@correctbrain_ profile picture
CorrectBrain @correctbrain_

Agentic System | Visual Explainers | EBooks | Graphic Guides| Tikz | Animations | LLMs act as the Designers |

Joined December 2023
  • Tweets 41
  • Following 0
  • Followers 52
38 Photos and videos
Pinned Tweet
CorrectBrain
@correctbrain_
2h
x.com/i/article/206622043098…
1
29
Data Flow Through the Original Transformer Architecture... Step-by-Step Execution Trace with Example English-to-French Translation.,,
1
1
2
2,137
Why height: auto; is Strictly Necessary... The complete image is generated in TikZ by a multi-agent system, requiring only 0% to 5% human involvement, which our tools make simple and efficient. @emollick @hxiao
3
1,125
Neural Network Layers: The Output Layer your goal dictates the output layer's size and activation function...
1
4
844
Loss Function: Cross-Entropy. Multi-Agent Generation of a Complete Image in TikZ.. Designed by an LLM and created in TikZ,, for cat and dog image we use nano banana pro..@SebastienBubeck
1
1
1
1,500
CSS Grid: Names and Placements...
1
4
684
The Holy Grail Layout...
1
1
1,579
The Generalization Test: Performance on Unseen Data...
3
548
Principal Component Analysis (PCA) In Machine Learning...
2
345
Framing ML Problems as Linear Equations..
1
315
principal component analysis in machine learning.. Created in TikZ with LLM..
1
1
201
Understanding the Scaled Dot-Product mathematically and visually.... Understanding the Scaled Dot-Product Attention in LLMs and preventing the ”Vanishing Gradient” problem....
1
1
12
6,759
How WO Mixes Information in Multi-Head Attention... The WO (Output Weight) matrix is the ”Blender”. It takes isolated, specialized features from different attention heads and merges them back into a single, context-rich unified representation. Transforming independent, isolated insights from h attention heads into a unified, rich representation.. #Transformer #LLMs,
1
8
4,980
Transformer Model in Nlp part 7.... The limitation of a single attention head... Why is a single attention head insufficient for capturing the complex nuances of language?...
1
9
4,313
CSS Part 6... Why is height: auto strictly necessary when max-width is applied, and what happens to the image distortion if it is omitted?... #correctbrain #WebDevelopment
2
828
JavaScript part 1... How do the web API, call stack, and task queues fit into the event loop?.... The "Stack → Web API → Queue" flow is true for Macrotasks (like timers and network requests). However, Microtasks (like Promises) usually skip the Web API "waiting room" and go straight from the Stack to the Microtask Queue....
5
330
Rust Memory Safety...part 1... Achieving Safety via Static Analysis (Ownership & Borrowing)... #rust #correctbrain
6
567
Low-Rank Adaptation (LoRA) in LLMs Comparing Full Rank Update ∆W vs. Low Rank Decomposition BA correctbrain.com/ #correctbrain #nlp #llm
1
2
7
4,762
CSS Part 5... Why use Donut Scope? Part 1... It allows precise targeting of ”slices” of the DOM. You can style a component wrapper without accidentally breaking nested components (like a 3rd party widget inside your .content) that might use similar tag names (like <img>)...,,
2
592
Load more