Training a 4B model to produce 81% faster query plans than Postgres - Rohan Bansal
rohanbansal.com · 9,838 words · saved by 1 readers
...or how to make Qwen learn query optimization via agentic reinforcement learning
How good are query optimizers, really? Leis et al. asked this exact question in 2015. Then, they asked it again 10 years later. Despite an enormous body of research spanning a decade since their original exploration, they found that query optimizers continue to leave much to be desired. I was surprised when I first learned about this. A Postgres database should know everything about the stuff that lives in its tables, no? How hard can it be? As it turns out: enormously hard. In fact, one particular task a query optimizer needs to do, join ordering, is known to be NP-hard. So query…
saved by
related reading
- Training a 4B model to produce 81% faster query plans than Postgresrohanbansal.com
- On-Policy Distillation - Thinking Machines Labthinkingmachines.ai
- Why Tool AIs Want to Be Agent AIs · Gwern.netgwern.net
- PostTrainBenchposttrainbench.com
- Efficient LLM inferencefinbarrtimbers.substack.com
- [2510.11696] QeRL: Beyond Efficiency -- Quantization-enhanced Reinforcement Learning for LLMsarxiv.org
- Spending Inference Time - Kevin Lukevinlu.ai
- Putting Task Expertise into RL Achieves State-of-the-Art Performance on Text-to-SQLthinkingmachines.ai
- State of RL for reasoning LLMs | A. Weersaweers.de
- Scaling PostgreSQL to power 800 million ChatGPT users | OpenAIopenai.com
- Speeding up RL with high-leverage samples | Applied Computeappliedcompute.com
- Q-learning is not yet scalableseohong.me