Paper's languages and terms: drop es/sw; request (not turn), backward (not regression), R1..R10 (#2)

- drop Spanish and Swahili: the paper's Table 3 languages only (fr de zh hi he) (9c38b56ee46041d34b5b713821d27db36523f40f)
- the paper's terms: request (not turn), backward (not regression), R1..R10 graph ids (9bbbb960c3a931d9f58d9895c006e87cff775845)
- drop difficulty: it was "hard" on every follow-up (6c4cdea47566a77ec0bd5a430e815c6d49685376)


Co-authored-by: B Alon <BarA94@users.noreply.huggingface.co>

README.md

@@ -8,8 +8,6 @@ language:
   - zh
   - hi
   - he
-  - es
-  - sw
 task_categories:
   - text-generation
 tags:
@@ -38,10 +36,6 @@ configs:
         path: multilingual/evoweb_benchmark_hi.jsonl
       - split: he
         path: multilingual/evoweb_benchmark_he.jsonl
-      - split: es
-        path: multilingual/evoweb_benchmark_es.jsonl
-      - split: sw
-        path: multilingual/evoweb_benchmark_sw.jsonl
 ---
 
 # EvoWeb: a benchmark for coding agents on evolving web applications
@@ -73,7 +67,7 @@ itself, as one row per app.
 | config | rows | what |
 |---|---|---|
 | `default` | 32 | the benchmark: 32 apps x 10 requests, with the judge's checks |
-| `multilingual` (splits `fr de zh hi he es sw`) | 32 each | the same apps with the user's requests translated; the app's UI labels, seeded data and the judge's checks stay in English |
+| `multilingual` (splits `fr de zh hi he`) | 32 each | the same apps with the user's requests translated; the app's UI labels, seeded data and the judge's checks stay in English |
 
 ```python
 from datasets import load_dataset
@@ -93,15 +87,14 @@ instruction         the first (build) request
 Category, application_type      WebGen-Bench metadata, unused by the harness
 ui_instruct         [ {task, expected_result, task_category} ]  the judge's checks for the first request
 follow_ups          [ 9 x {                                   request k = follow_ups[k-2], k = 2..10
-    difficulty      "hard"
     archetype       derived-state | multi-surface-workflow | interactive-tool | design-restyle
     instruction     the user's request
     source_website, source_feature      the real product feature the request was grounded in
-    ui_instruct     [ 5 x {kind, task, expected_result, task_category, [copied_from_turn]} ]
+    ui_instruct     [ 5 x {kind, task, expected_result, task_category, [copied_from_request]} ]
   } ]
-dependency_graph    {nodes: [{id: "T1".."T10", label, archetype}], edges: [{from, to, what}]}
-global_rules        [ 3 x {rule, introduced_in_turn} ]   app-wide rules the requests introduce
-pinned_seed         {entity, introduced_in_turn, records_gist, [records]}
+dependency_graph    {nodes: [{id: "R1".."R10", label, archetype}], edges: [{from, to, what}]}
+global_rules        [ 3 x {rule, introduced_in_request} ]   app-wide rules the requests introduce
+pinned_seed         {entity, introduced_in_request, records_gist, [records]}
                     the fixed demo data every exact-value check is written against
                     (`records`, present for 3 apps, is a JSON-encoded string here)
 target_app_summary  prose description of the app the 10 requests build towards
@@ -110,7 +103,7 @@ target_app_summary  prose description of the app the 10 requests build towards
 Every follow-up request carries exactly five checks in a fixed order: three **functional**
 (`exact-value` or `feature`), one **integration** (the new feature combined with an earlier one,
 along an edge of `dependency_graph`, or one of the `global_rules` when the request has no incoming
-edge), and one **backward** (`regression`, a check copied from an earlier request).
+edge), and one **backward** (a check copied from an earlier request).
 
 ## How it was built