Update BENCHMARKS.md with v0.4.3 results

BENCHMARKS.md

@@ -8,22 +8,22 @@ Measured on Apple M3 Max, 36GB RAM.
 
 | Metric | Sediment | ChromaDB | Mem0 |
 |--------|----------|----------|------|
-| Recall@1 | 45.0% | 47.0% | 47.0% |
-| Recall@3 | **69.0%** | 69.0% | 69.0% |
-| Recall@5 | 78.0% | 78.5% | 78.5% |
-| Recall@10 | **90.5%** | 90.0% | 90.0% |
-| MRR | 59.0% | 60.8% | 60.8% |
+| Recall@1 | **50.0%** | 47.0% | 47.0% |
+| Recall@3 | 69.0% | 69.0% | 69.0% |
+| Recall@5 | 77.5% | 78.5% | 78.5% |
+| Recall@10 | 89.5% | 90.0% | 90.0% |
+| MRR | **61.9%** | 60.8% | 60.8% |
 | Recency@1 | **100%** | 14% | 14% |
 | Consolidation | **99%** | 0% | 0% |
-| Store p50 | 22ms | 696ms | 16ms |
-| Recall p50 | 26ms | 694ms | 8ms |
+| Store p50 | 49ms | 696ms | 16ms |
+| Recall p50 | 103ms | 694ms | 8ms |
 
 ## Key takeaways
 
-- **Retrieval quality**: Within 0.5pp of ChromaDB on Recall@5 (78.0% vs 78.5%), matching on Recall@3
+- **Retrieval quality**: Best R@1 (50.0%) and MRR (61.9%) — top result is correct more often than alternatives
 - **Temporal correctness**: 100% Recency@1 — updated memories always rank first. Others: 14%
 - **Deduplication**: 99% consolidation rate — near-duplicates auto-merged. Others: 0%
-- **Latency**: 32x faster store than ChromaDB (22ms vs 696ms). All operations local, no network
+- **Latency**: 14x faster store than ChromaDB (49ms vs 696ms). All operations local, no network
 
 ## Methodology