Kyunghyun Cho
@kchonyc
Followers
76K
Following
32K
Media
2K
Statuses
13K
a combination of a mediocre scientist, a mediocre manager, a mediocre advisor & a mediocre physicist at @nyuniversity (@CILVRatNYU) & @PrescientDesign
Manhattan, NY
Joined June 2009
btw, it took 12 years, but i did end up having a dinner with CL last year
some absolutely not-plagiarised text snippets in my dissertations:. * "it would not have been possible for me to survive long, freezing, snowy winter of Finland without songs from four girls of 2NE1 (especially, the leader CL)."
17
7
348
vibe coding continues with a @MicrosoftEdge extension: the world's most expensive, slow and inaccurate ad blocker, featuring Gemini!
2
0
12
the perfect response from @GoogleDeepMind cannot be simpler: just halve the price of everything. @orf_bnw @_angie_chen please start with the Gemini Pro subscription. pretty please?.
5
2
54
RT @markchen90: Try GPT-5 today!. One of our top level goals was to marry the pretraining and reasoning paradigms - this required lots of h….
0
12
0
RT @iclr_conf: ICLR 2026 will take place in. 📍Rio de Janeiro, Brazil.📅 April 23–27, 2026. Save the date - see you in Rio! #ICLR2026 https:….
0
95
0
RT @Daniel_J_Im: It’s been a long road. Grateful to have worked on this with @kyunghyun_cho and coauthors. What excites me most is pushin….
0
1
0
love oai open model, especially the one that can run locally. now, it’s time for OAI to avoid building itanium.
1
0
13
RT @OpenAI: Our open models are here. Both of them.
openai.com
Advanced open-weight reasoning models to customize for any use case and run anywhere.
0
3K
0
RT @stephenrra: I was reminded by @kchonyc that it's been exactly 4 years to the day since we landed at @genentech. A lot has happened pers….
0
5
0
RT @ABosselut: The EPFL NLP lab is looking to hire a postdoctoral researcher on the topic of designing, training, and evaluating multilingu….
0
22
0
yes, yes, no theory. but, i don't think we should shy away from real(ish) problems because we can't do theory on them. paper:
openreview.net
Existing causal inference (CI) models are often restricted to data with low-dimensional confounders and singleton actions. We propose an autoregressive (AR) CI framework capable of handling complex...
1
0
6
together with @Daniel_J_Im , we worked on this for a couple of years already: use an autoregressive model à la LLM to model distributions over sequencified combinatorial variables and run MC to estimate causal effects. so many possibilities with such an approach (knowledge from
2
8
39
if you want to know more about NYU Global AI Frontier Lab, check out
1
0
2
at NYU Global AI Frontier Lab, ever now and then when we find some free time between research and life, we sit down together over tea or coffee and talk about research, life and restaurants/bars in NYC (if you know me, you'd know which one of these i would spend most time talking
1
1
56
when an analytical solution exists and if we know, an iterative approach often converges rapidly, resulting in comparable complexities.
1
0
9
RT @thoma_gu: Yeah… this is why we usually call it “Non-autoregressive sequence generation” in early days! 👀
nar-tutorial.github.io
0
6
0