1. Fix Thai PDF Github

30/08/2023

The new Fix Thai PDF by Wannaphong can fix ราชกิจานุเบกษา PDF text.

Untitled

Link: wannaphong/fixthaipdf: Fix Thai PDF (github.com)

2. Arabic LLM

30/08/2023

Untitled

Link: https://twitter.com/G42ai/status/1696773250162786377/photo/1

Technical Paper: [2308.16149] Jais and Jais-chat: Arabic-Centric Foundation and Instruction-Tuned Open Generative Large Language Models (arxiv.org)

White Paper: Whitepaper aLLM (inceptioniai.org)

Models: JAIS (base), JAIS-Chat (instruction finetuning)

Highlight: SOTA Arabicm, and comparable to Llama2 in English task, Jais-chat

Model: ALiBi Positional Encodings, MuP, SwiGLU, and Equal Importance English-Arabic tokens

Untitled

Training Datasets:

Untitled

Untitled