Hugging Face's logo
Jeremie Tisby's picture
1 41

Jeremie Tisby

Frobenius
21world's profile picture
Ā·

AI & ML interests

None yet

Recent Activity

replied to lewtun's post about 18 hours ago
We outperform Llama 70B with Llama 3B on hard math by scaling test-time compute šŸ”„ How? By combining step-wise reward models with tree search algorithms :) We show that smol models can match or exceed the performance of their much larger siblings when given enough "time to think" We're open sourcing the full recipe and sharing a detailed blog post. In our blog post we cover: šŸ“ˆ Compute-optimal scaling: How we implemented DeepMind's recipe to boost the mathematical capabilities of open models at test-time. šŸŽ„ Diverse Verifier Tree Search (DVTS): An unpublished extension we developed to the verifier-guided tree search technique. This simple yet effective method improves diversity and delivers better performance, particularly at large test-time compute budgets. šŸ§­ Search and Learn: A lightweight toolkit for implementing search strategies with LLMs and built for speed with vLLM Here's the links: - Blog post: https://huggingface.co/spaces/HuggingFaceH4/blogpost-scaling-test-time-compute - Code: https://github.com/huggingface/search-and-learn Enjoy!
Reacted to lewtun's post with šŸ”„ about 18 hours ago
We outperform Llama 70B with Llama 3B on hard math by scaling test-time compute šŸ”„ How? By combining step-wise reward models with tree search algorithms :) We show that smol models can match or exceed the performance of their much larger siblings when given enough "time to think" We're open sourcing the full recipe and sharing a detailed blog post. In our blog post we cover: šŸ“ˆ Compute-optimal scaling: How we implemented DeepMind's recipe to boost the mathematical capabilities of open models at test-time. šŸŽ„ Diverse Verifier Tree Search (DVTS): An unpublished extension we developed to the verifier-guided tree search technique. This simple yet effective method improves diversity and delivers better performance, particularly at large test-time compute budgets. šŸ§­ Search and Learn: A lightweight toolkit for implementing search strategies with LLMs and built for speed with vLLM Here's the links: - Blog post: https://huggingface.co/spaces/HuggingFaceH4/blogpost-scaling-test-time-compute - Code: https://github.com/huggingface/search-and-learn Enjoy!
View all activity

Organizations

None yet

Frobenius's activity

liked 5 models 4 months ago
liked a dataset 4 months ago
liked a model 4 months ago
liked 8 models 8 months ago
liked 5 models 9 months ago