Data Coverage Tests 5–13
LongMemEval Phase 2 · Memory Intelligence Validation
Critical Tests Progress0/5
Test 5
Temporal Continuity
Time anchoring
Critical
Test 6
Provenance Chain
Source tracing
Critical
Test 7
Recall Determinism
Query consistency
Critical
Test 8
Dedup Accuracy
Merge quality
Test 9
Contradiction Survivability
Conflict tolerance
Test 10
Confidence Calibration
Score integrity
Critical
Test 11
Memory Decay
Supersession
Test 12
Cross-Companion Transfer
Knowledge mesh
Test 13
Narrative Reconstruction
Story generation
Critical
Legend
Critical: Lock in truth plumbing first (5, 6, 7, 10, 13)
Structural: Tolerance & mesh features (8, 9, 11, 12)
Run Order: 5→6→7→9→11→12→13, then 10→8
Embeddings: Not required (can use regex/theme search)