awesome-deep-reasoning
github.com/modelscope/awesome-deep-reasoning ↗Collect every awesome work about r1!
Use this list with your AI agent
Add the Context Awesome MCP server to Claude, Cursor, or any MCP client, then ask:
"Show me replicates of deepseek-r1 and deepseek-r1-zero resources from awesome-deep-reasoning"
Installation instructions →What's inside
RelatedRepos
- 32B-DeepSeek-R1-ZeroReplicates of DeepSeek-R1 and DeepSeek-R1-Zero
- HuggingFace Open R1Replicates of DeepSeek-R1 and DeepSeek-R1-Zero
- Logic-RLReplicates of DeepSeek-R1 and DeepSeek-R1-Zero
Reproduce R1 Zero on Logic Puzzle
- oatllmReplicates of DeepSeek-R1 and DeepSeek-R1-Zero
- Open-R1-MultimodalAdvanced Reasoning for Multi-Modal
A fork to add multimodal model training to open-r1
- Open-Reasoner-ZeroReplicates of DeepSeek-R1 and DeepSeek-R1-Zero
Infra
- 3FS
[DeepSeek] A high-performance distributed file system designed to address the challenges of AI training and inference workloads.
- DeepGEMM
[DeepSeek] Clean and efficient FP8 GEMM kernels with fine-grained scaling
- DualPipe
[DeepSeek] DualPipe achieves full overlap of forward and backward computation-communication phases, also reducing pipeline bubbles.
- https://github.com/deepseek-ai/FlashMLA
- https://github.com/hiyouga/EasyR1
- https://github.com/huggingface/Math-Verify
Evaluation
- AIME-2024
This dataset contains problems from the American Invitational Mathematics Examination (AIME) 2024.
- AIME-VALIDATION
All 90 problems come from AIME 22, AIME 23, and AIME 24
- aimo-validation-amc
All 83 samples come from AMC12 2022, AMC12 2023
- Best practice for evaluating R1/o1-like reasoning models
- Codeforces-Python-Submissions
A dataset of Python submissions from Codeforces.
- GPQA-Diamond
Diamond subset from GPQA benchmark.
Papers
- AT WHICH TRAINING STAGE DOES CODE DATA HELP LLM REASONING?2024
- Competitive Programming with Large Reasoning Models2025.02
OpenAI: Competitive Programming with Large Reasoning Models
- DAPO2025.03
DAPO: An Open-Source LLM Reinforcement Learning System at Scale
- DeepSeek Math Tech-Report(GRPO)2024
- DeepSeek-R1-Tech-Report2025.01
- DeepSeek-V3 Tech-Report2025.02
Datasets
- BAAI-TACO
TACO is a benchmark for code generation with 26443 problems.
- Bespoke-Stratos-17k
A reasoning dataset of questions, reasoning traces, and answers.
- clevr_cogen_a_train
A R1-distilled visual reasoning dataset.
- Clevr_CoGenT_TrainA_R1
A multi-modal dataset for training MM R1 model.
- HuggingFace
ModelScope) - 800k samples dataset to train DeepSeek-R1 Distill models.
- HuggingFace
ModelScope)
News
- Bailian
- DAPO
DAPO: An Open-Source LLM Reinforcement Learning System at Scale
- deep-research
- o3-mini & o3-mini-high
- the DeepSeek-R1 model
- VSCode co-pilot
Resources
- ModelScope-r1-collection
HuggingFace-r1-collection
Showing a sample of 109 resources. View the full list on GitHub →