Neuro-Symbolic AI, RSM @MITIBMLab, @IBMResearch. Moved to bsky for English contents. My tweets don't represent the view of my organization.
Boston, MA, Planet Lisp
- Tweets 52,177
- Following 1,260
- Followers 1,082
- Likes 26,621
'(Masataro Asai Ph . D) @guicho271828@fedibird.com retweeted
🚲 Valtteri Bottas has completed another 161km on his cycle from Malaysia to Singapore.
Nearly there.
24
973
16,357
329,743
'(Masataro Asai Ph . D) @guicho271828@fedibird.com retweeted
Yea I’m riding to Singapore 🚲
On the gravel roads less travelled with my new @canyon_bikes Grail CFR.
See you there 🇸🇬
681
7,516
63,874
1,590,643
'(Masataro Asai Ph . D) @guicho271828@fedibird.com retweeted
Took an uber from Boston to NYC and it’s cheaper than Amtrak, Bus, Flying. Inefficient pricing exists everywhere.
73
37
2,990
416,923
'(Masataro Asai Ph . D) @guicho271828@fedibird.com retweeted
‼️New paper alert‼️ What does it take to recover the lost accuracy in non-prefix KVCache reuse? Simple averaging sometimes works surprising well! No fine-tunig, zero runtime cost. Article/paper:
linkedin.com/pulse/kvprior-e…
#RAG #LLM #KVCache #NLP #AI
1
1
3
403
Joint work with Bo Wu and Dan Gutfreund, and actually many others at the MIT-IBM Watson AI Lab.
Paper: zenodo.org/records/23106326 (It is submitted to Arxiv too, but it is on hold due to recent submission surge. We will update when it starts processing.)
1
71
Bonus: If you are annoyed by the current conference reviewing, check this out! I wrote it on a whim and it would probably not heal your illness :)
ICLR with 60k+ submission is a clear sign that peer review is broken. Here is a proposal for Contrastive Review System (CRS): reviewers rank a group of papers and authors defend their positions by criticising other papers in the group.
zenodo.org/records/22903811
52
