Krisztina Sinkovics
@KristaSinkovics
Followers
3,549
Following
162
Media
8
Statuses
59
AI Researcher. ML @ Cambridge University
Cambridge, England
Joined October 2021
Don't wanna be here?
Send us removal request.
Explore trending content on Musk Viewer
Texas
• 231106 Tweets
England
• 149844 Tweets
Mustafa Kemal
• 140900 Tweets
Notre Dame
• 82234 Tweets
MSNBC
• 78856 Tweets
Kentucky
• 33719 Tweets
Arkansas
• 29942 Tweets
Iowa
• 28972 Tweets
Sabalenka
• 26144 Tweets
Auburn
• 18339 Tweets
Marshall
• 17620 Tweets
#AEWAllOut
• 16321 Tweets
#الهلاليين_يرفضون_لودي
• 15083 Tweets
Pegula
• 14066 Tweets
T-岡田
• 11854 Tweets
Ozee
• 11791 Tweets
#UFCVegas97
• 10841 Tweets
#ボクらの時代
• 10462 Tweets
Go Green
• 10311 Tweets
At ICML you can bump into your heroes. And today it happened! I met
@AlexanderTong7
whose Flow Matching with Optimal Transport (OT-CFM) has been so inspirational for my work
7
21
1K
#FlowMatching
breakfast
@icml
. Great chat about flow matching, diffusion, consistently models, stochastic Interpolants and all in between
2
3
273
Weeks of reading papers could not outmatch the insights from diffusion circles today. Thank you
@sedielem
for sharing your wisdom and gathering us all together
1
0
49
Many great ideas in ML are inspired by Physics, we've seen these Conservation Laws underpinning different areas in ML, take energy score in diffusion models for mappings between distributions in time as example. I'm curious to see how the last one goes
5
6
42
Getting insights into Material Science at Cambridge Graphene Centre before our workshop on Machine Learning in Life and Material Science at
@icmlconf
2
0
35
Fantastic talk by
@FerranAlet
presenting their work at DeepMind on using Graph Neural Networks to predict weather at a massive scale, beating all other methods. GraphCast was featured on Science Magazine cover.
GenCast added diffusion to get smooth predictions with uncertainty.
0
3
23
If you are into
#FlowMatching
and happen to be at the
#icml24
, join us for a breakfast get together tomorrow (25 Jul) 10am at BRENNER cafe
1
0
21
Next token prediction + full sequence generation is a combo I've been looking forward to see and here it comes
Introducing Diffusion Forcing, a new way of training sequence generative models that unifies next-token prediction (think LLM) and full-sequence diffusion (think video diffusion models)! I’m super excited about this - it has a number of unique skills! (1/n)
5
65
533
3
0
12
The paper at the basis of OpenAI's Sora was rejected from CVR2023 for "lack of novelty". With the amount of attention that Sora generated, diffusion transformers will probably inspire a lot more papers that will show up in the upcoming conferences
The Diffusion Transformer paper, by my former-FAIR-and-current-NYU colleague
@sainingxie
and former-Berkeley-student-and-current-OpenAI engineer William Peebles, was rejected from CVR2023 for "lack of novelty", accepted at ICCV2023, and apparently forms the basis for Sora.
47
417
3K
0
0
11
Ukrainian Independence Day feels different when my country is under constant attack. So much gravity, so much symbolism.
#StandWithUkraine
#UkraineWar
0
0
6
@HannesStaerk
Thanks a lot for organising this talk! We briefly discussed the paper with Francesco Di Giovanni a couple of weeks ago, but an in-depth overview like this is much appreciated! Great work on extending Flow Matching!
0
0
3
It’s been 90 days since Russia attacked my home country. I haven’t noticed that summer is almost here, still stuck mentally between now and February 24
0
0
2
Wouldn’t wake up at 4 am for anything, except for flights. But for a good
#NeurIPS
social it’s worth it a hundred times
0
0
2
@matthewclifford
thank you for the great talk at the Cambridge AI Safety Hub!
My takeaways from the event:
- AI safety is the cornerstone of widespread adoption of AI in the economy.
- As the political landscape broadens the demand for AI safety and its portfolio will expand.
0
0
2
@FerranAlet
@RAEngNews
@GoogleDeepMind
Congratulations! It's exciting to see the impact of your team's work and the follow-up GenCast. Looking forward to seeing how you tackle new challenges and take it further
0
0
2
@im_prashant_rai
That sounds interesting. Are you open to share a bit more? I don't know much about mapless navigation and curious how this could be applied in this area
1
0
1
@Alpha7987
@icml
There's a Diffusion get together at 3pm Hall A. FM people will be there too.
We spoke about applications of FM to molecular and protein co-design, going beyond euclidean Interpolants, recent papers on FM extensions, good blog posts. I'll share some links if you are interested
2
0
1
@emaros96
Check out our workshop on ML for Life and Material Sciences on July 26th, there will be a number of posters along the areas of your interest.
0
0
1