GitLab

DeepSeek open-sourced DeepSeek-R1, an LLM fine-tuned with reinforcement learning (RL) to enhance reasoning capability. DeepSeek-R1 attains outcomes on par with OpenAI's o1 model on a number of benchmarks, including MATH-500 and SWE-bench.

DeepSeek-R1 is based on DeepSeek-V3, a mix of experts (MoE) model recently open-sourced by DeepSeek. This base model is fine-tuned utilizing Group Relative Policy Optimization (GRPO), a reasoning-oriented variant of RL. The research team likewise carried out understanding distillation from DeepSeek-R1 to open-source Qwen and Llama models and launched numerous versions of each; these models outperform larger designs, including GPT-4, on math and coding criteria.

[DeepSeek-R1 is] the first step towards improving language model thinking abilities using pure reinforcement knowing (RL). Our goal is to check out the potential of LLMs to develop thinking abilities without any monitored information, focusing on their self-evolution through a pure RL process...DeepSeek-R1 ... master a vast array of jobs, including creative writing, general question answering, modifying, summarization, and more. Additionally, DeepSeek-R1 demonstrates impressive performance on tasks needing long-context understanding, substantially outperforming DeepSeek-V3 on long-context benchmarks.

To establish the model, DeepSeek started with DeepSeek-V3 as a base. They first tried fine-tuning it only with RL, and with no monitored fine-tuning (SFT), producing a model called DeepSeek-R1-Zero, which they have also released. This design displays strong thinking performance, however" powerful reasoning behaviors, it faces numerous issues. For example, DeepSeek-R1-Zero fights with challenges like bad readability and language mixing."

To address this, the team used a short phase of SFT to prevent the "cold start" problem of RL. They gathered several thousand examples of chain-of-thought thinking to use in SFT of DeepSeek-V3 before running RL. After the RL procedure assembled, they then collected more SFT information utilizing rejection tasting, resulting in 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 variety of reasoning, mathematics, and coding criteria and compared it to other designs, consisting of Claude-3.5- Sonnet, GPT-4o, and o1. DeepSeek-R1 exceeded all of them on several of the standards, consisting of AIME 2024 and MATH-500.

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

Within a few days of its release, the LMArena announced that DeepSeek-R1 was ranked # 3 overall in the arena and # 1 in coding and math. It was likewise tied for # 1 with o1 in "Hard Prompt with Style Control" classification.

Django framework co-creator Simon Willison composed about his try outs one of the DeepSeek distilled Llama designs on his blog:

Each reaction starts with a ... pseudo-XML tag containing the chain of thought used to assist generate the action. [Given the prompt] "a joke about a pelican and a walrus who run a tea room together" ... It then thought for 20 paragraphs before outputting the joke! ... [T] he joke is horrible. But the process of getting there was such an intriguing insight into how these brand-new models work.

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

DeepSeek is rapidly becoming a strong contractor of open models. Not just are these designs fantastic entertainers, but their license allows usage of their outputs for distillation, possibly pushing forward the state of the art for language designs (and multimodal models) 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 models
  
  - Related Editorial
  
  Related Sponsored Content
  
  - [eBook] Getting Started with Azure Kubernetes Service
  
  Related Sponsor
  
  Free services for AI apps. Are you ready to try out innovative technologies? You can start building smart apps with complimentary 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 look for feedback from our readers to assist us improve InfoQ. Would you mind costs 2 minutes to share your feedback in our brief survey? Your feedback will straight assist us continuously evolve how we support you. The Take the survey
  
  Related Content
  
  The InfoQ Newsletter
  
  A round-up of recently's content on InfoQ sent out every Tuesday. Join a neighborhood of over 250,000 senior wiki.dulovic.tech developers.