diff --git a/scripts/benchmarks/hdbscan-benchmark.ts b/scripts/benchmarks/hdbscan-benchmark.ts
deleted file mode 100644
index 2a44d57..0000000
--- a/scripts/benchmarks/hdbscan-benchmark.ts
+++ /dev/null
@@ -1,173 +0,0 @@
-/**
- * HDBSCAN performance benchmark.
- * Compares native implementation against hdbscan-ts.
- */
-
-import { HDBSCAN } from '../../src/clusters/hdbscan.js';
-// @ts-ignore - hdbscan-ts types
-import { HDBSCAN as OldHDBSCAN } from 'hdbscan-ts';
-
-interface BenchmarkResult {
-  ms: number | string;
-  numClusters?: number;
-  noiseCount?: number;
-}
-
-function generateRandomEmbeddings(n: number, dim: number): number[][] {
-  const embeddings: number[][] = [];
-
-  // Create 5 clusters
-  const numClusters = 5;
-  const pointsPerCluster = Math.floor(n / numClusters);
-
-  for (let c = 0; c < numClusters; c++) {
-    // Random center for each cluster
-    const center = new Array(dim).fill(0).map(() => Math.random() * 20 - 10);
-
-    for (let i = 0; i < pointsPerCluster; i++) {
-      const point = center.map((v) => v + (Math.random() - 0.5) * 2);
-      embeddings.push(point);
-    }
-  }
-
-  // Add remaining points as noise
-  while (embeddings.length < n) {
-    embeddings.push(new Array(dim).fill(0).map(() => Math.random() * 100 - 50));
-  }
-
-  return embeddings;
-}
-
-function benchmarkFn<T>(fn: () => T): { ms: number; value: T } {
-  const start = performance.now();
-  const value = fn();
-  const ms = performance.now() - start;
-  return { ms: Math.round(ms), value };
-}
-
-async function benchmarkAsyncFn<T>(fn: () => Promise<T>): Promise<{ ms: number; value: T }> {
-  const start = performance.now();
-  const value = await fn();
-  const ms = performance.now() - start;
-  return { ms: Math.round(ms), value };
-}
-
-async function runBenchmarks() {
-  console.log('HDBSCAN Performance Benchmark');
-  console.log('=============================\n');
-
-  const dim = 128; // Typical embedding dimension
-  const sizes = [100, 250, 500, 750, 1000, 1500, 2000];
-
-  console.log('| Size | Native (parallel) | Native (single) | hdbscan-ts |');
-  console.log('|------|-------------------|-----------------|------------|');
-
-  for (const size of sizes) {
-    const embeddings = generateRandomEmbeddings(size, dim);
-
-    // Native parallel (default)
-    let nativeParallel: BenchmarkResult;
-    try {
-      const result = await benchmarkAsyncFn(async () => {
-        const h = new HDBSCAN({ minClusterSize: 4, parallel: true });
-        return h.fit(embeddings);
-      });
-      nativeParallel = { ms: result.ms, numClusters: result.value.numClusters };
-    } catch (_e) {
-      nativeParallel = { ms: 'error' };
-    }
-
-    // Native single-threaded
-    let nativeSingle: BenchmarkResult;
-    try {
-      const result = await benchmarkAsyncFn(async () => {
-        const h = new HDBSCAN({ minClusterSize: 4, parallel: false });
-        return h.fit(embeddings);
-      });
-      nativeSingle = { ms: result.ms, numClusters: result.value.numClusters };
-    } catch (_e) {
-      nativeSingle = { ms: 'error' };
-    }
-
-    // Old hdbscan-ts (skip for large sizes due to performance)
-    let oldLib: BenchmarkResult;
-    if (size <= 1000) {
-      try {
-        const result = benchmarkFn(() => {
-          const h = new OldHDBSCAN({ minClusterSize: 4, minSamples: 4 });
-          return h.fit(embeddings);
-        });
-        const labels = result.value as number[];
-        const numClusters = new Set(labels.filter((l: number) => l >= 0)).size;
-        oldLib = { ms: result.ms, numClusters };
-      } catch (_e) {
-        oldLib = { ms: 'error' };
-      }
-    } else {
-      oldLib = { ms: 'skip' };
-    }
-
-    console.log(
-      `| ${size.toString().padEnd(4)} | ${String(nativeParallel.ms).padEnd(17)}ms | ${String(nativeSingle.ms).padEnd(15)}ms | ${String(oldLib.ms).padEnd(10)}ms |`,
-    );
-  }
-
-  console.log('\n--- Large Dataset Test ---');
-
-  // Test with larger dataset
-  const largeSize = 3000;
-  console.log(`\nClustering ${largeSize} points (${dim}D)...`);
-
-  const largeData = generateRandomEmbeddings(largeSize, dim);
-
-  const largeResult = await benchmarkAsyncFn(async () => {
-    const h = new HDBSCAN({ minClusterSize: 4, parallel: false });
-    return h.fit(largeData);
-  });
-
-  console.log(`Duration: ${largeResult.ms}ms`);
-  console.log(`Clusters: ${largeResult.value.numClusters}`);
-  console.log(`Noise: ${largeResult.value.noiseCount}`);
-
-  // Test with different options
-  console.log('\n--- Option Comparison (1000 points) ---');
-  const testData = generateRandomEmbeddings(1000, dim);
-
-  // Default (parallel)
-  const defaultResult = await benchmarkAsyncFn(async () => {
-    const h = new HDBSCAN({ minClusterSize: 4 });
-    return h.fit(testData);
-  });
-  console.log(`Default (parallel): ${defaultResult.ms}ms`);
-
-  // Single-threaded
-  const singleResult = await benchmarkAsyncFn(async () => {
-    const h = new HDBSCAN({ minClusterSize: 4, parallel: false });
-    return h.fit(testData);
-  });
-  console.log(`Single-threaded: ${singleResult.ms}ms`);
-
-  // With approximate k-NN
-  const approxResult = await benchmarkAsyncFn(async () => {
-    const h = new HDBSCAN({ minClusterSize: 4, approximateKNN: true, parallel: false });
-    return h.fit(testData);
-  });
-  console.log(`Approximate k-NN: ${approxResult.ms}ms`);
-
-  console.log('\n--- Cluster Quality Comparison ---');
-  console.log(
-    `Native clusters: ${defaultResult.value.numClusters}, noise: ${defaultResult.value.noiseCount}`,
-  );
-
-  try {
-    const oldH = new OldHDBSCAN({ minClusterSize: 4, minSamples: 4 });
-    const oldLabels = oldH.fit(testData) as number[];
-    const oldClusters = new Set(oldLabels.filter((l: number) => l >= 0)).size;
-    const oldNoise = oldLabels.filter((l: number) => l < 0).length;
-    console.log(`hdbscan-ts clusters: ${oldClusters}, noise: ${oldNoise}`);
-  } catch (_e) {
-    console.log('hdbscan-ts: error');
-  }
-}
-
-runBenchmarks().catch(console.error);
diff --git a/test/fixtures/labeled-pairs.json b/test/fixtures/labeled-pairs.json
deleted file mode 100644
index 1bab4b5..0000000
--- a/test/fixtures/labeled-pairs.json
+++ /dev/null
@@ -1,20 +0,0 @@
-{
-  "pairs": [
-    {
-      "chunkIdA": "sess-001-chunk-1",
-      "chunkIdB": "sess-001-chunk-2",
-      "label": "related",
-      "confidence": "high",
-      "source": "same-session-adjacent"
-    },
-    {
-      "chunkIdA": "sess-001-chunk-1",
-      "chunkIdB": "sess-002-chunk-1",
-      "label": "unrelated",
-      "confidence": "high",
-      "source": "cross-project-random"
-    }
-  ],
-  "generatedAt": "2026-01-15T10:00:00.000Z",
-  "chunkCount": 4
-}