Hugging Face's logo

Spaces:
FlameF0X
/
Simple-Math-Benchmark
Running

App Files Community
FlameF0X commited on
Commit
edf774b
·
verified ·
1 Parent(s): 42011dd

Update index.html

Browse files
Files changed (1) hide show
  1. index.html +238 -193
index.html CHANGED
@@ -2,245 +2,290 @@
2
  <html lang="en">
3
  <head>
4
  <meta charset="UTF-8">
5
- <title>Model Benchmark: First Grade Math</title>
 
6
  <style>
7
- :root {
8
- --primary: #4a6fa5;
9
- --secondary: #6b8cbe;
10
- --accent: #ff6b6b;
11
- --background: #f5f7fa;
12
- --card-bg: #ffffff;
13
- --text-dark: #2d3748;
14
- --text-light: #4a5568;
15
- --success: #48bb78;
16
- --error: #e53e3e;
17
- }
18
-
19
  body {
20
- font-family: 'Segoe UI', Roboto, -apple-system, BlinkMacSystemFont, sans-serif;
21
- background: var(--background);
22
- margin: 0;
23
- padding: 0;
24
- color: var(--text-dark);
25
  line-height: 1.6;
26
- }
27
-
28
- .container {
29
  max-width: 1000px;
30
  margin: 0 auto;
31
- padding: 40px 20px;
 
32
  }
33
-
34
- header {
35
- background: linear-gradient(135deg, var(--primary), var(--secondary));
36
- color: white;
37
- padding: 30px 0;
38
- border-radius: 12px;
39
- margin-bottom: 30px;
40
- box-shadow: 0 4px 12px rgba(0,0,0,0.1);
41
  }
42
-
43
  h1 {
44
- margin: 0;
45
- font-size: 2.4rem;
46
- font-weight: 700;
 
 
 
47
  }
48
-
49
  h2 {
50
- color: var(--primary);
51
  font-size: 1.8rem;
52
- margin-top: 40px;
53
- margin-bottom: 20px;
54
- position: relative;
55
- padding-bottom: 10px;
56
- }
57
-
58
- h2::after {
59
- content: '';
60
- position: absolute;
61
- bottom: 0;
62
- left: 0;
63
- width: 60px;
64
- height: 4px;
65
- background: var(--accent);
66
- border-radius: 2px;
67
  }
68
-
69
- .chart-container {
70
- background: var(--card-bg);
71
- border-radius: 12px;
72
- padding: 30px;
73
- box-shadow: 0 8px 20px rgba(0,0,0,0.06);
74
- margin: 30px 0;
75
  }
76
-
77
- img {
78
- max-width: 100%;
79
- height: auto;
80
  border-radius: 8px;
81
- box-shadow: 0 4px 12px rgba(0,0,0,0.08);
 
82
  }
83
-
84
  .model-info {
85
- background: var(--card-bg);
86
- border-radius: 12px;
87
- padding: 30px;
88
- box-shadow: 0 8px 20px rgba(0,0,0,0.06);
89
- }
90
-
91
- .model-cards {
92
  display: flex;
93
- flex-wrap: wrap;
94
- gap: 20px;
95
- justify-content: center;
96
- margin-top: 20px;
97
- }
98
-
99
- .model-card {
100
- background: var(--card-bg);
101
- border-radius: 10px;
102
- padding: 20px;
103
- box-shadow: 0 4px 10px rgba(0,0,0,0.05);
104
- border-left: 5px solid var(--primary);
105
- width: calc(33% - 20px);
106
- min-width: 250px;
107
- transition: transform 0.2s, box-shadow 0.2s;
108
- }
109
-
110
- .model-card:hover {
111
- transform: translateY(-5px);
112
- box-shadow: 0 8px 15px rgba(0,0,0,0.1);
113
  }
114
-
115
  .model-name {
116
- font-weight: 700;
117
  font-size: 1.2rem;
118
- color: var(--primary);
119
- margin-bottom: 10px;
120
  }
121
-
122
- .stats {
 
 
 
 
 
 
 
 
 
 
 
 
 
123
  display: flex;
124
  justify-content: space-between;
125
- margin-top: 15px;
 
 
126
  }
127
-
128
- .stat {
129
- text-align: center;
130
  flex: 1;
 
 
 
 
 
131
  }
132
-
133
- .stat-value {
134
- font-size: 1.5rem;
135
- font-weight: 700;
136
  }
137
-
138
- .correct {
139
- color: var(--success);
 
 
140
  }
141
-
142
- .incorrect {
143
- color: var(--error);
 
144
  }
145
-
146
- .stat-label {
147
- font-size: 0.85rem;
148
- color: var(--text-light);
149
  }
150
-
151
- .accuracy-bar {
152
- height: 8px;
153
- background: #f0f0f0;
154
- border-radius: 4px;
155
- margin-top: 10px;
156
- overflow: hidden;
157
  }
158
-
159
- .accuracy-fill {
160
- height: 100%;
161
- background: var(--primary);
162
- border-radius: 4px;
163
  }
164
-
165
- footer {
 
 
166
  margin-top: 40px;
167
  text-align: center;
168
- color: var(--text-light);
169
  font-size: 0.9rem;
170
  }
 
 
 
 
 
 
 
 
 
 
 
171
  </style>
172
  </head>
173
  <body>
174
  <div class="container">
175
- <header>
176
- <h1>Model Benchmark: First Grade Math</h1>
177
- <p>Performance comparison across 1,000 questions per model</p>
178
- </header>
179
 
180
- <div class="chart-container">
181
- <h2>Performance Overview</h2>
182
- <img src="download.png" alt="Model Benchmark Chart">
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
183
  </div>
184
 
185
- <div class="model-info">
186
- <h2>Models Benchmarked</h2>
187
- <div class="model-cards">
188
- <div class="model-card">
189
- <div class="model-name">FlameF0X/MathGPT2</div>
190
- <div class="stats">
191
- <div class="stat">
192
- <div class="stat-value correct">763</div>
193
- <div class="stat-label">Correct</div>
194
- </div>
195
- <div class="stat">
196
- <div class="stat-value incorrect">237</div>
197
- <div class="stat-label">Incorrect</div>
198
- </div>
199
- </div>
200
- <div class="accuracy-bar">
201
- <div class="accuracy-fill" style="width: 76.3%"></div>
202
- </div>
203
- </div>
204
-
205
- <div class="model-card">
206
- <div class="model-name">FlameF0X/Muffin-2.9b-1C25</div>
207
- <div class="stats">
208
- <div class="stat">
209
- <div class="stat-value correct">9</div>
210
- <div class="stat-label">Correct</div>
211
- </div>
212
- <div class="stat">
213
- <div class="stat-value incorrect">991</div>
214
- <div class="stat-label">Incorrect</div>
215
- </div>
216
- </div>
217
- <div class="accuracy-bar">
218
- <div class="accuracy-fill" style="width: 0.9%"></div>
219
- </div>
220
- </div>
221
-
222
- <div class="model-card">
223
- <div class="model-name">FlameF0X/MuffinFace-2</div>
224
- <div class="stats">
225
- <div class="stat">
226
- <div class="stat-value correct">8</div>
227
- <div class="stat-label">Correct</div>
228
- </div>
229
- <div class="stat">
230
- <div class="stat-value incorrect">992</div>
231
- <div class="stat-label">Incorrect</div>
232
- </div>
233
- </div>
234
- <div class="accuracy-bar">
235
- <div class="accuracy-fill" style="width: 0.8%"></div>
236
- </div>
237
- </div>
238
  </div>
 
239
  </div>
 
 
240
 
241
- <footer>
242
- <p>Benchmark results as of April 2025</p>
243
- </footer>
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
244
  </div>
245
  </body>
246
  </html>
 
2
  <html lang="en">
3
  <head>
4
  <meta charset="UTF-8">
5
+ <meta name="viewport" content="width=device-width, initial-scale=1.0">
6
+ <title>Small But Mighty: Analyzing Compact Math Models</title>
7
  <style>
 
 
 
 
 
 
 
 
 
 
 
 
8
  body {
9
+ font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, Oxygen, Ubuntu, Cantarell, 'Open Sans', 'Helvetica Neue', sans-serif;
 
 
 
 
10
  line-height: 1.6;
11
+ color: #333;
 
 
12
  max-width: 1000px;
13
  margin: 0 auto;
14
+ padding: 20px;
15
+ background-color: #f9f9f9;
16
  }
17
+ h1, h2, h3, h4 {
18
+ color: #2d3748;
 
 
 
 
 
 
19
  }
 
20
  h1 {
21
+ font-size: 2.5rem;
22
+ margin-bottom: 2rem;
23
+ text-align: center;
24
+ color: #1a365d;
25
+ border-bottom: 3px solid #4299e1;
26
+ padding-bottom: 0.5rem;
27
  }
 
28
  h2 {
 
29
  font-size: 1.8rem;
30
+ margin-top: 2rem;
31
+ border-bottom: 1px solid #e2e8f0;
32
+ padding-bottom: 0.5rem;
 
 
 
 
 
 
 
 
 
 
 
 
33
  }
34
+ .container {
35
+ background-color: white;
36
+ border-radius: 8px;
37
+ padding: 20px;
38
+ box-shadow: 0 4px 6px rgba(0, 0, 0, 0.1);
39
+ margin-bottom: 20px;
 
40
  }
41
+ .model-card {
42
+ background-color: #f0f5ff;
43
+ padding: 15px;
 
44
  border-radius: 8px;
45
+ margin-bottom: 15px;
46
+ border-left: 4px solid #4299e1;
47
  }
 
48
  .model-info {
 
 
 
 
 
 
 
49
  display: flex;
50
+ justify-content: space-between;
51
+ margin-bottom: 10px;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
52
  }
 
53
  .model-name {
54
+ font-weight: bold;
55
  font-size: 1.2rem;
56
+ color: #2b6cb0;
 
57
  }
58
+ .model-params {
59
+ background-color: #ebf8ff;
60
+ padding: 3px 8px;
61
+ border-radius: 12px;
62
+ font-size: 0.9rem;
63
+ color: #2c5282;
64
+ }
65
+ .performance-highlight {
66
+ background-color: #e6fffa;
67
+ padding: 10px;
68
+ border-radius: 4px;
69
+ margin-top: 10px;
70
+ border-left: 3px solid #38b2ac;
71
+ }
72
+ .chart-container {
73
  display: flex;
74
  justify-content: space-between;
75
+ flex-wrap: wrap;
76
+ gap: 20px;
77
+ margin: 30px 0;
78
  }
79
+ .chart {
 
 
80
  flex: 1;
81
+ min-width: 300px;
82
+ background: white;
83
+ padding: 15px;
84
+ border-radius: 8px;
85
+ box-shadow: 0 2px 4px rgba(0, 0, 0, 0.05);
86
  }
87
+ .chart img {
88
+ width: 100%;
89
+ height: auto;
90
+ border-radius: 4px;
91
  }
92
+ .chart-caption {
93
+ text-align: center;
94
+ margin-top: 10px;
95
+ font-style: italic;
96
+ color: #718096;
97
  }
98
+ table {
99
+ width: 100%;
100
+ border-collapse: collapse;
101
+ margin: 20px 0;
102
  }
103
+ th, td {
104
+ padding: 12px 15px;
105
+ text-align: left;
106
+ border-bottom: 1px solid #e2e8f0;
107
  }
108
+ th {
109
+ background-color: #edf2f7;
110
+ font-weight: bold;
 
 
 
 
111
  }
112
+ tr:hover {
113
+ background-color: #f7fafc;
 
 
 
114
  }
115
+ .highlight {
116
+ background-color: #fffbeb;
117
+ }
118
+ .footer {
119
  margin-top: 40px;
120
  text-align: center;
121
+ color: #718096;
122
  font-size: 0.9rem;
123
  }
124
+ .key-finding {
125
+ background-color: #ebf8ff;
126
+ padding: 15px;
127
+ border-radius: 8px;
128
+ margin: 20px 0;
129
+ border-left: 4px solid #3182ce;
130
+ }
131
+ .key-finding h3 {
132
+ margin-top: 0;
133
+ color: #2b6cb0;
134
+ }
135
  </style>
136
  </head>
137
  <body>
138
  <div class="container">
139
+ <h1>Small But Mighty: Compact Mathematical Reasoning Models</h1>
 
 
 
140
 
141
+ <div class="key-finding">
142
+ <h3>Key Insight</h3>
143
+ <p>The data shows that extremely small language models (under 200M parameters) can demonstrate measurable mathematical abilities, challenging the assumption that large parameter counts are required for basic math reasoning.</p>
144
+ </div>
145
+
146
+ <h2>Model Overview</h2>
147
+ <p>This analysis compares the performance of various language models on mathematical problems, with a focus on very compact models:</p>
148
+
149
+ <div class="model-card">
150
+ <div class="model-info">
151
+ <span class="model-name">PingVortex/VLM-1</span>
152
+ <span class="model-params">124M parameters</span>
153
+ </div>
154
+ <p>The highest performer among tested models, demonstrating noteworthy mathematical abilities despite its tiny parameter count. Shows particular strength in division operations with 16.7% accuracy.</p>
155
+ <div class="performance-highlight">
156
+ <strong>Overall math accuracy:</strong> 6.9% on 1000 test questions
157
+ </div>
158
+ </div>
159
+
160
+ <div class="model-card">
161
+ <div class="model-info">
162
+ <span class="model-name">FlameF0X/MathGPT2</span>
163
+ <span class="model-params">81.9M parameters</span>
164
+ </div>
165
+ <p>The second best performer, scoring 2.2% overall accuracy. Shows more balanced performance across addition, subtraction and multiplication operations.</p>
166
+ <div class="performance-highlight">
167
+ <strong>Operation strength:</strong> 5.4% accuracy on multiplication
168
+ </div>
169
  </div>
170
 
171
+ <div class="model-card">
172
+ <div class="model-info">
173
+ <span class="model-name">aquiffoo/aquif-r1-1b</span>
174
+ <span class="model-params">1.14B parameters + CoT reasoning</span>
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
175
  </div>
176
+ <p>Despite being significantly larger and incorporating chain-of-thought reasoning capabilities, this model shows no measurable performance on the tested mathematical problems.</p>
177
  </div>
178
+
179
+ <h2>Performance Analysis</h2>
180
 
181
+ <div class="chart-container">
182
+ <div class="chart">
183
+ <img src="/api/placeholder/450/350" alt="Chart showing model accuracy by operation type">
184
+ <div class="chart-caption">Figure 1: Accuracy by Operation Type (%)</div>
185
+ </div>
186
+ <div class="chart">
187
+ <img src="/api/placeholder/450/350" alt="Chart showing model performance on math problems">
188
+ <div class="chart-caption">Figure 2: Correct vs Incorrect Answers (1000 questions each)</div>
189
+ </div>
190
+ </div>
191
+
192
+ <h2>Detailed Accuracy Results</h2>
193
+
194
+ <table>
195
+ <thead>
196
+ <tr>
197
+ <th>Model</th>
198
+ <th>Addition (+)</th>
199
+ <th>Subtraction (-)</th>
200
+ <th>Multiplication (*)</th>
201
+ <th>Division (/)</th>
202
+ <th>Exponentiation (**)</th>
203
+ <th>Overall</th>
204
+ </tr>
205
+ </thead>
206
+ <tbody>
207
+ <tr class="highlight">
208
+ <td>VLM-1 (124M)</td>
209
+ <td>2.6%</td>
210
+ <td>5.7%</td>
211
+ <td>5.4%</td>
212
+ <td>16.7%</td>
213
+ <td>0.0%</td>
214
+ <td>6.9%</td>
215
+ </tr>
216
+ <tr>
217
+ <td>MathGPT2 (81.9M)</td>
218
+ <td>0.8%</td>
219
+ <td>2.1%</td>
220
+ <td>5.4%</td>
221
+ <td>0.9%</td>
222
+ <td>0.0%</td>
223
+ <td>2.2%</td>
224
+ </tr>
225
+ <tr>
226
+ <td>aquif-r1-1b (1.14B)</td>
227
+ <td>0.0%</td>
228
+ <td>0.0%</td>
229
+ <td>0.0%</td>
230
+ <td>0.0%</td>
231
+ <td>0.0%</td>
232
+ <td>0.0%</td>
233
+ </tr>
234
+ <tr>
235
+ <td>BrainrotLM2-Assistant-362M</td>
236
+ <td>0.0%</td>
237
+ <td>0.0%</td>
238
+ <td>0.0%</td>
239
+ <td>0.0%</td>
240
+ <td>0.0%</td>
241
+ <td>0.0%</td>
242
+ </tr>
243
+ <tr>
244
+ <td>gonzalez-v1</td>
245
+ <td>0.0%</td>
246
+ <td>0.0%</td>
247
+ <td>0.0%</td>
248
+ <td>0.0%</td>
249
+ <td>0.0%</td>
250
+ <td>0.0%</td>
251
+ </tr>
252
+ <tr>
253
+ <td>Snowflake-G0-Release</td>
254
+ <td>0.0%</td>
255
+ <td>0.0%</td>
256
+ <td>0.0%</td>
257
+ <td>0.0%</td>
258
+ <td>0.0%</td>
259
+ <td>0.0%</td>
260
+ </tr>
261
+ </tbody>
262
+ </table>
263
+
264
+ <h2>Key Observations</h2>
265
+ <ul>
266
+ <li><strong>Size doesn't always matter:</strong> VLM-1 with only 124M parameters outperforms larger models on mathematical tasks.</li>
267
+ <li><strong>Operation specialization:</strong> VLM-1 shows particularly strong performance on division operations (16.7%).</li>
268
+ <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>
269
+ <li><strong>Zero performance:</strong> Four of the tested models showed no measurable mathematical ability on this test set.</li>
270
+ <li><strong>Chain-of-thought limitations:</strong> Despite having CoT capabilities, aquif-r1-1b did not demonstrate mathematical reasoning abilities in this benchmark.</li>
271
+ </ul>
272
+
273
+ <div class="key-finding">
274
+ <h3>Research Implications</h3>
275
+ <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>
276
+ </div>
277
+
278
+ <h2>Conclusion</h2>
279
+ <p>This analysis demonstrates that extremely small language models can exhibit measurable mathematical reasoning abilities, with models as small as 81.9M and 124M parameters showing the ability to solve basic arithmetic problems. The standout performer, VLM-1 with only 124M parameters, achieved nearly 7% accuracy on a diverse set of 1000 mathematical questions.</p>
280
+
281
+ <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>
282
+
283
+ <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>
284
+
285
+ <div class="footer">
286
+ <p>Data analysis based on benchmark results for VLM-1 (124M), MathGPT2 (81.9M), aquif-r1-1b (1.14B), and other models</p>
287
+ <p>© 2025 • Created for educational purposes</p>
288
+ </div>
289
  </div>
290
  </body>
291
  </html>