tasks/todo.md

Leaderboard Card Fixes — Task List

Source: annotated mockup review (2026-06-02). Targets index.html lines 161–292.

EnterpriseOps-Gym card

• T1 — Clickable title. Wrap <h3>EnterpriseOps-Gym</h3> (L165) in a link to the EOG webpage. Add hover-over info.
• T2 — Drop "cascade". Anthropic · cascade → Anthropic (L171).
• T3 — Rename metric label. Success rate · Oracle mode → Task Success Rate · Oracle mode (L175).
• T4 — Metric hover. Add hover-over def on the metric: "A task passes only if all verification conditions are met."

EVA-Bench card

• T5 — Clickable title. Wrap <h3>EVA-Bench</h3> (L217) in a link to the EVA webpage. Add hover-over info.
• T6 — Accuracy section label. Accuracy → EVA-Accuracy (L224).
• T7 — Accuracy metric tag. EVA-A · PASS@3 → Pass@1 (L225).
• T8 — Accuracy metric hover. Def: "Scores for accuracy. All values normalized to 0–1 (higher is better). 95% bootstrap confidence intervals shown for each value."
• T9 — Cascade subtitle. cascade · mixed (L230) — keep as-is (annotation just maps it to "Mixed Models | Cascade"; current value already correct). Confirm no change.
• T10 — Experience section label. Experience → EVA-Experience (L256).
• T11 — Experience metric tag. EVA-X · PASS@3 → Pass@1 (L257).
• T12 — Experience metric hover. Def: "Scores for conversational experience. All values normalized to 0–1 (higher is better). 95% bootstrap confidence intervals shown for each value."
• T13 — S2S subtitle. Google · S2S (L262) — keep as-is (annotation maps it to "Google | Speech-to-Speech"; current already correct). Confirm no change.

Open questions (need answers before implementing)

1. Title link URLs. Use same as the "View full leaderboard" links? EOG → https://enterpriseops-gym.github.io/, EVA → https://servicenow.github.io/eva/. Confirm.
2. Hover mechanism. Native title="" attribute (simple, less styled) vs custom CSS/JS tooltip (accessible, on-brand, more work)? Recommend custom accessible tooltip.
3. T7/T11. Confirm full replace of EVA-A · PASS@3 / EVA-X · PASS@3 with just Pass@1 (drops the EVA-A/EVA-X code and changes @3 → @1).