qa: bare-name guard, lazy-anchor demote, game tie-in noisy markers

Three layered fixes for the JP-dinosaurs lazy-anchor failure mode where
"what dinosaurs were in the first jurassic park film?" returned HYBRID
5/8 with citations to a video-game tie-in chunk (Operation Genesis):

1. Bare-name claim guard. New policy field
   `claim_lattice_min_claim_content_tokens` (default 2). Claims with
   fewer content tokens (>=4 chars, post-spotlight-stopword) classify
   as SCHEMA_INVALID. "Triceratops. [E16]" had 1 content token; lexical
   overlap on a single token is too thin to verify "X is in the FIRST
   film" semantics. Forcing sentence shape raises the coverage bar so
   off-topic chunks can no longer satisfy the citation. Threshold of 2
   keeps narrow factoids viable (e.g. "Steve Jobs co-founded Apple"
   tokenizes to 3 content tokens after stopword strip).

2. Smell → demote. The lazy-anchor smell sidecar was advisory only;
   the verdict could be STRICT while every claim cited the same
   magnet chunk. New policy fields
   `claim_lattice_lazy_anchor_demote_threshold` (0.5) and
   `claim_lattice_lazy_anchor_demote_min_pairs` (3) cap audit_mode at
   HYBRID when ratio + pair-count both met. STRICT now requires
   diverse anchoring across pointers; emits a LAZY_ANCHOR_DEMOTE
   violation record so the auditor can see what fired.

3. Noisy title markers extended. `_NOISY_TITLE_MARKERS` now includes
   "operation genesis", "the game", "video games" so spinoff/tie-in
   articles (Jurassic Park: Operation Genesis, X (NES game), etc.)
   classify as noisy_background_source. Default
   claim_lattice_allowed_source_roles excludes that role, so noisy
   chunks never enter the evidence map for pointer-mode queries.
   Brittle (case-by-case markers) but addresses the JP failure
   directly.

Live JP-dinosaurs after fix: UNGROUNDED 0/11 (was HYBRID 5/8 with
bogus citations). Honest refusal beats false confidence — model could
recover by writing sentence-shape claims; bare-name shortcut blocked.

Test fixtures using minimal placeholder claims ("Trex appears. [E1]")
opt out of the bare-name check via `min_claim_content_tokens=0` since
those tests exercise orthogonal behaviors. 460 tests pass.

aborist/qa/query.py

@@ -381,6 +381,9 @@ DEFAULT_QUERY_POLICY = {
     # cached records.
     "claim_lattice_max_chunks_per_source": 2,
     "claim_lattice_min_citation_coverage": 0.30,
+    "claim_lattice_min_claim_content_tokens": 2,
+    "claim_lattice_lazy_anchor_demote_threshold": 0.5,
+    "claim_lattice_lazy_anchor_demote_min_pairs": 3,
     # JSON variant — `answer_mode="claim_lattice"`. Pairs with grammar-
     # constrained inference (vLLM guided_json, Claude/GPT-4 native
     # JSON, Qwen 3.6 reasoner). Lenient pre-parser keeps the path
@@ -444,9 +447,15 @@ SOURCE_ROLE_BUDGET_WEIGHTS = {
 }
 
 # Title patterns that demote a source's role. Lower-cased substring match.
+# 2026-04-30: extended to catch tie-in spinoff titles. The JP-dinosaurs
+# query lazy-anchored on "Jurassic Park: Operation Genesis" (a video game)
+# whose enumerative dinosaur tables pattern-matched the question shape
+# more cleanly than the actual film article's prose. Adding the explicit
+# game subtitle plus generic markers ("the game", "video games") so
+# similar tie-ins classify as noisy and drop out of the evidence map.
 _NOISY_TITLE_MARKERS = (
-    "score", "music", "soundtrack", "video game", "merchandise",
-    "discography",
+    "score", "music", "soundtrack", "video game", "video games",
+    "merchandise", "discography", "operation genesis", "the game",
 )
 _SEQUEL_TITLE_MARKERS = (
     " iii", " ii)", " ii ", " iv", " v ", " v)", "lost world", "sequel",
@@ -1526,6 +1535,15 @@ def query(
                 min_citation_coverage=float(policy.get(
                     "claim_lattice_min_citation_coverage", 0.30
                 )),
+                min_claim_content_tokens=int(policy.get(
+                    "claim_lattice_min_claim_content_tokens", 3
+                )),
+                lazy_anchor_demote_threshold=float(policy.get(
+                    "claim_lattice_lazy_anchor_demote_threshold", 0.5
+                )),
+                lazy_anchor_demote_min_pairs=int(policy.get(
+                    "claim_lattice_lazy_anchor_demote_min_pairs", 3
+                )),
             )
             rendered = verdict["rendered_text"]
             answer_text = rendered if rendered else raw_answer

aborist/qa/runner.py

@@ -204,6 +204,25 @@ DEFAULT_POLICY = {
     # like "Steve Jobs co-founded Apple" still pass. Folds into
     # governance_policy_hash on change.
     "claim_lattice_min_citation_coverage": 0.30,
+    # Bare-name claim guard. A claim with fewer than this many content
+    # tokens (>=4 chars, post-spotlight-stopword) is rejected as
+    # SCHEMA_INVALID. Catches the JP-dinosaurs lazy-anchor where
+    # "Triceratops. [E16]" passes lexical overlap on a single token
+    # even when E16 is a video-game tie-in chunk rather than the film
+    # article. Default 2: bare-entity-name claims (one content token
+    # after stopword strip) fail; sentence-shape claims pass. Note
+    # ``_content_tokens`` already filters "appears", "shown", etc. so
+    # "Trex appears" → 1 content token (filtered), "Trex appears in
+    # the film" → 2 content tokens (passes). Folds into
+    # governance_policy_hash on change.
+    "claim_lattice_min_claim_content_tokens": 2,
+    # Lazy-anchor smell auto-demote. When >= threshold of verified
+    # pointer-pairs cite a single pointer AND there are >= min_pairs
+    # total, cap audit_mode at HYBRID. The smell sidecar was advisory
+    # only pre-2026-04-30; now it's load-bearing. STRICT requires
+    # diverse anchoring across pointers.
+    "claim_lattice_lazy_anchor_demote_threshold": 0.5,
+    "claim_lattice_lazy_anchor_demote_min_pairs": 3,
     # JSON variant — `answer_mode="claim_lattice"`. Mirrors the
     # multi-source query path. Pairs with grammar-constrained inference
     # (vLLM guided_json, Claude/GPT-4 native JSON, Qwen 3.6 reasoner).
@@ -517,6 +536,15 @@ def ask(
             min_citation_coverage=float(policy.get(
                 "claim_lattice_min_citation_coverage", 0.30
             )),
+            min_claim_content_tokens=int(policy.get(
+                "claim_lattice_min_claim_content_tokens", 3
+            )),
+            lazy_anchor_demote_threshold=float(policy.get(
+                "claim_lattice_lazy_anchor_demote_threshold", 0.5
+            )),
+            lazy_anchor_demote_min_pairs=int(policy.get(
+                "claim_lattice_lazy_anchor_demote_min_pairs", 3
+            )),
         )
         # Rendered prose (literal spans interpolated) is the user-facing
         # answer text — never the model's raw pointer-line output. If

aborist/qa/verify.py

@@ -825,6 +825,9 @@ def _claim_textually_overlaps_evidence(
 
 
 DEFAULT_MAX_POINTERS_PER_CLAIM = 2
+DEFAULT_MIN_CLAIM_CONTENT_TOKENS = 2
+DEFAULT_LAZY_ANCHOR_DEMOTE_THRESHOLD = 0.5
+DEFAULT_LAZY_ANCHOR_DEMOTE_MIN_PAIRS = 3
 
 
 def verify_claim_lattice(
@@ -834,6 +837,9 @@ def verify_claim_lattice(
     allowed_source_roles: tuple[str, ...] = DEFAULT_ALLOWED_SOURCE_ROLES,
     max_pointers_per_claim: int = DEFAULT_MAX_POINTERS_PER_CLAIM,
     min_citation_coverage: float = DEFAULT_MIN_CITATION_COVERAGE,
+    min_claim_content_tokens: int = DEFAULT_MIN_CLAIM_CONTENT_TOKENS,
+    lazy_anchor_demote_threshold: float = DEFAULT_LAZY_ANCHOR_DEMOTE_THRESHOLD,
+    lazy_anchor_demote_min_pairs: int = DEFAULT_LAZY_ANCHOR_DEMOTE_MIN_PAIRS,
 ) -> dict:
     """Deterministic verifier for ``answer_mode="claim_lattice_pointer"``.
 
@@ -974,6 +980,38 @@ def verify_claim_lattice(
             evidence_id_pairs.append([])
             continue
 
+        # Bare-name claim guard. A claim like "Tyrannosaurus rex. [E15]"
+        # has 1 content token; the citation passes any span that mentions
+        # T-rex anywhere, even when E15 is a video-game-behavior chunk
+        # rather than a film-context one. Forcing a sentence-shape claim
+        # ("Tyrannosaurus rex appeared in the first JP film") raises the
+        # token-coverage bar so an off-topic chunk can no longer satisfy
+        # the citation. Folds into governance_policy_hash via
+        # ``claim_lattice_min_claim_content_tokens``.
+        from aborist.qa.evidence import _content_tokens as _ct
+        claim_content_tokens = _ct(claim_text)
+        if len(claim_content_tokens) < min_claim_content_tokens:
+            violations.append({
+                "kind": "SCHEMA_INVALID",
+                "claim_idx": idx,
+                "reason": (
+                    f"bare-name claim ({len(claim_content_tokens)} content "
+                    f"tokens < {min_claim_content_tokens}); write a sentence"
+                ),
+            })
+            claim_statuses.append({
+                "claim_idx": idx,
+                "text": claim_text,
+                "pointer_ids": pointer_ids,
+                "evidence_ids": [],
+                "status": "SCHEMA_INVALID",
+                "reasons": ["bare-name claim — write a sentence with predicate"],
+            })
+            n_pairs += len(pointer_ids)
+            evidence_id_pairs.append([])
+            unverified.append(claim_text)
+            continue
+
         # Pointer-count cap (Rule 9). Catches the encyclopedic-mega-
         # claim where the model produces one line citing every
         # pointer at once. Counts every pointer toward the denominator
@@ -1115,6 +1153,28 @@ def verify_claim_lattice(
         else 0.0
     )
 
+    # Smell → demote. Pre-2026-04-30 the lazy-anchor signal was advisory
+    # only; the verdict could still be STRICT while every claim cited
+    # the same magnet chunk. Now: when ratio ≥ threshold AND total pairs
+    # ≥ floor, cap audit_mode at HYBRID. STRICT becomes unreachable for
+    # answers where one pointer carries every claim — that pattern is
+    # almost never honest verbatim grounding. UNGROUNDED is left alone
+    # (a verdict with zero verified pairs has no smell to flag).
+    lazy_anchor_demoted = False
+    if (
+        audit_mode == "STRICT"
+        and total_pointers >= lazy_anchor_demote_min_pairs
+        and lazy_anchor_ratio >= lazy_anchor_demote_threshold
+    ):
+        audit_mode = "HYBRID"
+        lazy_anchor_demoted = True
+        violations.append({
+            "kind": "LAZY_ANCHOR_DEMOTE",
+            "ratio": round(lazy_anchor_ratio, 3),
+            "min_pairs": lazy_anchor_demote_min_pairs,
+            "threshold": lazy_anchor_demote_threshold,
+        })
+
     return {
         "n_quotes": n_pairs,
         "n_verified": n_pairs_verified,
@@ -1128,6 +1188,7 @@ def verify_claim_lattice(
         "evidence_id_pairs": evidence_id_pairs,
         "pointer_id_distribution": pointer_distribution,
         "lazy_anchor_ratio": lazy_anchor_ratio,
+        "lazy_anchor_demoted": lazy_anchor_demoted,
     }
 
 

tests/test_claim_lattice.py

@@ -264,7 +264,9 @@ def test_no_evidence_pointer_downgrades():
         "Trex appears in the film.\n"  # no pointer
         "Velociraptors are shown. [E2]\n"
     )
-    v = verify_claim_lattice(answer, em)
+    # Opt out of bare-name check (Velociraptors stems to 1 content token
+    # after spotlight stopword strip); test exercises NO_EVIDENCE_POINTER.
+    v = verify_claim_lattice(answer, em, min_claim_content_tokens=0)
     # Mixed: one verified, one NO_EVIDENCE_POINTER → HYBRID.
     assert v["audit_mode"] == "HYBRID"
     assert any(viol["kind"] == "NO_EVIDENCE_POINTER" for viol in v["violations"])
@@ -279,7 +281,9 @@ def test_partial_resolution_yields_hybrid():
         "Tyrannosaurus rex appears. [E1]\n"
         "Made-up. [E99]\n"
     )
-    v = verify_claim_lattice(answer, em)
+    # Test exercises mixed-pointer-resolution behavior; bare-name check
+    # is orthogonal so opt out.
+    v = verify_claim_lattice(answer, em, min_claim_content_tokens=0)
     assert v["audit_mode"] == "HYBRID"
     assert v["n_verified"] == 1
     assert v["n_quotes"] == 2
@@ -288,7 +292,7 @@ def test_partial_resolution_yields_hybrid():
 def test_partial_pointer_within_claim_yields_partial():
     em = build_evidence_map(_sample_chunks())
     answer = "Tyrannosaurus rex appears. [E1,E99]\n"
-    v = verify_claim_lattice(answer, em)
+    v = verify_claim_lattice(answer, em, min_claim_content_tokens=0)
     assert v["audit_mode"] == "HYBRID"
     assert v["claim_statuses"][0]["status"] == "EVIDENCE_LINKED_PARTIAL"
     # Renderer keeps only resolved ids.
@@ -895,7 +899,10 @@ def test_lazy_anchor_ratio_one_when_all_share_pointer():
         "Tyrannosaurus rex roars. [E1]\n"
         "Tyrannosaurus rex broke out. [E1]\n"
     )
-    v = verify_claim_lattice(answer, em)
+    # Test exercises the smell ratio counter; bare-name check is
+    # orthogonal so opt out so all three claims reach EVIDENCE_LINKED
+    # and populate pointer_distribution.
+    v = verify_claim_lattice(answer, em, min_claim_content_tokens=0)
     assert v["lazy_anchor_ratio"] == 1.0
     assert v["pointer_id_distribution"] == {"E1": 3}
 
@@ -908,7 +915,7 @@ def test_lazy_anchor_ratio_one_third_when_diversified():
         "Velociraptors stalk workers. [E2]\n"
         "The climactic Tyrannosaurus moment. [E1]\n"
     )
-    v = verify_claim_lattice(answer, em)
+    v = verify_claim_lattice(answer, em, min_claim_content_tokens=0)
     # 2 cite E1, 1 cites E2 → max share 2/3.
     assert v["pointer_id_distribution"] == {"E1": 2, "E2": 1}
     assert abs(v["lazy_anchor_ratio"] - 2 / 3) < 1e-9
@@ -942,7 +949,9 @@ def test_citation_mismatch_partial_when_one_pointer_overlaps():
     # E1 contains "Tyrannosaurus rex"; E2 contains "Velociraptors".
     # Claim mentions Tyrannosaurus → E1 overlaps, E2 does not.
     answer = "Tyrannosaurus rex appears. [E1,E2]\n"
-    v = verify_claim_lattice(answer, em)
+    # Test exercises CITATION_MISMATCH on a single pointer; bare-name
+    # check is orthogonal so opt out.
+    v = verify_claim_lattice(answer, em, min_claim_content_tokens=0)
     assert v["audit_mode"] == "HYBRID"
     assert v["n_verified"] == 1
     assert v["claim_statuses"][0]["status"] == "EVIDENCE_LINKED_PARTIAL"
@@ -962,7 +971,10 @@ def test_pure_stopword_claim_passes_overlap_check_vacuously():
     # below has effectively no content tokens after stopword filter
     # (just function words), so the overlap check shouldn't fail it.
     answer = "It was so. [E1]\n"
-    v = verify_claim_lattice(answer, em)
+    # Bare-name check would reject this for 0 content tokens; opt out
+    # since this test is specifically about the overlap rule's
+    # vacuous-True branch for claims with no content tokens.
+    v = verify_claim_lattice(answer, em, min_claim_content_tokens=0)
     # Overlap check passed; claim is short but valid.
     assert v["claim_statuses"][0]["status"] == "EVIDENCE_LINKED"