3 ms·
>we've had powerful summarization techniques for decades now (deep-learning, or otherwise) - and they aren't susceptible to the hallucinations of present day Ge
by dontupvoteme 3y ago
>we've had powerful summarization techniques for decades now (deep-learning, or otherwise) - and they aren't susceptible to the hallucinations of present day Generative LLMs.
do you mean BERT, Seq2Seq and similar?
My biggest takeaway from summarization tasks was that the benchmarks/evals seemed rather.. lacking.