# Sibyl 365d 500c Category Benchmark

Run: `run-2026-06-06-sibyl-365d-500c-50q-category-retrieval`

## Scope

- System: Sibyl local memory, sync tier.
- Corpus: 500 companies, 1500 stakeholders, 365 simulated days.
- Stored records: 191000 final records.
- Write calls: 209000.
- Questions: 350, 50 per category.
- Categories: status, milestone, context_stat, role, marker, temporal_topic, negative_trap.
- Retrieval strategy: app-side hybrid exact fallback by question intent.
- Answer mode: `skipped`.

## Summary

| Metric | Value |
| --- | ---: |
| Retrieval passed | 350 / 350 |
| Avg context rows | 2.10 |
| Avg context tokens | 227.76 |
| DB size bytes | 287375360 |
| Ingest elapsed seconds | 47.604 |
| Sonnet answers passed | Skipped |
| Sonnet estimated cost USD | 0.000000 |

## Retrieval By Category

| Category | Retrieval passed |
| --- | ---: |
| status | 50 / 50 |
| milestone | 50 / 50 |
| context_stat | 50 / 50 |
| role | 50 / 50 |
| marker | 50 / 50 |
| temporal_topic | 50 / 50 |
| negative_trap | 50 / 50 |

## Answer By Category

Skipped

## Failures

```json
[]
```