Card: document which estimator produces the published belief-F1 column

README.md

@@ -51,6 +51,15 @@ Key columns: `experiment_name` (append-safe grouping), `cell_id`, `arm`, `deadli
 `instance_id`, `seed`, `deal_rate`, `normalized_primary` (realized normalized utilitarian welfare over the
 instance's feasible ceiling; a no-deal scores 0), `rounds_to_agreement`, `belief_auc_accept_set_f1`.
 
+### Reproducing the headline belief-accuracy numbers
+
+The belief-F1 figures quoted in research note 0061 (0.210 / 0.228 / 0.241 / 0.244 / 0.241 across r4-r64) are
+**`cells.csv.belief_auc_accept_set_f1`** — a per-episode time-average, then aggregated across episodes.
+Taking a flat mean of the `accept_set_f1` column in `belief_turns.csv.gz` instead gives
+0.207 / 0.225 / 0.239 / 0.243 / 0.240, about 0.003 lower at every deadline, because that weights turns
+equally rather than episodes and so over-weights long episodes. Both are defensible summaries of the same
+data and the saturation conclusion is identical under either; use `cells.csv` to match the published table.
+
 ## Experiment-name mapping
 
 | experiment_name | description |