DeepSeek-OCR Notes
This paper is not simply about using OCR to read a document. Rather, it treats OCR as a proxy task for studying vision-text compression. It does not claim that vision tokens can fully replace text …
Agentic Ai
This section groups the core model and system techniques that support agentic workflows, including language models, retrieval augmentation, and transformer-based architectures.
Articles
Effective context engineering for AI agents
Context engineering is becoming increasingly important. For modern LLMs, the overall context matters more than any single word or phrase. Context refers to all of the tokens included in the model w…
Retrieval-Augmented Generation (RAG) — Notes
- Large pre-trained language models (e.g., BART, T5) store factual knowledge in parameters. - Limitations of parametric-only models: - Knowledge is difficult to update. - Lack of interpretability. …
AN IMAGE IS WORTH 16X16 WORDS: Transformers For Image Recognition At Scale
This study shows that the reliance on CNN (Convolutional Neural Network) is unnecessary for computer vision. Vit (Vision Transformer) attains excellent results when pre-trained on large amounts of …
Attention is All You Need Paper Note
paper: https://arxiv.org/pdf/1706.03762 The dominant sequence transduction models are based on complex recurrent or convolutional neural networks that include an encoder and a decoder. The best per…