Update NVFP4 blog sharing links for improved clarity and engagement

- Revised Twitter and LinkedIn sharing links to feature a more engaging article title and summary.
- Enhanced the content to better reflect the focus on NVIDIA's NVFP4 and its benefits for training LLMs with 4-bit precision.

Author: vukrosic

app/blog/pretrain-llm-with-nvfp4/page.tsx

@@ -303,15 +303,15 @@ export default function NVFP4Project() {
                       </div>
                     </div>
 
-                    <a href="https://x.com/intent/tweet?text=I%20wrote%20an%20article%20about%20NVIDIA%27s%20NVFP4%20-%20how%20to%20train%20LLMs%20with%204-bit%20precision%20%F0%9F%9A%80%0A%0A2-3x%20faster%20training%2C%2050%25%20less%20memory.%20Covers%20architecture%2C%20implementation%2C%20and%20benchmarks.%0A%0A%23AI%20%23MachineLearning%20%23DeepLearning&url=https://opensuperintelligencelab.com/blog/pretrain-llm-with-nvfp4/" 
+                    <a href="https://x.com/intent/tweet?text=Check%20out%20this%20article%20about%20NVIDIA%27s%20NVFP4%20-%20how%20to%20train%20LLMs%20with%204-bit%20precision%20%F0%9F%9A%80%0A%0A2-3x%20faster%20training%2C%2050%25%20less%20memory.%20Covers%20architecture%2C%20implementation%2C%20and%20benchmarks.%0A%0A%23AI%20%23MachineLearning%20%23DeepLearning&url=https://opensuperintelligencelab.com/blog/pretrain-llm-with-nvfp4/" 
                        target="_blank" 
                        rel="noopener noreferrer"
                        className="text-slate-400 hover:text-green-400 transition-colors">
                       <svg className="w-5 h-5" fill="currentColor" viewBox="0 0 24 24">
                         <path d="M18.244 2.25h3.308l-7.227 8.26 8.502 11.24H16.17l-5.214-6.817L4.99 21.75H1.68l7.73-8.835L1.254 2.25H8.08l4.713 6.231zm-1.161 17.52h1.833L7.084 4.126H5.117z"/>
                       </svg>
                     </a>
-                    <a href="https://www.linkedin.com/sharing/share-offsite/?url=https://opensuperintelligencelab.com/blog/pretrain-llm-with-nvfp4/&title=I%20wrote%20an%20article%20about%20NVIDIA%20NVFP4%3A%204-Bit%20LLM%20Training&summary=I%20wrote%20a%20technical%20article%20about%20NVIDIA%27s%20NVFP4%20for%204-bit%20LLM%20training.%20Achieves%202-3x%20faster%20training%20speeds%20and%2050%25%20memory%20reduction.%20Includes%20architecture%20details%2C%20implementation%20guide%2C%20and%20benchmarks." 
+                    <a href="https://www.linkedin.com/sharing/share-offsite/?url=https://opensuperintelligencelab.com/blog/pretrain-llm-with-nvfp4/&title=Check%20out%20this%20article%20about%20NVIDIA%20NVFP4%3A%204-Bit%20LLM%20Training&summary=Check%20out%20this%20technical%20article%20about%20NVIDIA%27s%20NVFP4%20for%204-bit%20LLM%20training.%20Achieves%202-3x%20faster%20training%20speeds%20and%2050%25%20memory%20reduction.%20Includes%20architecture%20details%2C%20implementation%20guide%2C%20and%20benchmarks." 
                        target="_blank" 
                        rel="noopener noreferrer"
                        className="text-slate-400 hover:text-green-400 transition-colors">

public/content/pretrain-llm-with-nvfp4/pretrain-llms-with-fp4-content.md

@@ -36,7 +36,7 @@ The presence of outliers - ex. `50` in `[0.5, -0.2, 1.1, -0.8, 50.0]` - is a maj
 -   **The Outlier:** The value `50.0` is a significant outlier.
 -   **Scaling:** To quantize this block into FP4, which has a maximum representable value of `6.0`, we must scale every number down with the same scaling factor. `Scale Factor = 6.0 / 50.0 = 0.12`. The scaling factor is based on the largest number (taken absolute value)
 -   **Result:** After scaling (multiplying), our block becomes `[0.06, -0.024, 0.132, -0.096, 6.0]`.
-Here, the values that can be represented in FP4 (E2M1) are ±0, ±0.5, ±1, ±1.5, ±2, ±3, ±4, and ±6. This means that after scaling, any value in the block will be rounded to the nearest of these discrete numbers. The representable range is thus from -6 to +6, with only these specific values available.
+Here, the values that can be represented in FP4 (E2M1) are`±0, ±0.5, ±1, ±1.5, ±2, ±3, ±4, and ±6`. This means that after scaling, any value in the block will be rounded to the nearest of these discrete numbers. The representable range is thus from -6 to +6, with only these specific values available.
 -   **Information Loss:** When these new values are converted to the closest representable FP4 number (e.g., `±0`, `±0.5`, `±1.0`...), the first four values are so small that they will likely all be rounded to zero. The original information they contained is lost. Only the outlier retains its significance. NVFP4's techniques are designed to mitigate exactly this problem.
 
 ## Technical Advantages of NVFP4