Reproducing 21 paper-linked candidates
Author code, accessible inputs, measured outputs, and comparisons from the candidate reproduction campaign. Cole / @ajhcs directed the work; domain agents performed the runs and other agents reviewed the evidence. Human scientific review remains pending.
A completed code run and a match to a published target are reported separately. Some paper-linked artifacts differ from the article or no longer supply the required runtime, trained model, data, or reference output. These differences and failed attempts remain part of the record.
| Recorded outcome | Tools |
|---|---|
| Selected paper target compared | 5 |
| Author code executed | 6 |
| Adapted run executed | 6 |
| Saved outputs analysed; simulation not rerun | 1 |
| Attempted; blocked | 3 |
Pass/fail in the table applies only to the comparison defined in its guide, including exploratory or archived-output checks. A passing selected paper target additionally needs the corresponding execution classification.
The preCICE and deal.II candidates share one coupled experiment. Counts describe tools; they do not imply independent experiments or reproduction of every result in a paper.
| Candidate | Execution | Recorded comparison (scope in guide) |
|---|---|---|
| Generate a published MOVi-A dataset scene with Kubric and Blender | Author code executed | Passed within stated rule |
| Gazebo | Author code executed | Did not pass |
| Bullet / PyBullet | Adapted run executed | Did not pass |
| CARLA | Adapted run executed | Did not pass |
| preCICE | Attempted; blocked | Not established |
| FEniCSx / DOLFINx | Selected paper target compared | Passed within stated rule |
| deal.II | Attempted; blocked | Not established |
| AutoCkt public op-amp: author-code training and held-out evaluation | Adapted run executed | Did not pass |
| DualSPHysics | Adapted run executed | Not established |
| ORE central-clearing study: fresh simulation blocked by missing author generator | Attempted; blocked | Not established |
| Carroll author-associated HARK replication: executed with failed strict table comparison | Author code executed | Did not pass |
| TaxFuncIntegr: fresh macro steady states from public fitted inputs | Author code executed | Passed within stated rule |
| The published M/M/1 queue and its replication interval | Selected paper target compared | Passed within stated rule |
| Published ring-network switching states and full-day comparison | Selected paper target compared | Did not pass |
| Archived energy-system metric comparison and verified optimizer blocker | Saved outputs analysed; simulation not rerun | Passed within stated rule |
| Historical LuST city-model run with three seeds | Adapted run executed | Not established |
| Tellurium Calzone archive: selected Figure4B Wee1c comparison | Selected paper target compared | Passed within stated rule |
| Historical Brian eye-movement author model with exploratory Figure3 comparison | Author code executed | Passed within stated rule |
| Unison: Table 2 and thread determinism | Selected paper target compared | Passed within stated rule |
| SimGrid: Figure 13 endpoint workloads | Author code executed | Passed within stated rule |
| Smets–Wouters maintained model: fresh mode evaluation and shock decomposition | Adapted run executed | Not established |
The structured campaign record preserves all candidate outcomes. Each guide links numeric outputs and a compact run record; full environments, sources, native results and failed attempts remain in the development evidence directory. Follow the contribution guide when extending a result.