Leo Boytsov    @srchvrs    11/23/2021      

Very interesting work, but I would like to highlight that massive pre-training doesn't help retrieval much (see Table 8). I personally played with base and large BERT rankers & recently with nearly 1B-param one, but I don't get much benefit (only modest improvements).
  
    1         14




  Hashtags  







 
  Related  

Dani S. Bassett    @DaniSBassett    12/3/2021      

Preparing for my favorite class of the semester. Who doesn't want to start their day with massive networks in rainbow colors? #NetworkScience
  
    3         173



Anna Rogers    @annargrs    11/10/2021      

A highlight from @zacharylipton's talk at https://t.co/axQAhW5YdF on positive/negative results in engineering papers: If X works, even for no reason, community wants to know about it. But not if X doesn't work - since there was no reason why it would work in the first place. /1
  
    9         82



antonio vergari    @tetraduzione    12/1/2021      

discovered it just now, might miss the first talk, but the others are all super interesting
  
          5



Thomas G. Dietterich    @tdietterich    12/4/2021      

Interesting new conference starting up in AutoML
  
          5