Mulberry: Empowering MLLM with o1-like Reasoning and Reflection via Collective Monte Carlo Tree Search Yao et al.: https://
arxiv.org/abs/2412.18319 #ArtificialIntelligence #DeepLearning #MachineLearning
LLMS
-
Mulberry: O1-Like Reasoning for Multimodal Large Language Models
By
–
-
Elastic AI Assistant Evolution: From RAG to Agentic Capabilities
By
–
Has been cool to see Elastic AI Assistant evolve from RAG to be more agentic over the past year I particularly like their work on "ES|QL query generation" – using LLMs to generate a particular DSL is a pretty good use case imo
-
OpenAI’s Packaging and Deployment Moat Exceeds Competition
By
–
The moat is in packaging + deployment – this is where OAI by far exceeds by miles, with Anthropic coming second. Loads of work to do to get there for open source, 2025 will be critical!
-
Chapter 3 Technical Complexity and Multi-Head Attention Implementation
By
–
It say Ch 3 might be the most technical one (like building the engine of a car) but it gets easier from here! You were wondering about some of the design choices. The implementation follows the original, popular, and widely used multi-head attention one (so we can also load
-
Improving LLM Reasoning: Test-Time Compute Scaling Over Parameters
By
–
And since o1 and the upcoming o3 are such important topics these days, I wanted to share a few good intros into improving the reasoning capabilities of LLMs I read earlier this year: 1. "Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameters"
-
Open-Weight Reasoning Models Development Predictions for 2025
By
–
It will take time, but as both a wish and a prediction for 2025, I hope to see more interesting approaches to developing open-weight (o1 and o3-like) reasoning models through improved post-training recipes, data, and inference compute.
-
Understanding Mixture of Experts Architecture and FFN Weights
By
–
Based on how MoEs work, I believe this is possible. Each expert is like a FFN with it's own weights.
-
Qwen 2.5 0.5B: The Sweet Spot for Efficient Language Models
By
–
Qwen 2.5 0.5B is the sweet spot, but you do you
-
Try All State-of-the-Art LLMs Free for One Month
By
–
Try all SOTA LLMs in one place – DeepSeekV3
– o1/GPT-4o
– Sonnet 3.5
– Gemini Flash 2.0 FREE for a month: -
Understanding Mixture of Experts: How AI Models Work Together
By
–
DeepSeekV3, Gemini, Mixtral and many others are all Mixture of Experts (MoEs). But what exactly are MoEs? A Mixture of Experts (MoE) is a machine learning framework that resembles a team of specialists, each adept at handling different aspects of a complex task. It's like