BrandoHablando Profile Banner
Brando Miranda @ ICML 2025 Profile
Brando Miranda @ ICML 2025

@BrandoHablando

Followers
1K
Following
2K
Media
87
Statuses
1K

CS Ph.D. @Stanford, researching data quality, foundation models, and ML for Theorem Proving. Prev: @MIT, @MIT_CBMM, @IllinoisCS, @IBM. Opinions are mine. 🇲🇽

Stanford, CA
Joined March 2013
Don't wanna be here? Send us removal request.
@BrandoHablando
Brando Miranda @ ICML 2025
3 days
🚨 Can your LLM really do math—or is it cramming the test set?. 📢 Meet Putnam-AXIOM, a advanced mathematics contamination-resilient benchmark that finally hurts FMs. 1. 2. #ICML2025 East Exhibition Hall A-B, #E-2502. 🧵1/14
Tweet media one
Tweet media two
3
16
56
@BrandoHablando
Brando Miranda @ ICML 2025
4 hours
Ok, now that ICML is officially over. I guess it's time to delete my twitter app again and lock in to deep life/work? :)💪🧠.
0
0
4
@BrandoHablando
Brando Miranda @ ICML 2025
1 day
RT @allenainie: Come to EXAIT today for our best paper talk at 9:05 and poster session 11:45-2-15! The workshop also has a great set of tal….
0
5
0
@BrandoHablando
Brando Miranda @ ICML 2025
1 day
RT @pli_cachete: Terence Tao on the supposed Gold from OpenAI at IMO
Tweet media one
Tweet media two
Tweet media three
0
498
0
@BrandoHablando
Brando Miranda @ ICML 2025
1 day
@arimorcos this is the workshop paper were we show your takeaways from ICML that target distribution is "all you need" (sorry can't remember how you actually said it! Didn't mean to be cheesy). #ICML2025.
0
0
2
@BrandoHablando
Brando Miranda @ ICML 2025
1 day
Sorry I meant this picture! :D. Not the same one twice, X doesn't let me update my post.
Tweet media one
0
1
4
@BrandoHablando
Brando Miranda @ ICML 2025
1 day
Come to my second poster session about Data centric Machine Learning (DMLR)! . At 209-2010!. #ICML2025
Tweet media one
Tweet media two
@BrandoHablando
Brando Miranda @ ICML 2025
1 day
Come to 208-209 ICML data workshop and chat with me about how to use data optimally! Scale isn't everything! . Ask me how to use it beyond post-training ;). - Scale isn’t enough: LLM performance rises with training‑task alignment more than with data volume. - Robust Alignment
Tweet media one
Tweet media two
1
1
8
@BrandoHablando
Brando Miranda @ ICML 2025
1 day
RT @_akhaliq: Beyond Scale: the Diversity Coefficient as a Data Quality Metric Demonstrates LLMs are Pre-trained on Formally Diverse Data….
0
12
0
@BrandoHablando
Brando Miranda @ ICML 2025
1 day
work at/with @sanmikoyejo @stai_research !.
0
1
1
@BrandoHablando
Brando Miranda @ ICML 2025
1 day
@_akhaliq @_alycialee Joint work with @ObbadElyas Mario Krrish Aryan @sanmikoyejo Me Sudarsan at @stai_research !. Thank you!. 🧵3/3.
1
2
3
@BrandoHablando
Brando Miranda @ ICML 2025
1 day
We first demonstrated scale isn't enough in our Beyond Scale paper using the diversity coefficient! . thanks for featuring us @_akhaliq !. Work led by @_alycialee et al!. 🧵 2/3.
@_akhaliq
AK
2 years
Beyond Scale: the Diversity Coefficient as a Data Quality Metric Demonstrates LLMs are Pre-trained on Formally Diverse Data. paper page: Current trends to pre-train capable Large Language Models (LLMs) mostly focus on scaling of model and dataset size.
Tweet media one
2
3
15
@BrandoHablando
Brando Miranda @ ICML 2025
1 day
Come to 208-209 ICML data workshop and chat with me about how to use data optimally! Scale isn't everything! . Ask me how to use it beyond post-training ;). - Scale isn’t enough: LLM performance rises with training‑task alignment more than with data volume. - Robust Alignment
Tweet media one
Tweet media two
3
1
5
@BrandoHablando
Brando Miranda @ ICML 2025
1 day
Come to Convention Center West room 208-209 2nd floor to learn about optimal data selection using compression like gzip! . tldr; you can learn much faster if you use gzip compression distances to select data given a task!. DM if you are interested or what to use the code!.
@ObbadElyas
Elyas Obbad
9 months
🚨 What’s the best way to select data for fine-tuning LLMs effectively?. 📢Introducing ZIP-FIT—a compression-based data selection framework that outperforms leading baselines, achieving up to 85% faster convergence in cross-entropy loss, and selects data up to 65% faster. 🧵1/8
Tweet media one
0
4
7
@BrandoHablando
Brando Miranda @ ICML 2025
1 day
RT @allenainie: If you missed @wanqiao_xu’s presentation, here are some of our slides! (The workshop will post full slides later on their w….
0
18
0
@BrandoHablando
Brando Miranda @ ICML 2025
2 days
RT @heeney_luke: Academia must be the only industry where extremely high-skilled PhD students spend much of their time doing low value work….
0
123
0
@BrandoHablando
Brando Miranda @ ICML 2025
2 days
RT @ai4mathworkshop: It's happening today!.📍Location: West Ballroom C, Vancouver Convention Center.⌚️Time: 8:30 am - 6:00 pm.🎥 Livestream:….
0
11
0
@BrandoHablando
Brando Miranda @ ICML 2025
2 days
@HenryJamesBosch Picture with Leni! Coauthor of Veribench and long time collaborator!. Main mind behind Pantograph! #TACAS2025
Tweet media one
0
1
2
@BrandoHablando
Brando Miranda @ ICML 2025
2 days
Another poster by @zhankezhou and @tmlrgroup that I believe couldn't come due to visa detals
Tweet media one
0
1
3
@BrandoHablando
Brando Miranda @ ICML 2025
2 days
CoDaPo awesome joint work with @zhankezhou and @tmlrgroup !
Tweet media one
1
1
7
@BrandoHablando
Brando Miranda @ ICML 2025
2 days
🚶‍♀️come talk to me live!. Thanks to @HenryJamesBosch for helping set and for the advertisment! :)
Tweet media one
1
1
3