Machine Quest

370 posts

Machine Quest banner
Machine Quest

Machine Quest

@machine_quest

I doubt therefore I think therefore I exist

Entrou em Mayıs 2022
852 Seguindo34 Seguidores
Tweet fixado
Machine Quest
Machine Quest@machine_quest·
Happy to have been part of this paper. This is my first EACL paper, but wont be the last. I thank everyone who enabled me to reach this height, especially my supervisors Dr. @ransurangika and Dr. @NisansaDdS
Nisansa de Silva@NisansaDdS

In our upcoming @eaclmeeting paper on the utilization potential of the quality of web-mined corpora, we discuss how you may build better translation models by automatically sorting the training samples and using the top samples. Paper: arxiv.org/abs/2402.07446 1/n

English
0
0
0
281
deep Manifold
deep Manifold@BetaTomorrow·
These are important points and worth further exploration. Learning complexity arises from data complexity, which is largely rooted in high-order nonlinearity. This is precisely where traditional supercomputing historically hit its limits, and AI systems are now encountering the same fundamental barrier.
deep Manifold tweet mediadeep Manifold tweet mediadeep Manifold tweet mediadeep Manifold tweet media
English
2
1
5
680
Sebastian Raschka
Sebastian Raschka@rasbt·
One of the underrated papers this year: "Small Batch Size Training for Language Models: When Vanilla SGD Works, and Why Gradient Accumulation Is Wasteful" (arxiv.org/abs/2507.07101) (I can confirm this holds for RLVR, too! I have some experiments to share soon.)
Sebastian Raschka tweet media
English
27
174
1.4K
90.4K
merve
merve@mervenoyann·
Chapter 5 of the Vision Language Models book is out and it's all about pre-training, illustrated and hands-on 🧑🏻‍🎨👩🏻‍🎨👨🏻‍🎨 here's a sneak peek 👀
merve tweet media
English
15
31
452
25.3K
Edoardo Ponti
Edoardo Ponti@PontiEdoardo·
@machine_quest @FSoudan This work takes inspiration from BLT, H-Net, as well as from earlier work, including Dynamic Token Pooling (DTP) aclanthology.org/2023.acl-long.… As far as I know, in DTP we were the first to introduce the idea of end-to-end tokenisation for autoregressive Transformers.
English
1
0
3
69
Edoardo Ponti
Edoardo Ponti@PontiEdoardo·
Finally, you can count the r's in strawberry and check if 3.11 is higher than 3.9 without tokenisation interfering: Here's Bolmo, a fully open byte-level LLM with latent tokenisation, derived from a SOTA LLM (Olmo 3). Promising on coding and char-level understanding!
Ai2@allen_ai

Introducing Bolmo, a new family of byte-level language models built by "byteifying" our open Olmo 3—and to our knowledge, the first fully open byte-level LM to match or surpass SOTA subword models across a wide range of tasks. 🧵

English
2
7
44
4.2K
Edoardo Ponti
Edoardo Ponti@PontiEdoardo·
@FSoudan That's the magic of latent tokenisation! The sequence length is shortened in internal layers. So, depending on the compression ratio, the model may be *even faster* than its subword-level counterpart.
English
1
0
1
99
Marc Lelarge 🌻
Marc Lelarge 🌻@marc_lelarge·
Ready for tomorrow’s lecture: 𝗣𝗿𝗼𝗴𝗿𝗮𝗺𝗺𝗶𝗻𝗴 𝗼𝗻 𝗚𝗣𝗨𝘀
Marc Lelarge 🌻 tweet media
English
7
57
709
27.9K
Marc Lelarge 🌻
Marc Lelarge 🌻@marc_lelarge·
Learn 𝗚𝗣𝗨 𝗽𝗿𝗼𝗴𝗿𝗮𝗺𝗺𝗶𝗻𝗴 from the ground up: begin with Numba for low-level control, then progress to Triton to write high-performance kernels in a Python-like language. A hands-on Jupyter notebook to get you started quickly.
Marc Lelarge 🌻 tweet mediaMarc Lelarge 🌻 tweet media
English
12
125
1.2K
54.4K
Machine Quest
Machine Quest@machine_quest·
@marc_lelarge Really looking forward to this and diving into this during my weekend. I appreciate your opensource policy!
English
1
0
1
159
Mor Geva
Mor Geva@megamor2·
✨ New course materials: Interpretability of LLMs✨ This semester I'm teaching an active-learning grad course at @TelAvivUni on LLM interpretability, co-developed with my student @dhgottesman. We're releasing the materials as we go, so they can serve as a resource for anyone curious about how LLMs work from the inside: github.com/mega002/llm-in…
English
14
109
796
49.8K
Machine Quest
Machine Quest@machine_quest·
@elliotarledge This is awesome, such an awesome material to dive into pytorch as well
English
0
0
1
113
Machine Quest
Machine Quest@machine_quest·
@jxmnop All the very best!!! Would love to read your thesis when its public
English
0
0
0
26
Machine Quest
Machine Quest@machine_quest·
@jxmnop @goyal__pramod Wow! This is brilliant. Question: the written column in your spreadsheet, is it conference date or the arxiv date ?
English
0
0
1
28
dr. jack morris
dr. jack morris@jxmnop·
from 2017–2020 i was learning ML. i didn't publish any research (and hadn't yet); i just trained a lot of tiny models and read a paper every single day i maintained a giant spreadsheet with notes about each paper along with random thoughts. was a great way to learn
dr. jack morris tweet mediadr. jack morris tweet mediadr. jack morris tweet mediadr. jack morris tweet media
English
50
125
1.8K
123.8K
Machine Quest retweetou
Pedro Domingos
Pedro Domingos@pmddomingos·
Hinton is no longer afraid of superintelligence.
English
570
622
4.1K
814.9K
Machine Quest
Machine Quest@machine_quest·
@zouharvi Congrats, this is truly an exceptional acheivement, would love to hear your advice on applying for such prestigious awards.
English
0
0
1
223
Vilém Zouhar @ EACL
Vilém Zouhar @ EACL@zouharvi·
Grateful to receive the Google PhD Fellowship!🙂 I am not secretive about having applied to 4 similar fellowships during my PhD before and didn't succeed. Still, refining my research statement (part of the application) helped me tremendously in finding out the real interesting..
Google.org@Googleorg

🎉 We're excited to announce the 2025 Google PhD Fellows! @GoogleOrg is providing over $10 million to support 255 PhD students across 35 countries, fostering the next generation of research talent to strengthen the global scientific landscape. Read more: goo.gle/43wJWw8

English
12
13
268
31.5K
saurabh
saurabh@saurabhtwq·
lion only care about nanochat for next few days.
saurabh tweet media
English
3
1
74
15.1K