Card: remove selection/provenance process detail; keep results, cost, usage

README.md

@@ -33,8 +33,7 @@ Duo picks per query. On a mixed workload it beats **both** constituents.
 ## Results
 
 MAP@10, full corpus, all ground-truth queries, one evaluator (`pytrec_eval map_cut.10`),
-paired bootstrap 10,000 resamples. Routing policy frozen on development data before any
-benchmark was scored.
+paired bootstrap 10,000 resamples.
 
 | benchmark | MODA | MODA Pro Lite+ | **MODA Duo** |
 |---|---:|---:|---:|
@@ -86,14 +85,11 @@ The router is a callable — replace it with any policy that maps a query to a c
 duo = Duo(router=lambda q: "moda" if looks_like_a_description(q) else "moda_pro_lite_plus")
 ```
 
-## Selection and provenance
+## Evaluation
 
-- Constituents: MODA and MODA Pro Lite+, each with its own published harness, unchanged.
-- The routing policy was chosen on OpenVTON validation and GLAMI development data by
-  minimising worst-regime regret against the per-regime oracle, frozen, then applied once to
-  the benchmarks above. No benchmark was consulted during selection.
-- Full receipts, per-query results, and the sensitivity curve are in the
-  [repository](https://github.com/hopit-ai/Moda).
+All figures are full corpus, all ground-truth queries, MAP@10 under one evaluator
+(`pytrec_eval map_cut.10`), paired bootstrap with 10,000 resamples. Per-query results are in
+the [repository](https://github.com/hopit-ai/Moda).
 
 ## Related