Article
From Zero to Reasoning Hero: How DeepSeek-R1 Leverages Reinforcement Learning to Master Complex Reasoning
By
•
•
6Sugato Ray PRO
sugatoray
AI & ML interests
None yet
Recent Activity
upvoted
an
article
about 9 hours ago
From Zero to Reasoning Hero: How DeepSeek-R1 Leverages Reinforcement Learning to Master Complex Reasoning
updated
a collection
about 12 hours ago
AV LLMs
liked
a model
about 12 hours ago
Zyphra/Zonos-v0.1-transformer
Organizations
sugatoray's activity
upvoted
an
article
about 9 hours ago
updated
a
collection
about 12 hours ago
liked
a
model
about 12 hours ago
updated
a
model
about 12 hours ago
Update README with color formatted Python code block
#11 opened about 12 hours ago
by
sugatoray
updated
a
collection
about 12 hours ago
liked
a
model
about 12 hours ago
upvoted
an
article
about 13 hours ago
Article
Open R1: Update #2
By
and 6 others
•
•
121
updated
2
collections
about 13 hours ago
liked
a
dataset
about 13 hours ago
reacted to
lewtun's
post with ❤️
about 13 hours ago
Post
2625
Introducing OpenR1-Math-220k!
open-r1/OpenR1-Math-220k
The community has been busy distilling DeepSeek-R1 from inference providers, but we decided to have a go at doing it ourselves from scratch 💪
What’s new compared to existing reasoning datasets?
♾ Based on AI-MO/NuminaMath-1.5: we focus on math reasoning traces and generate answers for problems in NuminaMath 1.5, an improved version of the popular NuminaMath-CoT dataset.
🐳 800k R1 reasoning traces: We generate two answers for 400k problems using DeepSeek R1. The filtered dataset contains 220k problems with correct reasoning traces.
📀 512 H100s running locally: Instead of relying on an API, we leverage vLLM and SGLang to run generations locally on our science cluster, generating 180k reasoning traces per day.
⏳ Automated filtering: We apply Math Verify to only retain problems with at least one correct answer. We also leverage Llama3.3-70B-Instruct as a judge to retrieve more correct examples (e.g for cases with malformed answers that can’t be verified with a rules-based parser)
📊 We match the performance of DeepSeek-Distill-Qwen-7B by finetuning Qwen-7B-Math-Instruct on our dataset.
🔎 Read our blog post for all the nitty gritty details: https://huggingface.co/blog/open-r1/update-2
open-r1/OpenR1-Math-220k
The community has been busy distilling DeepSeek-R1 from inference providers, but we decided to have a go at doing it ourselves from scratch 💪
What’s new compared to existing reasoning datasets?
♾ Based on AI-MO/NuminaMath-1.5: we focus on math reasoning traces and generate answers for problems in NuminaMath 1.5, an improved version of the popular NuminaMath-CoT dataset.
🐳 800k R1 reasoning traces: We generate two answers for 400k problems using DeepSeek R1. The filtered dataset contains 220k problems with correct reasoning traces.
📀 512 H100s running locally: Instead of relying on an API, we leverage vLLM and SGLang to run generations locally on our science cluster, generating 180k reasoning traces per day.
⏳ Automated filtering: We apply Math Verify to only retain problems with at least one correct answer. We also leverage Llama3.3-70B-Instruct as a judge to retrieve more correct examples (e.g for cases with malformed answers that can’t be verified with a rules-based parser)
📊 We match the performance of DeepSeek-Distill-Qwen-7B by finetuning Qwen-7B-Math-Instruct on our dataset.
🔎 Read our blog post for all the nitty gritty details: https://huggingface.co/blog/open-r1/update-2
upvoted
an
article
about 13 hours ago
Article
Fine-tune Deepseek-R1 with a Synthetic Reasoning Dataset
By
•
•
21
upvoted
an
article
2 days ago
Article
How to deploy and fine-tune DeepSeek models on AWS
•
40
updated
a
collection
2 days ago
upvoted
a
paper
2 days ago
upvoted
an
article
3 days ago
Article
DABStep: Data Agent Benchmark for Multi-step Reasoning
•
42
updated
a
collection
3 days ago
upvoted
a
paper
3 days ago
updated
a
collection
3 days ago