Open the Experiment groups panel above and either type a regex/substring (e.g. ^pi05-tabletop-) and click Select matched, or click individual run chips to multi-select. Name the group, click Add group from selection, and the charts below will appear with the aggregated statistics.
1. Top-line numbers by split
Per-split mean SR & Score across all tasks & (within-group) seeds
2. Train → Test generalization
x = Validation-Train, y = Test · y=x reference
3. Cluster radar
Each axis = one cluster from the bundled taxonomy
4. Cluster breakdowns
Per-category cluster aggregates · Test by default
5. Generalization dimensions
Background · Instruction · Object · Mix — Test split by default