Steelskull commited on
Commit
c685d05
·
verified ·
1 Parent(s): 877988f

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +73 -1
README.md CHANGED
@@ -265,6 +265,67 @@ code {
265
  grid-template-columns: repeat(3, 1fr);
266
  gap: 15px;
267
  }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
268
  </style>
269
  <html lang="en">
270
  <head>
@@ -292,6 +353,17 @@ Nevoria = mix of EVA-LLAMA-0.1, EURYALE-v2.3, Anubis-v1, and Negative_LLAMA
292
  <p>The choice to use the lorablated model as a base was intentional - while it might seem counterintuitive, this approach creates unique interactions between the weights, similar to what was achieved in the original <a href="https://huggingface.co/Steelskull/L3-MS-Astoria-70b" target="_blank">Astoria model</a> and <a href="https://huggingface.co/Steelskull/L3.1-MS-Astoria-70b-v2" target="_blank">Astoria V2 model</a> . Rather than simply removing refusals, this "weight twisting" effect that occurs when subtracting the lorablated base model from the other models during the merge process creates an interesting balance in the final model's behavior. While this approach differs from traditional sequential application of components, it was chosen for its unique characteristics in the model's responses.</p>
293
  <h2>UGI-Benchmark Results:</h2>
294
  <div class="benchmark-container">
 
 
 
 
 
 
 
 
 
 
 
295
  <!-- Core Metrics -->
296
  <div class="metrics-section">
297
  <h3>Core Metrics</h3>
@@ -301,7 +373,7 @@ Nevoria = mix of EVA-LLAMA-0.1, EURYALE-v2.3, Anubis-v1, and Negative_LLAMA
301
  <span class="value">56.75</span>
302
  </div>
303
  <div class="metric-box">
304
- <span class="label">Wisdom Score</span>
305
  <span class="value">7.5/10</span>
306
  </div>
307
  <div class="metric-box">
 
265
  grid-template-columns: repeat(3, 1fr);
266
  gap: 15px;
267
  }
268
+
269
+ /* Notification styling */
270
+ .benchmark-notification {
271
+ background: rgba(255, 0, 255, 0.15);
272
+ border: 1px solid rgba(255, 0, 255, 0.3);
273
+ border-radius: 8px;
274
+ margin-bottom: 20px;
275
+ padding: 12px;
276
+ animation: glowPulse 2s infinite;
277
+ }
278
+
279
+ .notification-content {
280
+ display: flex;
281
+ align-items: center;
282
+ justify-content: center;
283
+ gap: 10px;
284
+ text-align: center;
285
+ }
286
+
287
+ .notification-icon {
288
+ font-size: 20px;
289
+ }
290
+
291
+ .notification-text {
292
+ color: #FFE1FF;
293
+ font-size: 16px;
294
+ font-weight: 500;
295
+ display: flex;
296
+ flex-direction: column;
297
+ align-items: center;
298
+ gap: 5px;
299
+ }
300
+
301
+ .benchmark-link {
302
+ color: #00FFFF;
303
+ text-decoration: none;
304
+ font-size: 14px;
305
+ padding: 4px 8px;
306
+ border-radius: 4px;
307
+ transition: all 0.3s ease;
308
+ border: 1px solid rgba(0, 255, 255, 0.3);
309
+ }
310
+
311
+ .benchmark-link:hover {
312
+ background: rgba(0, 255, 255, 0.1);
313
+ border-color: rgba(0, 255, 255, 0.5);
314
+ color: #00FFFF;
315
+ text-shadow: 0 0 5px rgba(0, 255, 255, 0.5);
316
+ }
317
+
318
+ @keyframes glowPulse {
319
+ 0% {
320
+ box-shadow: 0 0 5px rgba(255, 0, 255, 0.3);
321
+ }
322
+ 50% {
323
+ box-shadow: 0 0 15px rgba(255, 0, 255, 0.5);
324
+ }
325
+ 100% {
326
+ box-shadow: 0 0 5px rgba(255, 0, 255, 0.3);
327
+ }
328
+ }
329
  </style>
330
  <html lang="en">
331
  <head>
 
353
  <p>The choice to use the lorablated model as a base was intentional - while it might seem counterintuitive, this approach creates unique interactions between the weights, similar to what was achieved in the original <a href="https://huggingface.co/Steelskull/L3-MS-Astoria-70b" target="_blank">Astoria model</a> and <a href="https://huggingface.co/Steelskull/L3.1-MS-Astoria-70b-v2" target="_blank">Astoria V2 model</a> . Rather than simply removing refusals, this "weight twisting" effect that occurs when subtracting the lorablated base model from the other models during the merge process creates an interesting balance in the final model's behavior. While this approach differs from traditional sequential application of components, it was chosen for its unique characteristics in the model's responses.</p>
354
  <h2>UGI-Benchmark Results:</h2>
355
  <div class="benchmark-container">
356
+ <div class="benchmark-notification">
357
+ <div class="notification-content">
358
+ <span class="notification-icon">🏆</span>
359
+ <span class="notification-text">
360
+ Ranked 4th globally behind Llama 405B, Gemini-exp, and Chat-4o as of 1/17/2025
361
+ <a href="https://huggingface.co/spaces/DontPlanToEnd/UGI-Leaderboard" target="_blank" class="benchmark-link">
362
+ View Full Leaderboard →
363
+ </a>
364
+ </span>
365
+ </div>
366
+ </div>
367
  <!-- Core Metrics -->
368
  <div class="metrics-section">
369
  <h3>Core Metrics</h3>
 
373
  <span class="value">56.75</span>
374
  </div>
375
  <div class="metric-box">
376
+ <span class="label">Willingness Score</span>
377
  <span class="value">7.5/10</span>
378
  </div>
379
  <div class="metric-box">