Tracking instruction-tuned LLM openness. Paper: Liesenfeld, Andreas, Alianda Lopez, and Mark Dingemanse. 2023. “Opening up ChatGPT: Tracking Openness, Transparency, and Accountability in Instruction-Tuned Text Generators.” In Proceedings of the 5th International Conference on Conversational User Interfaces. doi:10.1145/3571884.3604316.
Many models come in variants that differ only in the number of parameters (7B/30B etc) and are otherwise the same on all criteria.
So far we simply have picked one, which is fine. But it can be useful to have them besides one another. Only then it'd be nice to also be able to collapse a family so that the table doesn't become cluttered.
Would require adding a variable family and checking for that. Would also require some html and CSS adjustments.
Many models come in variants that differ only in the number of parameters (7B/30B etc) and are otherwise the same on all criteria.
So far we simply have picked one, which is fine. But it can be useful to have them besides one another. Only then it'd be nice to also be able to collapse a family so that the table doesn't become cluttered.
Would require adding a variable
familyand checking for that. Would also require some html and CSS adjustments.