5631] | 16/32 = 0.5 [0.3189, 0.6811] | 0.5625 | 3 / 29 / 0 | 14/16 | $0.1 |
| 6 | BO | 9/32 = 0.2812 [0.1375, 0.4675] | 4/16 = 0.25 [0.0727, 0.5238] | 16/32 = 0.5 [0.3189, 0.6811] | 15/32 = 0.4688 [0.2909, 0.6526] | 0.5625 | 14 / 18 / 0 | 7/16 | $0.0 |
| 7 | DOE | 9/32 = 0.2812 [0.1375, 0.4675] | 8/16 = 0.5 [0.2465, 0.7535] | 10/32 = 0.3125 [0.1612, 0.5001] | 24/32 = 0.75 [0.566, 0.8854] | 0.5312 | 12 / 20 / 0 | 4/16 | $0.0 |
| 8 | z-ai/glm-5.3-flash | 8/32 = 0.25 [0.1146, 0.434] | 1/16 = 0.0625 [0.0016, 0.3023] | 13/32 = 0.4062 [0.237, 0.5936] | 16/32 = 0.5 [0.3189, 0.6811] | 0.5625 | 6 / 26 / 0 | 11/16 | $0.54 |
Behaviour (medians unless a share)
| Behaviour | Opus 5.5 | Sonnet 5.5 | GPT-6 Astra | Gemini 4 Argon | DeepSeek 1017 (base) | BO | DOE | z-ai/glm-5.3-flash |
|---|---|---|---|---|---|---|---|---|
| routes_built | 7.0 | 13.0 | 13.5 | 10.5 | 6.0 | 17.5 | 19.0 | 8.5 |
| loops | 4.5 | 5.0 | 4.5 | 5.0 | 6.5 | 3.0 | 3.0 | 4.5 |
| levers_varied | 6.0 | 5.0 | 9.0 | 6.5 | 4.0 | 9.0 | 7.0 | 4.0 |
| one_factor_pairs | 3.5 | 8.0 | 4.0 | 6.0 | 8.0 | 2.0 | 3.5 | 5.0 |
| single_lever_sweep_share | 0.219 | 0.438 | 0.406 | 0.469 | 0.344 | 0.469 | 0.688 | 0.156 |
| anneal_programs | 1.0 | 2.0 | 2.0 | 1.0 | 1.0 | 4.0 | 4.0 | 2.0 |
| replicated_nomination_share | 0.8 | 0.55 | 0.636 | 0.812 | 0.667 | 0.357 | 0.417 | 0.667 |
| experiments_accepted | 7.0 | 10.0 | 9.0 | 7.5 | 6.0 | 9.0 | 9.0 | 7.0 |
| experiments_rejected | 0.03 | 0.0 | 0.12 | 0.06 | 0.84 | 0.0 | 0.0 | 1.91 |
| experiments_failed | 0.0 | 0.0 | 0.0 | 0.0 | 0.22 | 1.0 | 0.0 | 0.25 |
| credits | 503.5 | 524.5 | 569.0 | 561.0 | 556.0 | 443.0 | 544.0 | 535.5 |
| bodies | 7.0 | 7.0 | 9.0 | 8.0 | 7.0 | 9.0 | 11.0 | 6.0 |
| minutes_to_first_experiment | 98.7 | 30.05 | 1.8 | 53.35 | 59.8 | None | None | 65.6 |
| models_adopted | 1.0 | 1.0 | 2.0 | 2.0 | 1.0 | 0.0 | 0.0 | 3.0 |
| model_revisions_after_first_return | 0.0 | 0.0 | 1.0 | 1.0 | 0.0 | 0.0 | 0.0 | 1.0 |
| previews | 1.0 | 0.5 | 0.0 | 4.0 | 2.0 | 0.0 | 0.0 | 0.0 |
| forecast_numeric_share | 0.993 | 0.956 | 1.0 | 1.0 | 0.777 | 0.0 | 0.0 | 0.717 |
| bash_minutes | 164.95 | 34.05 | 7.95 | 21.15 | 38.3 | 0.0 | 0.0 | 42.95 |
| background_job_share | 1.0 | 0.938 | 0.062 | 0.312 | 0.719 | 0.0 | 0.0 | 0.875 |
| supplied_package_use_share | 1.0 | 1.0 | 0.906 | 1.0 | 1.0 | 0.0 | 0.0 | 1.0 |
| commands_stopped_by_bound | 4 | 3 | 3 | 6 | 15 | 0 | 0 | 5 |
| commands_stopped_by_own_timeout | 0 | 0 | 1 | 0 | 0 | 0 | 0 | 0 |
| claim_file_outside_workspace_root | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 |
| commands_killed_by_memory | 1 | 0 | 0 | 3 | 8 | 0 | 0 | 2 |
| compaction_share | 0.031 | 0.0 | 0.062 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 |