What is the value of proprietary information included in the training of an LLM?
- What is the value of proprietary information[ake] included in the training of an LLM?
- That information helps the LLM perform better, but how much?
- How much worse would the LLM be if you hadn't included that marginal bit of data?
- Would a human even notice?
- Someone pointed out that the Shapely Value might be a useful conceptual lens to try to get a handle on this.