I mean that everyone uses these shortcuts to get good numbers because everyone before them did. You have to get better numbers if you want to publish (even though a negative result is still science), and you have to publish to graduate or advance your career.
On Wed, Sep 15, 2021, 10:53 AM <[email protected]> wrote: > I had seen several papers but none say exactly how they compute Perplexity > really, even at all, I'm pretty sure. > > Matt, I am confused, you said, in the same sentence, a contradictory > sentence, in, the same, single sentence, there is no way to know which way > you meant.....you said above essentially they need to get a good grade > hence they cheat and post a slightly better score using different methods > like removing stopsymbols, but then you at the same time said in the very > same sentence (LOL!!!) that "but you have to use the same methodology to > show that your algorithm works better than prior work"............I know > people probably lean to both ways especially when rules are not stated on > their benchmark I posted, but your opinion is not clear. > *Artificial General Intelligence List <https://agi.topicbox.com/latest>* > / AGI / see discussions <https://agi.topicbox.com/groups/agi> + > participants <https://agi.topicbox.com/groups/agi/members> + > delivery options <https://agi.topicbox.com/groups/agi/subscription> > Permalink > <https://agi.topicbox.com/groups/agi/Tc9a99c50a9ec758e-M92656736d8e72c9a63c06ce3> > ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/Tc9a99c50a9ec758e-M3835da91a2db55d1dc351c32 Delivery options: https://agi.topicbox.com/groups/agi/subscription
