DeepSeek open-sourced DeepSeek-R1, an LLM fine-tuned with reinforcement learning (RL) to improve thinking ability. DeepSeek-R1 attains results on par with OpenAI's o1 design on several standards, consisting of MATH-500 and SWE-bench.

DeepSeek-R1 is based upon DeepSeek-V3, a mixture of professionals (MoE) model just recently open-sourced by DeepSeek. This base design is fine-tuned utilizing Group Relative Policy Optimization (GRPO), a reasoning-oriented version of RL. The research group also performed knowledge distillation from DeepSeek-R1 to open-source Qwen and Llama models and launched several variations of each; these bigger designs, consisting of GPT-4, on math and coding benchmarks.

[DeepSeek-R1 is] the primary step toward improving language model thinking capabilities utilizing pure reinforcement learning (RL). Our objective is to check out the capacity of LLMs to develop thinking capabilities with no supervised data, focusing on their self-evolution through a pure RL process...DeepSeek-R1 ... excels in a vast array of jobs, consisting of imaginative writing, basic concern answering, modifying, summarization, and more. Additionally, DeepSeek-R1 shows outstanding performance on tasks requiring long-context understanding, substantially exceeding DeepSeek-V3 on long-context standards.

To establish the model, DeepSeek started with DeepSeek-V3 as a base. They initially tried fine-tuning it only with RL, and with no supervised fine-tuning (SFT), producing a model called DeepSeek-R1-Zero, which they have also released. This model shows strong reasoning efficiency, however" effective thinking behaviors, it faces several issues. For example, DeepSeek-R1-Zero struggles with difficulties like bad readability and language mixing."

To address this, the team utilized a short phase of SFT to prevent the "cold start" issue of RL. They collected several thousand examples of chain-of-thought reasoning to utilize in SFT of DeepSeek-V3 before running RL. After the RL procedure assembled, they then gathered more SFT information utilizing rejection sampling, leading to a dataset of 800k samples. This dataset was used for additional fine-tuning and to produce the distilled models from Llama and Qwen.

DeepSeek examined their design on a range of reasoning, mathematics, and coding standards and compared it to other models, including Claude-3.5- Sonnet, GPT-4o, pediascape.science and o1. DeepSeek-R1 exceeded all of them on numerous of the benchmarks, consisting of AIME 2024 and MATH-500.

DeepSeek-R1 Performance. Image Source: DeepSeek-R1 Technical Report

Within a couple of days of its release, the LMArena announced that DeepSeek-R1 was ranked # 3 total in the arena and # 1 in coding and mathematics. It was also connected for # 1 with o1 in "Hard Prompt with Style Control" category.

Django structure co-creator Simon Willison blogged about his experiments with one of the DeepSeek distilled Llama models on his blog site:

Each response starts with a ... pseudo-XML tag containing the chain of thought utilized to help produce the reaction. [Given the timely] "a joke about a pelican and a walrus who run a tea space together" ... It then believed for 20 paragraphs before outputting the joke! ... [T] he joke is terrible. But the process of arriving was such a fascinating insight into how these new designs work.

Andrew Ng's newsletter The Batch wrote about DeepSeek-R1:

DeepSeek is rapidly becoming a strong builder of open models. Not only are these models terrific entertainers, but their license permits use of their outputs for distillation, potentially pushing forward the state of the art for language designs (and multimodal designs) of all sizes.

The DeepSeek-R1 designs are available on HuggingFace.

About the Author

Anthony Alford

Rate this Article

This content remains in the AI, ML & Data Engineering subject

Related Topics:

- AI, ML & Data Engineering

• Generative AI
• Large language designs
  
  - Related Editorial
  
  Related Sponsored Content
  
  - [eBook] Getting Going with Azure Kubernetes Service
  
  Related Sponsor
  
  Free services for AI apps. Are you ready to experiment with advanced technologies? You can start constructing smart apps with totally free Azure app, information, and AI services to minimize upfront costs. Find out more.
  
  How could we enhance? Take the InfoQ reader survey
  
  Each year, we seek feedback from our readers to assist us enhance InfoQ. Would you mind spending 2 minutes to share your feedback in our brief survey? Your feedback will straight help us constantly develop how we support you. The InfoQ Team Take the survey
  
  Related Content
  
  The InfoQ Newsletter
  
  A round-up of last week's content on InfoQ sent out every Tuesday. Join a community of over 250,000 senior developers.