From d3f641072adad21910813bb912a8b439c4f81c11 Mon Sep 17 00:00:00 2001 From: Archie Hansell Date: Fri, 4 Apr 2025 02:39:49 +0200 Subject: [PATCH] Add 'DeepSeek Open-Sources DeepSeek-R1 LLM with Performance Comparable To OpenAI's O1 Model' --- ...R1-LLM-with-Performance-Comparable-To-OpenAI%27s-O1-Model.md | 2 ++ 1 file changed, 2 insertions(+) create mode 100644 DeepSeek-Open-Sources-DeepSeek-R1-LLM-with-Performance-Comparable-To-OpenAI%27s-O1-Model.md diff --git a/DeepSeek-Open-Sources-DeepSeek-R1-LLM-with-Performance-Comparable-To-OpenAI%27s-O1-Model.md b/DeepSeek-Open-Sources-DeepSeek-R1-LLM-with-Performance-Comparable-To-OpenAI%27s-O1-Model.md new file mode 100644 index 0000000..feb1d6d --- /dev/null +++ b/DeepSeek-Open-Sources-DeepSeek-R1-LLM-with-Performance-Comparable-To-OpenAI%27s-O1-Model.md @@ -0,0 +1,2 @@ +
DeepSeek open-sourced DeepSeek-R1, an LLM fine-tuned with [reinforcement learning](https://gitea.xiaolongkeji.net) (RL) to improve thinking ability. DeepSeek-R1 attains results on par with OpenAI's o1 design on a number of benchmarks, consisting of MATH-500 and SWE-bench.
+
DeepSeek-R1 is based on DeepSeek-V3, a mixture of professionals (MoE) model recently open-sourced by DeepSeek. This base model is fine-tuned utilizing Group Relative Policy Optimization (GRPO), a [reasoning-oriented](http://www.xn--9m1b66aq3oyvjvmate.com) version of RL. The research study group also carried out knowledge distillation from DeepSeek-R1 to [open-source](https://gitlab.rail-holding.lt) Qwen and Llama models and released several variations of each \ No newline at end of file