fix(kaggle): update dataset path to match user's custom kaggle dataset output

benchmark/smriti_longmemeval_kaggle.ipynb

@@ -32,7 +32,7 @@
    "id": "title",
    "metadata": {},
    "source": [
-    "# Smriti × LongMemEval — Official Benchmark Run\n",
+    "# Smriti \u00d7 LongMemEval \u2014 Official Benchmark Run\n",
     "\n",
     "**What this notebook does:**\n",
     "- Evaluates Smriti's pure-code memory algorithm (Gap=0.08, MaxCutoff=0.52) against the official LongMemEval benchmark (ICLR 2025)\n",
@@ -45,8 +45,8 @@
     "**Algorithm:** 100% pure code, zero LLM calls at query time.\n",
     "```\n",
     "Vector Search (threshold=0.85)\n",
-    "  → Entity/SVO Filter (zero-overlap candidates dropped)\n",
-    "  → Bayesian Gap Cutoff (gap=0.08, max_cutoff=0.52)\n",
+    "  \u2192 Entity/SVO Filter (zero-overlap candidates dropped)\n",
+    "  \u2192 Bayesian Gap Cutoff (gap=0.08, max_cutoff=0.52)\n",
     "```"
    ]
   },
@@ -57,7 +57,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "# ── Install dependencies ──────────────────────────────────────────────────────\n",
+    "# \u2500\u2500 Install dependencies \u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\n",
     "import subprocess, sys\n",
     "\n",
     "pkgs = [\n",
@@ -77,11 +77,11 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "# ── Configuration ─────────────────────────────────────────────────────────────\n",
+    "# \u2500\u2500 Configuration \u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\n",
     "import os\n",
     "\n",
-    "# Dataset paths — update if your Kaggle dataset has a different name\n",
-    "DATA_DIR        = '/kaggle/input/longmemeval'\n",
+    "# Dataset paths \u2014 update if your Kaggle dataset has a different name\n",
+    "    DATA_DIR        = '/kaggle/input/longmemeval-clean/data/longmemeval'\\n",
     "S_SPLIT_PATH    = os.path.join(DATA_DIR, 'longmemeval_s_cleaned.json')\n",
     "M_SPLIT_PATH    = os.path.join(DATA_DIR, 'longmemeval_m_cleaned.json')\n",
     "OUTPUT_DIR      = '/kaggle/working'\n",
@@ -104,7 +104,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "# ── In-memory stores (self-contained, no smriti_core import needed) ───────────\n",
+    "# \u2500\u2500 In-memory stores (self-contained, no smriti_core import needed) \u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\n",
     "import uuid\n",
     "from dataclasses import dataclass, field\n",
     "from typing import List, Optional, Dict, Any\n",
@@ -176,7 +176,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "# ── Smriti Mock Server (pure-code algorithm) ──────────────────────────────────\n",
+    "# \u2500\u2500 Smriti Mock Server (pure-code algorithm) \u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\n",
     "import re\n",
     "import threading\n",
     "import time\n",
@@ -306,7 +306,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "# ── Evaluation Engine ─────────────────────────────────────────────────────────\n",
+    "# \u2500\u2500 Evaluation Engine \u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\n",
     "import json\n",
     "import math\n",
     "import time as time_mod\n",
@@ -333,7 +333,7 @@
     "\n",
     "def run_eval(cases, label='S'):\n",
     "    print(f'\\n{\"=\"*60}')\n",
-    "    print(f'  Running LongMemEval-{label} — {len(cases)} cases')\n",
+    "    print(f'  Running LongMemEval-{label} \u2014 {len(cases)} cases')\n",
     "    print(f'{\"=\"*60}')\n",
     "\n",
     "    results     = []\n",
@@ -408,7 +408,7 @@
     "            running_f1  = sum(r['f1'] for r in results) / len(results)\n",
     "            print(f'  [{i+1:4d}/{len(cases)}]  Sub={running_sub:.3f}  F1={running_f1:.3f}  lat={lat_ms:.0f}ms')\n",
     "\n",
-    "    # ── Summary ────────────────────────────────────────────────────────────────\n",
+    "    # \u2500\u2500 Summary \u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\n",
     "    n         = len(results)\n",
     "    total_sub = sum(r['substring_match'] for r in results) / n\n",
     "    total_f1  = sum(r['f1'] for r in results) / n\n",
@@ -461,7 +461,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "# ── Load datasets ─────────────────────────────────────────────────────────────\n",
+    "# \u2500\u2500 Load datasets \u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\n",
     "def load_split(path):\n",
     "    with open(path, 'r', encoding='utf-8') as f:\n",
     "        data = json.load(f)\n",
@@ -479,7 +479,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "# ── Run LongMemEval-S (500 cases, ~30 mins on GPU) ───────────────────────────\n",
+    "# \u2500\u2500 Run LongMemEval-S (500 cases, ~30 mins on GPU) \u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\n",
     "summary_s = run_eval(cases_s, label='S')\n",
     "\n",
     "# Save results\n",
@@ -501,7 +501,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "# ── Run LongMemEval-M (500 cases, ~3 hrs on GPU) ─────────────────────────────\n",
+    "# \u2500\u2500 Run LongMemEval-M (500 cases, ~3 hrs on GPU) \u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\n",
     "# NOTE: This is the massive split (500 sessions per question, ~1.5M tokens each)\n",
     "# It will take 2-4 hours. Kaggle allows up to 12 hours.\n",
     "if cases_m:\n",
@@ -517,7 +517,7 @@
     "\n",
     "    print('M-split done and saved.')\n",
     "else:\n",
-    "    print('M-split file not found — skipping.')\n",
+    "    print('M-split file not found \u2014 skipping.')\n",
     "    summary_m = None"
    ]
   },
@@ -528,7 +528,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "# ── Generate BENCHMARKS.md ────────────────────────────────────────────────────\n",
+    "# \u2500\u2500 Generate BENCHMARKS.md \u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\n",
     "from datetime import datetime\n",
     "\n",
     "def make_benchmarks_md(summary_s, summary_m=None):\n",
@@ -545,7 +545,7 @@
     "        '## LongMemEval (ICLR 2025)',\n",
     "        '',\n",
     "        '> **Dataset:** `xiaowu0162/longmemeval-cleaned`  ',\n",
-    "        '> **Metric:** Substring Match (Recall) — does the retrieved context contain the gold answer?',\n",
+    "        '> **Metric:** Substring Match (Recall) \u2014 does the retrieved context contain the gold answer?',\n",
     "        '',\n",
     "    ]\n",
     "\n",
@@ -601,4 +601,4 @@
    ]
   }
  ]
-}
+}