Spaces:
Running
Running
File size: 14,208 Bytes
1df72c3 edf774b ef29ae4 1df72c3 edf774b 4e141cd edf774b 4e141cd edf774b 4e141cd edf774b 4e141cd edf774b 1df72c3 4e141cd edf774b 4e141cd edf774b 4e141cd edf774b 4e141cd edf774b 1df72c3 4e141cd edf774b 4e141cd edf774b 4e141cd edf774b 4e141cd edf774b 4e141cd edf774b 4e141cd edf774b 4e141cd edf774b 4e141cd edf774b 32a93fb 4e141cd edf774b 4e141cd edf774b 4e141cd edf774b 4e141cd edf774b 4e141cd edf774b 4e141cd edf774b 4e141cd edf774b 4e141cd 1df72c3 edf774b 1df72c3 4e141cd ef29ae4 4e141cd edf774b 4ac51bd edf774b 357ed4f edf774b 357ed4f edf774b 58a5019 edf774b 58a5019 edf774b 9a1d981 58a5019 edf774b 58a5019 edf774b 58a5019 edf774b 4e141cd edf774b 9a1d981 58a5019 4e141cd edf774b 4e141cd edf774b 360d20a 58a5019 edf774b 360d20a 58a5019 edf774b 32a93fb edf774b 357ed4f 58a5019 357ed4f edf774b 58a5019 edf774b 58a5019 edf774b 58a5019 edf774b 58a5019 edf774b 58a5019 edf774b 58a5019 edf774b 58a5019 edf774b 1df72c3 32a93fb 6383139 32a93fb 1df72c3 4e141cd |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 |
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>Simple Math Benchmark</title>
<style>
body {
font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, Oxygen, Ubuntu, Cantarell, 'Open Sans', 'Helvetica Neue', sans-serif;
line-height: 1.6;
color: #333;
max-width: 1000px;
margin: 0 auto;
padding: 20px;
background-color: #f9f9f9;
}
h1, h2, h3, h4 {
color: #2d3748;
}
h1 {
font-size: 2.5rem;
margin-bottom: 2rem;
text-align: center;
color: #1a365d;
border-bottom: 3px solid #4299e1;
padding-bottom: 0.5rem;
}
h2 {
font-size: 1.8rem;
margin-top: 2rem;
border-bottom: 1px solid #e2e8f0;
padding-bottom: 0.5rem;
}
.container {
background-color: white;
border-radius: 8px;
padding: 20px;
box-shadow: 0 4px 6px rgba(0, 0, 0, 0.1);
margin-bottom: 20px;
}
.model-card {
background-color: #f0f5ff;
padding: 15px;
border-radius: 8px;
margin-bottom: 15px;
border-left: 4px solid #4299e1;
}
.model-info {
display: flex;
justify-content: space-between;
margin-bottom: 10px;
}
.model-name {
font-weight: bold;
font-size: 1.2rem;
color: #2b6cb0;
}
.model-params {
background-color: #ebf8ff;
padding: 3px 8px;
border-radius: 12px;
font-size: 0.9rem;
color: #2c5282;
}
.performance-highlight {
background-color: #e6fffa;
padding: 10px;
border-radius: 4px;
margin-top: 10px;
border-left: 3px solid #38b2ac;
}
.chart-container {
display: flex;
justify-content: space-between;
flex-wrap: wrap;
gap: 20px;
margin: 30px 0;
}
.chart {
flex: 1;
min-width: 300px;
background: white;
padding: 15px;
border-radius: 8px;
box-shadow: 0 2px 4px rgba(0, 0, 0, 0.05);
}
.chart img {
width: 100%;
height: auto;
border-radius: 4px;
cursor: zoom-in;
transition: transform 0.3s ease;
}
.modal {
display: none;
position: fixed;
z-index: 1000;
left: 0;
top: 0;
width: 100%;
height: 100%;
overflow: auto;
background-color: rgba(0,0,0,0.9);
}
.modal-content {
margin: auto;
display: block;
max-width: 90%;
max-height: 90%;
}
.close {
position: absolute;
top: 15px;
right: 35px;
color: #f1f1f1;
font-size: 40px;
font-weight: bold;
transition: 0.3s;
cursor: pointer;
}
.close:hover,
.close:focus {
color: #bbb;
text-decoration: none;
}
.chart-caption {
text-align: center;
margin-top: 10px;
font-style: italic;
color: #718096;
}
table {
width: 100%;
border-collapse: collapse;
margin: 20px 0;
}
th, td {
padding: 12px 15px;
text-align: left;
border-bottom: 1px solid #e2e8f0;
}
th {
background-color: #edf2f7;
font-weight: bold;
}
tr:hover {
background-color: #f7fafc;
}
.highlight {
background-color: #fffbeb;
}
.footer {
margin-top: 40px;
text-align: center;
color: #718096;
font-size: 0.9rem;
}
.key-finding {
background-color: #ebf8ff;
padding: 15px;
border-radius: 8px;
margin: 20px 0;
border-left: 4px solid #3182ce;
}
.key-finding h3 {
margin-top: 0;
color: #2b6cb0;
}
</style>
</head>
<body>
<div class="container">
<h1>Simple Math Benchmark</h1>
<div class="key-finding">
<h3>Key Insight</h3>
<p>The data shows that some small language models (under 0.5B parameters) can demonstrate significant mathematical abilities, challenging the assumption that large parameter counts are required for basic math reasoning. Models tested using this <a href="https://github.com/FlameF0X/Simple-Math-Benchmark">link</a>.</p>
</div>
<h2>Model Overview</h2>
<p>This analysis compares the performance of various language models on mathematical problems, with a focus on compact models:</p>
<div class="model-card">
<div class="model-info">
<span class="model-name">FlameF0X/MathGPT2</span>
<span class="model-params">81.9M parameters</span>
</div>
<p>The highest performer among tested models, demonstrating remarkable mathematical abilities despite its tiny parameter count. Shows particular strength in addition operations with 58.3% accuracy and subtraction with 57.1% accuracy.</p>
<div class="performance-highlight">
<strong>Overall math accuracy:</strong> 42.0% on 100 test questions
</div>
</div>
<div class="model-card">
<div class="model-info">
<span class="model-name">aquiffoo/aquif-moe-800m</span>
<span class="model-params">800M parameters</span>
</div>
<p>The second best performer, scoring 39.0% overall accuracy. Shows exceptional performance in subtraction (76.2%) and solid performance in addition (54.5%).</p>
<div class="performance-highlight">
<strong>Operation strength:</strong> 76.2% accuracy on subtraction
</div>
</div>
<div class="model-card">
<div class="model-info">
<span class="model-name">GoofyLM/BrainrotLM-Assistant-362M</span>
<span class="model-params">362M parameters</span>
</div>
<p>Shows moderate mathematical abilities with 12.0% overall accuracy. Demonstrates particular strength in division operations (38.9%) and subtraction (22.7%).</p>
<div class="performance-highlight">
<strong>Operation strength:</strong> 38.9% accuracy on division
</div>
</div>
<h2>Performance Analysis</h2>
<div class="chart-container">
<div class="chart">
<img src="acc.png" alt="Chart showing model accuracy by operation type">
<div class="chart-caption">Figure 1: Accuracy by Mathematical Operation (%)</div>
</div>
<div class="chart">
<img src="numb.png" alt="Chart showing model performance on math problems">
<div class="chart-caption">Figure 2: Correct vs Incorrect Answers (100 questions each)</div>
</div>
</div>
<!-- Modal for image zoom -->
<div id="imageModal" class="modal">
<span class="close">×</span>
<img class="modal-content" id="zoomedImage">
</div>
<h2>Detailed Accuracy Results</h2>
<table>
<thead>
<tr>
<th>Model</th>
<th>Addition (+)</th>
<th>Subtraction (-)</th>
<th>Multiplication (*)</th>
<th>Division (/)</th>
<th>Exponentiation (**)</th>
<th>Overall</th>
</tr>
</thead>
<tbody>
<tr class="highlight">
<td>MathGPT2 (81.9M)</td>
<td>58.3%</td>
<td>57.1%</td>
<td>45.0%</td>
<td>24.1%</td>
<td>0.0%</td>
<td>42.0%</td>
</tr>
<tr>
<td>aquif-moe-800m (800M)</td>
<td>54.5%</td>
<td>76.2%</td>
<td>21.9%</td>
<td>18.2%</td>
<td>0.0%</td>
<td>39.0%</td>
</tr>
<tr>
<td>BrainrotLM-Assistant-362M (362M)</td>
<td>0.0%</td>
<td>22.7%</td>
<td>0.0%</td>
<td>38.9%</td>
<td>0.0%</td>
<td>12.0%</td>
</tr>
<tr>
<td>gonzalez-v1</td>
<td>5.3%</td>
<td>8.3%</td>
<td>0.0%</td>
<td>0.0%</td>
<td>0.0%</td>
<td>3.0%</td>
</tr>
<tr>
<td>VLM-1 (124M)</td>
<td>3.4%</td>
<td>0.0%</td>
<td>0.0%</td>
<td>4.3%</td>
<td>0.0%</td>
<td>2.0%</td>
</tr>
<tr>
<td>gpt2</td>
<td>0.0%</td>
<td>7.4%</td>
<td>0.0%</td>
<td>0.0%</td>
<td>0.0%</td>
<td>2.0%</td>
</tr>
<tr>
<td>Snowflake-G0-Release</td>
<td>0.0%</td>
<td>0.0%</td>
<td>0.0%</td>
<td>0.0%</td>
<td>0.0%</td>
<td>0.0%</td>
</tr>
</tbody>
</table>
<h2>Key Observations</h2>
<ul>
<li><strong>Size doesn't always matter:</strong> MathGPT2 with only 81.9M parameters demonstrates impressive mathematical abilities, achieving 42.0% overall accuracy.</li>
<li><strong>Operation specialization:</strong> MathGPT2 excels at addition (58.3%) and subtraction (57.1%), while aquif-moe-800m shows exceptional strength in subtraction operations (76.2%).</li>
<li><strong>Architectural importance:</strong> The results suggest that architecture design and training approach may be more important than raw parameter count for specialized tasks.</li>
<li><strong>Zero performance:</strong> One of the tested models (Snowflake-G0-Release) showed no measurable mathematical ability on this test set.</li>
<li><strong>Division specialists:</strong> BrainrotLM-Assistant-362M shows specific strength in division operations (38.9%) despite lower performance in other areas.</li>
</ul>
<div class="key-finding">
<h3>Research Implications</h3>
<p>These results challenge conventional wisdom about language model scaling laws for mathematical reasoning capabilities. They suggest that highly specialized, compact models might offer a more efficient approach for specific reasoning tasks compared to general-purpose large models.</p>
</div>
<h2>Conclusion</h2>
<p>This analysis demonstrates that extremely small language models can exhibit significant mathematical reasoning abilities, with models as small as 81.9M parameters showing the ability to solve basic arithmetic problems. The standout performer, MathGPT2 with only 81.9M parameters, achieved an impressive 42.0% accuracy on a diverse set of 100 mathematical questions.</p>
<p>These findings suggest that efficient architectural design and specialized training approaches may be more important than raw parameter count when optimizing for specific reasoning capabilities. This could have significant implications for resource-constrained applications where deploying massive models is impractical.</p>
<p>Future research directions could include investigating what specific architectural choices enable these compact models to perform mathematical operations, and how these insights might be applied to develop more efficient specialized models for other reasoning tasks.</p>
<div class="footer">
<p>Data analysis based on benchmark results for MathGPT2 (81.9M), aquif-moe-800m (800M), BrainrotLM-Assistant-362M (362M), and other models</p>
<p>© 2025 • Created for educational purposes</p>
</div>
</div>
<script>
// Get the modal
var modal = document.getElementById("imageModal");
// Get the image and insert it inside the modal
var zoomableImages = document.querySelectorAll(".chart img");
var modalImg = document.getElementById("zoomedImage");
// Add click event to all zoomable images
zoomableImages.forEach(function(image) {
image.onclick = function() {
modal.style.display = "flex";
modalImg.src = this.src;
};
});
// Get the <span> element that closes the modal
var span = document.getElementsByClassName("close")[0];
// When the user clicks on <span> (x), close the modal
span.onclick = function() {
modal.style.display = "none";
};
// Close the modal when clicking outside the image
modal.onclick = function(event) {
if (event.target === modal) {
modal.style.display = "none";
}
};
</script>
</body>
</html> |