Per-tile deterministic chips · no LLM judge
- ⏱ latency
- end-to-end response time; orange = first turn per backend (carries model warmup).
- chart refs
- count of chart records cited — a COUNT, not a grounding/quality signal; the authoritative call is the human adjudication on each tile.
- not surfaced
- tokens / finish_reasons / response model — not returned by
/chat (OTel-deferred). - ranking
- drag tiles within a question to rank backends; rank + adjudication export to separate files.