/goal Phase 6 — six rubrics × N rounds — track score trend per topic and per dimension. Reads the JSONL evaluation files emitted by the codex evaluators (no build step).
Mean across all sub-scores in each topic, plotted at the timestamp of the evaluator R-round. Dashed line at 8.0 is the convergence target.
Each cell colored 0.0 (red) → 5.0 (amber) → 8.0+ (green = target). Click a row label to view its evolution.
| Topic | Dimension | Rn-1 | Rn | Δ | Direction |
|---|---|---|---|---|---|
| Loading… | |||||