We didn't have the compute required (GPUs powerful enough to parallelize forward and backward pass). This compute is what allows us to train from human knowledge or distillation.
I don't understand all the hype for generating SVG with LLM. The task is not really useful, doesn't seem that interesting in single shot as it's really hard, and no human could do it (it would be more useful if the model has visual feedback and could correct the result).
And also, since it becomes a popular task, companies will add the examples in their training set, so you're just benchmarking who has the better text to SVG training set, not the overall quality of the model.
Multi-step reasoning means that the LLM is giving a question (maths here), and generating an answer that consists of many intermediate words, before returning the solution.
Here, we don't want to tell the LLM how to solve the problem word-by-word. We want to tell it at the end, "correct" or "incorrect", and have the model learn on its own to generate intermediate steps, to reach the solution.
That's typically a setup where RL is desirable (even necessary): we have sparse rewards (only at the end) and give no details to the model on how to reach the solution. It's similar to training models to play chess against a specific opponent.
Those are tree search techniques, they are not metrics to assess the "human" complexity of a line. They could be used for this purpose but out of the box they just give you winning probability