diff --git a/tests/test_boman_scorer.py b/tests/test_boman_scorer.py
index 021fd58e..fa08d97f 100644
--- a/tests/test_boman_scorer.py
+++ b/tests/test_boman_scorer.py
@@ -6,6 +6,8 @@
 """
 from __future__ import annotations
 
+import math
+
 import pytest
 
 from openamp_foundry.scoring.boman import (
@@ -71,6 +73,11 @@ def test_mixed_canonical_and_unknown(self):
         # "KB" → K(2.465) + B(0.0) → mean 1.2325
         assert boman_index("KB") == pytest.approx(1.2325, abs=1e-4)
 
+    def test_all_non_canonical_returns_zero(self):
+        # Every residue maps to 0.0 via .get(aa, 0.0) → mean = 0.0
+        assert boman_index("XXX") == pytest.approx(0.0, abs=1e-4)
+        assert boman_index("BBBBB") == pytest.approx(0.0, abs=1e-4)
+
 
 class TestBomanActivityScore:
     def test_returns_between_zero_and_one(self):
@@ -104,6 +111,15 @@ def test_rounding(self):
         score = boman_activity_score("KWK")
         assert score == round(score, 4)
 
+    def test_tanh_formula_pinned_against_scorer(self):
+        # Pin the formula 0.5*(1+tanh(bi/2)) against the actual scorer output.
+        # K → boman_index = 2.465 (verified in TestBomanIndex::test_single_lysine)
+        # expected = 0.5 * (1 + tanh(2.465/2)) ≈ 0.9845
+        # If the divisor in the scorer were changed from 2.0 to 3.0 this test fails.
+        k_bi = boman_index("K")
+        expected = round(0.5 * (1.0 + math.tanh(k_bi / 2.0)), 4)
+        assert boman_activity_score("K") == pytest.approx(expected, abs=1e-4)
+
 
 class TestGravyScore:
     def test_empty_returns_zero(self):
diff --git a/tests/test_diversity.py b/tests/test_diversity.py
index 1d0f292c..76c31f14 100644
--- a/tests/test_diversity.py
+++ b/tests/test_diversity.py
@@ -1,5 +1,4 @@
 """Tests for src/openamp_foundry/analysis/diversity.py"""
-import pytest
 from openamp_foundry.analysis.diversity import (
     cluster_panel,
     diversity_stats,
diff --git a/tests/test_pipeline_filters.py b/tests/test_pipeline_filters.py
index 2a57b0d9..12ee3d5b 100644
--- a/tests/test_pipeline_filters.py
+++ b/tests/test_pipeline_filters.py
@@ -2,8 +2,9 @@
 from __future__ import annotations
 
 import json
+from pathlib import Path
 
-
+from openamp_foundry.config import load_config
 from openamp_foundry.pipeline import _passes_length_filter, run_ranking_pipeline, score_candidates
 
 
@@ -127,3 +128,32 @@ def test_novelty_filter_excludes_near_duplicates(tmp_path):
     # But should still be scored and appear in output
     assert len(rows) == 1
     assert rows[0]["scores"]["novelty"] == 0.0
+
+
+def test_phase3_config_has_stricter_safety_filter_than_pipeline():
+    """phase3.yaml max_safety_risk=0.40 (safety>=0.60) vs pipeline.yaml 0.70 (safety>=0.30).
+
+    A candidate with safety=0.50 passes pipeline.yaml but is blocked by phase3.yaml.
+    """
+    repo_root = Path(__file__).parents[1]
+    pipeline_cfg = load_config(repo_root / "configs" / "pipeline.yaml")
+    phase3_cfg = load_config(repo_root / "configs" / "phase3.yaml")
+
+    pipeline_min_safety = 1.0 - float(pipeline_cfg["selection"]["max_safety_risk"])
+    phase3_min_safety = 1.0 - float(phase3_cfg["selection"]["max_safety_risk"])
+
+    # Phase3 must be stricter
+    assert phase3_min_safety > pipeline_min_safety
+
+    # Verify the expected concrete thresholds (max_safety_risk 0.70 → 0.30, 0.40 → 0.60)
+    assert abs(pipeline_min_safety - 0.30) < 0.01, (
+        f"pipeline.yaml min_safety expected ~0.30, got {pipeline_min_safety}"
+    )
+    assert abs(phase3_min_safety - 0.60) < 0.01, (
+        f"phase3.yaml min_safety expected ~0.60, got {phase3_min_safety}"
+    )
+
+    # A candidate with safety=0.50 straddles the two configs
+    candidate_safety = 0.50
+    assert candidate_safety >= pipeline_min_safety, "Should pass pipeline filter"
+    assert candidate_safety < phase3_min_safety, "Should fail phase3 filter"