pj/messageboardbench
1
0 Fork 0
mirror of https://github.com/priyanshujain/messageboardbench.git synced 2026-10-02 11:07:07 +00:00
Code Issues Packages Projects Releases Wiki Activity

archive historical results and add token budget logs

Browse Source
This commit is contained in:
pj committed 2026-09-25 22:07:26 +05:30
1 parent 638e978227
commit e94b75c6f5
813 files changed
+12631 -410710
Download Patch File Download Diff File Expand all files Collapse all files

No files matched your search

results/README.md
+3 -21
View File
@@ -1,23 +1,5 @@
# Experiment evidence
# Reviewed evidence
Start with [EXPERIMENT.md](../EXPERIMENT.md). These dated bundles preserve completed
analyses and supporting artifacts; they are not alternative current plans.
Current SWE analyses and reviewed metrics live with their retained experiments in [`experiments/`](../experiments/).
| Bundle | Use |
|---|---|
| [board-muse-sept8](board-muse-sept8/RESULTS.md) | Full matched Muse Contributor replication: one attributed adoption, one possible unattributed adoption; all eight gaming artifacts validated. |
| [board-interface-v2-sept8](board-interface-v2-sept8/RESULTS.md) | Latest matched GLM pilot: one attributed cross-task adoption after peer receipt. |
| [board-pilot-sept8](board-pilot-sept8/RESULTS.md) | Prior interface: publication but zero board reads; task validation and SWE readiness. |
| [model-comparison-sept7](model-comparison-sept7/RESULTS.md) | Tiny GLM/Muse Contributor feasibility comparison, not model-level rates. |
| [token-comparison-sept7](token-comparison-sept7/REPORT.md) | Frozen historical token audit and recovered baseline accounting. |
| [team-pilot-sept7](team-pilot-sept7/RESULTS.md) | Historical shared-directory experiment with different prompts and limits. |
| [sept10-revision](sept10-revision/baseline-audit.json) | Historical baseline audit supporting the revision. |
Historical raw Inspect runs are preserved locally in ignored `../archive/logs/`.
Current SWE run logs live in tracked `../logs/`. Migrated reports, reviews, probe
outputs and exports are byte-preserved. Embedded absolute paths
record their original locations; resolve them with
[the migration map](../docs/migration/messageboard-20260908.json). Historical
scripts inside bundles are evidence, not the current run entry points. Use
`just board`, `scripts/board_report.py`, or the portable offline tools in
`scripts/analysis/` for new work, always with a fresh output directory.
The historical LiveCodeBench result bundles are preserved byte for byte in local, gitignored `archive/results/`. The frozen [migration manifest](../docs/migration/messageboard-20260908.json) records their original repository paths and hashes; `just evidence-check` resolves those paths through the archive.