Example trial
GPT 5.6 Sol
dc-legacy-alloc-full-stack
- Reward
- 78.0%
- Checks passed
- 6 / 8
- Duration
- 9 min
Scoring
| Check | Credit | Weight |
|---|---|---|
| contracts and scenario data are unchanged | 100.0% | 2% |
| allocator runs and returns a plan | 100.0% | 8% |
| plan is feasible under the contract | 100.0% | 15% |
| locked phases come back verbatim | 100.0% | 14% |
| fill matches the golden optimum | 100.0% | 25% |
| lateness matches the golden optimum | 0.0% | 12% |
| cost matches the golden optimum | 0.0% | 10% |
| re-run tracks landed lots and new locks | 100.0% | 14% |
The official reward includes any verifier gates that override individual checks.