WelcomeUser Guide
ToSPrivacyCanary
DonateBugsLicense

©2024 Poal.co

908
[–] 0 pt 4d

We introduce our first-generation reasoning models, DeepSeek-R1-Zero and DeepSeek-R1. DeepSeek-R1-Zero, a model trained via large-scale reinforcement learning (RL) without supervised fine-tuning (SFT) as a preliminary step, demonstrated remarkable performance on reasoning.

With RL, DeepSeek-R1-Zero naturally emerged with numerous powerful and interesting reasoning behaviors.

However, DeepSeek-R1-Zero encounters challenges such as endless repetition, poor readability, and language mixing.

To address these issues and further enhance reasoning performance, we introduce DeepSeek-R1, which incorporates cold-start data before RL. DeepSeek-R1 achieves performance comparable to OpenAI-o1 across math, code, and reasoning tasks.

To support the research community, we have open-sourced DeepSeek-R1-Zero, DeepSeek-R1, and six dense models distilled from DeepSeek-R1 based on Llama and Qwen.

DeepSeek-R1-Distill-Qwen-32B outperforms OpenAI-o1-mini across various benchmarks, achieving new state-of-the-art results for dense models.

[–] 1 pt 4d

That explains what the funding was used to do (cold start), but again, this problem was solved many years ago in fastai. https://medium.com/@jelaniwoods/fastai-lesson-4-collaborative-filtering-454064ffe0a2

[–] 0 pt 4d

Do you think DeepSeek is as jewed as OpenAI?

I checked its data cut-off and it seems to be October 2023.

[–] 1 pt 4d

It's jewed, that's for sure. I asked it some questions and it regurgitated (((answers))). The problem is that the English speaking world is so brainwashed all of the training data is kiked by default.

[–] 0 pt 4d

I doubt anything could be as jewed as openAI, but even if it's based, I guess what I'm saying is that it's old news being overblown, and both the US and China without a doubt have secret projects that are way beyond it.