Author name: Winston

Evaluating the Mathematical Reasoning Capabilities of Large Language Models: Limitations and Challenges

Leave a Comment / Paper Skimming / By

LLMs have made remarkable progress in various fields, including natural language processing, question answering, and creative tasks, even demonstrating the ability to solve mathematical problems. Recently, OpenAI’s o1 model, which uses CoT (Chain of Thought), has shown significant reasoning capabilities. However, for a long time, the commonly used GSM8K dataset has had a fixed set of questions

Evaluating the Mathematical Reasoning Capabilities of Large Language Models: Limitations and Challenges Read More »

Paper Skimming, , , ,

Unveiling AlphaFold 3: The Next Leap in Predicting Biomolecular Structures Across the Chemical Space

Leave a Comment / Research Highlights / By

On October 9, 2024, the Royal Swedish Academy of Sciences decided to award half of the 2024 Nobel Prize in Chemistry to Demis Hassabis and John Jumper for their development of AlphaFold2 in 2020, a model capable of predicting the structure of almost all 200 million proteins discovered by researchers. Here is the official scientific background: They have revealed proteins’ secrets through

Unveiling AlphaFold 3: The Next Leap in Predicting Biomolecular Structures Across the Chemical Space Read More »

Research Highlights, , , ,

Nobel Prize in Physics 2024

Leave a Comment / Posts / By

The Nobel Prize in Physics being awarded to pioneers in the field of machine learning may seem odd at first glance, but it is, in fact, well-deserved. If one merely thinks that today’s artificial intelligence is achieved through computers, it’s natural to find this decision hard to understand. However, at its core, the two scientists

Nobel Prize in Physics 2024 Read More »

Posts, , , ,

Beginner’s guide to fine-tuning models using MLX on Apple Silicon.

Leave a Comment / Tutorial / By

This article is also available in Simplified Chinese. Popular Python fine-tuning packages for large language models (LLMs), such as Unsloth and Lamini, do not support GPU acceleration on Apple M-series chips. Using MLX for fine-tuning on Mac with Apple Silicon is a great alternative. MLX is a machine learning framework developed by Apple, specifically optimized for

Beginner’s guide to fine-tuning models using MLX on Apple Silicon. Read More »

Tutorial, , ,

The technology itself is neither right nor wrong. Cal Gov. vetos SB 1047

Leave a Comment / Posts / By

The technology itself is neither right nor wrong. If legislation can be enacted to restrict its application, similar to nuclear technology, then tracking should be strengthened at the hardware level, rather than restricting the development of the technology itself and hindering innovation. https://www.engadget.com/ai/california-gov-newsom-vetoes-bill-sb-1047-that-aims-to-prevent-ai-disasters-220826827.html

The technology itself is neither right nor wrong. Cal Gov. vetos SB 1047 Read More »

Posts, ,

MLX Framework FAQ Explained: Model Support, Fine-Tuning, Conversion, and MLX Community

Leave a Comment / Tutorial / By

1. What machine learning models does MLX support? The MLX framework supports a variety of popular machine learning and deep learning models, primarily including large language models (LLM) and text generation models, such as LLaMA, Mistral, Phi-2, and Qwen; image generation models like Stable Diffusion; speech recognition models such as OpenAI’s Whisper; and models for

MLX Framework FAQ Explained: Model Support, Fine-Tuning, Conversion, and MLX Community Read More »

Tutorial, ,

Balancing AI Reasoning Power and Efficiency

Leave a Comment / Thoughts and Reflections / By

The Crucial Challenge of OpenAI’s o1 Model And All of Us Over the past day, countless people have excitedly participated in testing OpenAI’s o1. The release of the new model has undoubtedly become an internet sensation. The leaks and hints surrounding “Strawberry” in the early stages, the increasingly competitive market environment, and the public’s focus

Balancing AI Reasoning Power and Efficiency Read More »

Thoughts and Reflections, ,

Revolutionizing Protein Research with ESM3

Leave a Comment / Research Highlights / By

Traditional biological research, often characterized by labor-intensive experiments, struggles to reveal the intricate mechanisms behind protein folding and function. The advent of large neural networks offers a transformative approach by uncovering hidden patterns and making accurate predictions, particularly in protein biology—life’s fundamental code. Biology is fundamentally programmable. Every living organism shares the same genetic code

Revolutionizing Protein Research with ESM3 Read More »

Research Highlights, , , ,
pexels-photo-28608151-28608151.jpg

Grade-School Math and the Hidden Reasoning Process

Leave a Comment / Research Highlights / By

Currently, models like OpenAI’s GPT, Anthropic’s Claude, and Meta AI’s LLaMA have achieved over 90% accuracy on the GSM8K dataset. But how do they accomplish this? Is it through memorization of data and problems, or do they truly understand the content of the questions? GSM8K, short for “Grade School Math 8K,” comprises 8,000 math problems

Grade-School Math and the Hidden Reasoning Process Read More »

Research Highlights, , , ,
Crop chemist holding in hands molecule model

Curiosity and Hands-On Exploration: Fueling Innovation Through Deep Understanding

Leave a Comment / Thoughts and Reflections / By

Seeing Adam Majmudar meticulously build his TinyGPU reminds me of my own experience with manual PCR in the lab. Back when I was working on molecular biology research, our lab faced a severe shortage of PCR machines—two had broken down, and two others were stuck in customs. In those desperate days, with no machine available,

Curiosity and Hands-On Exploration: Fueling Innovation Through Deep Understanding Read More »

Thoughts and Reflections, , ,
Scroll to Top