Source drift radar
Source changes become review signals.
Schema movement, row disappearance, column rename/parser changes, and unusual score shifts are grouped by source.
| Source | Severity | Captured | Parser | Status | Reasons |
|---|---|---|---|---|---|
| Arena | ok | 2026-10-03 | 0.6.0 | verified | No drift warning |
| LiveBench | ok | 2026-10-03 | 0.5.0 | verified | No drift warning |
| LLMBase | ok | 2026-10-03 | 0.2.0 | relay | No drift warning |
| MTEB | ok | 2026-04-22 | 0.2.0 | verified | No drift warning |
| OpenCompass | ok | 2026-04-22 | 0.2.0 | verified | No drift warning |
| Provider official evals | ok | 2026-04-19 | unrecorded | verified | No drift warning |
| Terminal-Bench | ok | 2026-10-03 | 0.3.0 | verified | No drift warning |
| Vals AI | ok | 2026-10-03 | 0.2.0 | verified | No drift warning |
| Artificial Analysis | warning | 2026-10-03 | 0.5.0 | verified | source snapshot changed: Source snapshot changed |
| BridgeBench | warning | 2026-05-15 | 0.3.0 | verified | scores shifted: Benchmark coverage or values moved |
| Scale Labs | warning | 2026-10-03 | 0.2.0 | verified | source snapshot changed: Source snapshot changed |