Grass
@grass
Our mission is to redefine Internet incentive structures
加入 March 2023
3 正在關注 544.6K 粉絲
The world is coming to realize what we've always known.
The difference is the data.
This is notable. DeepSeek, a lab usually first to pioneer novel algorithms and architectures, is saying that at this point, the ROI of improving data quality far exceeds that of working on novel post-training algorithms.
I think this has already been true for some time for non-lab practitioners. If you're doing llm post-training, 80% of your effort should go into looking at your data.
This means:
- Hiring experts to dig through your RL tasks
- Sifting through rollouts and sft data by hand to remove suspicious samples. Make sure all tasks are actually passable.
- Making sure your data is diverse in both difficulty and category.
顯示更多
