Fabian Gloeckle

🔥Very happy to introduce BigO(Bench) dataset on @hf.co 🤗

✨3,105 coding problems and 1,190,250 solutions from CodeContests

✨Time/Space Complexity labels and curve coefficients

✨Up to 5k Runtime/Memory Footprint measures for each solution

huggingface.co/datasets/fac...

Causal multi-token prediction at scale!

x.com/francoisfleure…

Still a bit to go

x.com/DavidJAlba94/s…

The PyTorch reference transformer implementation for researchers. Amazing work!

x.com/mathuvu_/statu…

New generation of cross-lingual embedding models -- the same could work for code + math

x.com/JoaoMJaneiro/s…

🤩

x.com/TacoCohen/stat…

GitHub - facebookresearch/Evariste: HyperTree Proof Search for Neural Theorem Proving -- "La science est l'œuvre de l'esprit humain, qui est plutôt destiné à étudier qu'à connaître, à chercher qu'à trouver la vérité." HyperTree Proof Search for Neural Theorem Proving -- "La science est l'œuvre de l'esprit humain, qui est plutôt destiné à étudier qu'à connaître, à chercher qu&...

by @GuillaumeLample @tlacroix6 @MaLachaux @AurR0d @Amaury_Hayat @LavrilThibaut Gabriel Ebner @javier_m and others!

github.com/facebookresear…

awesome folks, awesome event - thanks @rajko_rad !

x.com/rajko_rad/stat…

#ICML Looking for recommendations for posters and orals on
- LM pretraining losses
- non-autoregressive architectures, energy-based architectures
- planning, subgoals, task decomposition
- ML for math and reasoning

Please shamelessly self-promote, promote others, comment or DM!

At #ICML this week to present Multi-token Prediction with @byoubii: poster session Wednesday 11.30 am - 1 pm Hall C 4-9 #915 - drop by to chat about pretraining losses with us!

icml.cc/virtual/2024/p…

Please throw all your interpretability research and decoding tricks on these

x.com/syhw/status/18…

great work!

x.com/AdrienBardes/s…

science starts in the appendix