# mlsanity

> A map of machine learning research: 469k core ML papers on arXiv since 2021, grouped into 15 areas and 193 research directions, with what each paper builds on and what built on it. Updated daily; data through 7 Oct 2026.

How the map is made: https://mlsanity.com/about/
Read API (JSON, CORS open): https://mlsanity.com/api/  (OpenAPI: https://mlsanity.com/openapi.json)
Data downloads and licence: https://mlsanity.com/data/
Any arXiv paper: https://mlsanity.com/abs/<arXiv id> (redirects to its page)

## Areas

- [World models](https://mlsanity.com/wm/index.md): 11 directions
- [LLM reasoning](https://mlsanity.com/lr/index.md): 17 directions
- [Multimodal LLMs / vision-language models](https://mlsanity.com/mllm/index.md): 16 directions
- [Vision-language-action models](https://mlsanity.com/vla/index.md): 13 directions
- [Retrieval-augmented generation](https://mlsanity.com/rag/index.md): 15 directions
- [Diffusion models](https://mlsanity.com/diffusion/index.md): 14 directions
- [End-to-end autonomous driving](https://mlsanity.com/driving/index.md): 9 directions
- [LLM agents and tool use](https://mlsanity.com/agents/index.md): 16 directions
- [LLM alignment / RLHF / preference optimisation](https://mlsanity.com/alignment/index.md): 10 directions
- [Mixture-of-experts](https://mlsanity.com/moe/index.md): 12 directions
- [State-space sequence models (S4, Mamba)](https://mlsanity.com/ssm/index.md): 10 directions
- [Synthetic data and self-improvement for LLMs](https://mlsanity.com/synth/index.md): 13 directions
- [Offline RL and imitation learning from data](https://mlsanity.com/offlinerl/index.md): 13 directions
- [Efficient LLM inference and serving](https://mlsanity.com/inference/index.md): 12 directions
- [LLM safety and security](https://mlsanity.com/safety/index.md): 12 directions

## Cite

mlsanity (2026). A map of machine learning research. https://mlsanity.com/ (data through 7 Oct 2026).