Understanding Papers

Understand every DeepSeek paper from first principles, in order.

0 / 35 papers complete

Complete = every question answered · 0 papers read

Start from the first paper

Founding

5 papers

The first DeepSeek LLMs, code, and math models: dense, scale-first, and already competitive.

01PaperCore
DeepSeek LLM: Scaling Open-Source Language Models with Longtermism
The 7B and 67B base models that opened the DeepSeek line, configured by scaling laws the authors re-measured from scratch.
2024 · 15 min theory · 6 questions
Not started
02PaperCore
DeepSeekMoE: Towards Ultimate Expert Specialization in Mixture-of-Experts Language Models
Split every expert into smaller pieces and pin a few experts as always-on, so parameters grow without compute growth and each expert stays specialized.
2024 · 14 min theory · 5 questions
Not started
03PaperAdvanced
DeepSeek-Coder: When the Large Language Model Meets Programming - The Rise of Code Intelligence
A code model family trained from scratch on 2 trillion tokens, with repository structure and fill-in-the-middle objectives built into pretraining.
2024 · 12 min theory · 5 questions
Not started
04PaperCore
DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models
A 7B math model built on web data, plus GRPO (Group Relative Policy Optimization): the critic-free RL method that later powered DeepSeek's reasoning era.
2024 · 16 min theory · 6 questions
Not started
05PaperAdvanced
DeepSeek-VL: Towards Real-World Vision-Language Understanding
A vision-language family whose hybrid encoder reads 1024 x 1024 images into 576 tokens, trained so the language model never forgets how to read.
2024 · 13 min theory · 5 questions
Not started

Efficiency

9 papers

Mixture-of-experts routing, latent attention, and low-precision training: frontier quality at a fraction of the cost.

Reasoning

8 papers

Reinforcement learning with verifiable rewards: models taught to think before they answer.

Frontier

13 papers

Sparse attention, multimodal decoupling, and hybrid thinking at the current edge of the series.

Two passes per paper: theory first, built from first principles, then the paper itself with an implementation check. Answers and read marks save on this device and travel with your backup.

Understanding Papers

Founding

DeepSeek LLM: Scaling Open-Source Language Models with Longtermism

DeepSeekMoE: Towards Ultimate Expert Specialization in Mixture-of-Experts Language Models

DeepSeek-Coder: When the Large Language Model Meets Programming - The Rise of Code Intelligence

DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models

DeepSeek-VL: Towards Real-World Vision-Language Understanding

Efficiency

DeepSeek-V2: A Strong, Economical, and Efficient Mixture-of-Experts Language Model

DeepSeek-Coder-V2: Breaking the Barrier of Closed-Source Models in Code Intelligence

Let the Expert Stick to His Last: Expert-Specialized Fine-Tuning for Sparse Architectural Large Language Models

Fire-Flyer AI-HPC: A Cost-Effective Software-Hardware Co-Design for Deep Learning

Auxiliary-Loss-Free Load Balancing Strategy for Mixture-of-Experts

DeepSeek-VL2: Mixture-of-Experts Vision-Language Models for Advanced Multimodal Understanding

DeepSeek-V3 Technical Report

Insights into DeepSeek-V3: Scaling Challenges and Reflections on Hardware for AI Architectures

DualPath: Breaking the Storage Bandwidth Bottleneck in Agentic LLM Inference

Reasoning

DeepSeek-Prover: Advancing Theorem Proving in LLMs through Large-Scale Synthetic Data

DeepSeek-Prover-V1.5: Harnessing Proof Assistant Feedback for Reinforcement Learning and Monte-Carlo Tree Search

DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning

CodeI/O: Condensing Reasoning Patterns via Code Input-Output Prediction

Inference-Time Scaling for Generalist Reward Modeling

DeepSeek-Prover-V2: Advancing Formal Mathematical Reasoning via Reinforcement Learning for Subgoal Decomposition

DeepSeek-R1-0528 Report

DeepSeekMath-V2: Towards Self-Verifiable Mathematical Reasoning

Frontier

Janus: Decoupling Visual Encoding for Unified Multimodal Understanding and Generation

JanusFlow: Harmonizing Autoregression and Rectified Flow for Unified Multimodal Understanding and Generation

Janus-Pro: Unified Multimodal Understanding and Generation with Data and Model Scaling

Native Sparse Attention: Hardware-Aligned and Natively Trainable Sparse Attention

DeepSeek-V3.1 Release: Hybrid Inference for the Agent Era

DeepSeek-V3.2-Exp: Boosting Long-Context Efficiency with DeepSeek Sparse Attention

DeepSeek-OCR: Contexts Optical Compression

DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models

mHC: Manifold-Constrained Hyper-Connections

Conditional Memory via Scalable Lookup: A New Axis of Sparsity for Large Language Models

DeepSeek-OCR 2: Visual Causal Flow

DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence

DeepSeek-V4.1-Flash: Pushing the Limits of KV Cache Compression