bocoel
bocoel copied to clipboard
rentruewang
Bayesian Optimization as a Coverage Tool for Evaluating LLMs. Accurate evaluation (benchmarking) that's 10 times faster with just a few lines of modular code.
Currently, the factories are using enums. This isn't scalable.
Rather than using Bayesian, use MCTS to explore the search space.
I have decided to archive the repo. I am not feeling particularly motivated to work on this, and it has been inactive for a while now. For: #26 #29 :...
Since the optimizers mostly different from their sampling strategy, this change would make it more obvious.
Hi RenChu! This project looks really neat! I've looked through the docs, but I was wondering if you have any preliminary tech report that explains a bit more about how...
Metadata
Owner
Metadata
Bayesian Optimization as a Coverage Tool for Evaluating LLMs. Accurate evaluation (benchmarking) that's 10 times faster with just a few lines of modular code.