A model can top the MTEB average and be ordinary at retrieval
The number that sorts the MTEB leaderboard is one flat average over 58 datasets, weighting retrieval fifteen times as heavily as summarization purely because retrieval happens to have more of them.
A slightly wrong goal, optimized hard enough, can miss by everything
Eliezer Yudkowsky argues that a goal missing human value by a hair, optimized hard enough, ends in a galaxy tiled with molecular smiley faces.
LoRA matched full fine-tuning by training 4.7M of GPT-3's 175 billion weights
The 2021 Microsoft paper that made cheap fine-tuning routine reported parity only on the constrained language benchmarks it ran.
A strict output format can block the reasoning a model needs to be right
The program holding a model's output to a schema blocks every next token that would break the format, sometimes the very ones the model needed to reason with.
The FTC found DoNotPay never tested its robot lawyer against a human one
DoNotPay advertised an AI “robot lawyer” as operating “like a human lawyer,” a claim the FTC says it never backed with a single attorney or test.