Original title, "SYNTHETIC-1: Scaling Distributed Synthetic Data Generation for Verified Reasoning," edited to fit HN's "title" length limit.
From their blog post dated Feb. 6:
"Today, we are excited to introduce SYNTHETIC-1, a collaborative effort to create the largest open-source dataset of verified reasoning traces for math, coding and science, leveraging DeepSeek-R1. Our dataset consists of 1.4 million high-quality tasks and verifiers, designed to advance reasoning model training.
We invite everyone to contribute compute and join us in our effort to scale distributed reinforcement learning to o3-scale and beyond.
In our recent post, Decentralized Training in the Inference-Time-Compute Paradigm, we explored how this paradigm shift will fundamentally reshape compute infrastructure, making globally distributed training the way forward.
The DeepSeek-R1 paper highlights the importance of generating cold-start synthetic data for RL. As our first step toward state-of-the-art reasoning models, SYNTHETIC-1 generates verified reasoning traces across math, coding, and science using DeepSeek-R1."
1 comment
[ 2.0 ms ] story [ 11.3 ms ] threadFrom their blog post dated Feb. 6:
"Today, we are excited to introduce SYNTHETIC-1, a collaborative effort to create the largest open-source dataset of verified reasoning traces for math, coding and science, leveraging DeepSeek-R1. Our dataset consists of 1.4 million high-quality tasks and verifiers, designed to advance reasoning model training.
We invite everyone to contribute compute and join us in our effort to scale distributed reinforcement learning to o3-scale and beyond.
In our recent post, Decentralized Training in the Inference-Time-Compute Paradigm, we explored how this paradigm shift will fundamentally reshape compute infrastructure, making globally distributed training the way forward.
The DeepSeek-R1 paper highlights the importance of generating cold-start synthetic data for RL. As our first step toward state-of-the-art reasoning models, SYNTHETIC-1 generates verified reasoning traces across math, coding, and science using DeepSeek-R1."