Register ExtractBench Pro as a benchmark and restore RL discovery

Adds `eval.yaml` using the registered `extractbench` framework, with the overall mean and six challenge groups. Restores `rl-environment` alongside `harbor` to test Benchmark/RL badge coexistence.

Only `eval.yaml` and the dataset-card tag change; data and Harbor tasks are unchanged. YAML and task mappings were checked against the existing ExtractBench registration and Pro dataset configuration. Official Benchmark activation may require HF allowlisting; badge precedence remains to be checked after merge.

README.md

@@ -38,6 +38,7 @@ pretty_name: ExtractBench Pro
 size_categories:
 - n<1K
 tags:
+- rl-environment
 - harbor
 - document-extraction
 - structured-extraction

eval.yaml

@@ -0,0 +1,26 @@
+name: ExtractBench Pro
+description: ExtractBench Pro evaluates schema-guided extraction across six challenging
+  document categories.
+evaluation_framework: extractbench
+tasks:
+- id: mean
+  config: extract-bench-pro
+  split: mean
+- id: document_conventions
+  config: extract-bench-pro
+  split: document_conventions
+- id: house_vocabulary
+  config: extract-bench-pro
+  split: house_vocabulary
+- id: house_counting_rules
+  config: extract-bench-pro
+  split: house_counting_rules
+- id: drawn_relations
+  config: extract-bench-pro
+  split: drawn_relations
+- id: charts_inside_documents
+  config: extract-bench-pro
+  split: charts_inside_documents
+- id: long_chart_sheets
+  config: extract-bench-pro
+  split: long_chart_sheets