# Results

The public v0.1 page describes the benchmark methodology and placeholder
workload shape. Before publication as a measured claim, replace this file with
raw per-plan rows for each arm:

- Arm A: flat prompt library
- Arm B: public retrieval baseline
- Arm C: K2 scoped context

Required columns:

| plan_id | arm | target_id | targeting_precision | lineage_complete | reviewer_minutes | regression_hit | policy_scope_adherent | notes |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| placeholder | arm-c-k2-scoped-context | target-supportbot-v2.3 | pending | pending | pending | pending | pending | Replace after freeze. |

Do not publish customer-specific claims without customer-frozen tasks and
customer-owned corpora.