fix(ocr): fix same-page pass — filter locator captions, guard unnumbered formal captions

Two bugs in PrimarySamePagePass identified by review:

1. Previous-page locator captions (e.g. 'Fig. 1 (See legend on previous
   page)') were not filtered from same-page matching. Now skips any
   legend where ocr_figures._is_previous_page_legend_locator() returns
   True, preserving them for future locator-specific handling.

2. Formal captions with no parsed figure number crashed in
   _materialize_match() because _format_figure_id() requires int.
   Now produces figure_unknown_<n> fallback IDs, consistent with
   legacy behavior at ocr_figures.py:3918.

Tests added for both behaviors.

paperforge/worker/ocr_figure_vnext_passes.py

@@ -21,6 +21,8 @@ class PrimarySamePagePass:
             page = _resource_page(legend)
             if page is None:
                 continue
+            if ocr_figures._is_previous_page_legend_locator(legend):
+                continue
             page_groups = [g for g in state.candidate_index.candidate_groups if _resource_page(g) == page]
             for group in page_groups:
                 score = ocr_figures._score_legend_to_group(
@@ -70,9 +72,14 @@ class PrimarySamePagePass:
             for b in state.candidate_index.deduped_legends
             if str(b.get("block_id", "")) == legend.block_id
         )
+        figure_no = proposal.figure_no
         namespace = ocr_figures._extract_figure_namespace(legend_text)
+        if figure_no is None:
+            figure_id = f"figure_unknown_{len(state.matches):03d}"
+        else:
+            figure_id = ocr_figures._format_figure_id(namespace, figure_no)
         return {
-            "figure_id": ocr_figures._format_figure_id(namespace, proposal.figure_no),
+            "figure_id": figure_id,
             "figure_namespace": namespace,
             "figure_number": proposal.figure_no,
             "legend_block_id": legend.block_id,

tests/test_ocr_figure_vnext_passes.py

@@ -45,3 +45,49 @@ def test_primary_same_page_pass_prefers_higher_score_when_two_legends_compete_fo
 
     assert len(report.accepted) == 1
     assert report.accepted[0].figure_no == 2
+
+
+def test_primary_same_page_pass_ignores_previous_page_locator(monkeypatch):
+    blocks = [
+        {"block_id": "c1", "page": 1, "role": "figure_caption", "text": "Fig. 1 (See legend on previous page).", "bbox": [0, 100, 200, 150]},
+        {"block_id": "c2", "page": 1, "role": "figure_caption", "text": "Figure 2. Caption", "bbox": [0, 160, 200, 210]},
+        {"block_id": "a1", "page": 1, "role": "figure_asset", "bbox": [0, 0, 200, 90], "raw_label": "image"},
+    ]
+    corpus = FigureCorpus.from_blocks(blocks, page_width=1200)
+    index = FigureCandidateIndex.from_corpus(corpus)
+    if not index.candidate_groups:
+        index.candidate_groups = [{"group_id": "g1", "page": 1, "asset_block_ids": ["a1"], "media_blocks": [{"block_id": "a1"}], "group_type": "single_asset", "cluster_bbox": [0, 0, 200, 90]}]
+
+    monkeypatch.setattr("paperforge.worker.ocr_figures._score_legend_to_group", lambda *a, **k: {"score": 0.9, "decision": "matched", "evidence": [""]})
+    monkeypatch.setattr("paperforge.worker.ocr_figures.score_figure_caption", lambda *a, **k: {"score": 0.9})
+
+    state = FigurePipelineState(corpus=corpus, candidate_index=index, ledger=OwnershipLedger())
+    report = PrimarySamePagePass().run(state)
+
+    # Only the non-locator caption should match
+    assert len(report.accepted) == 1
+    assert report.accepted[0].figure_no == 2
+
+
+def test_primary_same_page_pass_unnumbered_formal_caption_does_not_crash(monkeypatch):
+    blocks = [
+        {"block_id": "c1", "page": 1, "role": "figure_caption", "text": "Figure. Detailed description of the experimental results showing significant differences between groups.", "bbox": [0, 100, 500, 200]},
+        {"block_id": "a1", "page": 1, "role": "figure_asset", "bbox": [0, 0, 200, 90], "raw_label": "image"},
+    ]
+    corpus = FigureCorpus.from_blocks(blocks, page_width=1200)
+    index = FigureCandidateIndex.from_corpus(corpus)
+    if not index.candidate_groups:
+        index.candidate_groups = [{"group_id": "g1", "page": 1, "asset_block_ids": ["a1"], "media_blocks": [{"block_id": "a1"}], "group_type": "single_asset", "cluster_bbox": [0, 0, 200, 90]}]
+
+    monkeypatch.setattr("paperforge.worker.ocr_figures._score_legend_to_group", lambda *a, **k: {"score": 0.9, "decision": "matched", "evidence": [""]})
+    monkeypatch.setattr("paperforge.worker.ocr_figures.score_figure_caption", lambda *a, **k: {"score": 0.9})
+
+    state = FigurePipelineState(corpus=corpus, candidate_index=index, ledger=OwnershipLedger())
+    report = PrimarySamePagePass().run(state)
+
+    assert len(report.accepted) == 1
+    assert report.accepted[0].figure_no is None
+    assert len(state.matches) == 1
+    # Must produce figure_unknown_* id, not crash
+    assert state.matches[0]["figure_id"].startswith("figure_unknown_")
+    assert state.matches[0]["figure_number"] is None