@xtimv

machine learning, reinforcement learning, programming languages, handstands (he/him)

 NYC
timvieira.github.io/blog
Joined June 2009
  • Tweets2.6K
  • Following1K
  • Followers3.8K
153Media
When you say, "This is a reinforcement learning problem," you should say it with the same excitement as "This is NP-hard."
7
37
8
224
Tim Vieira retweeted
Sasha Rush
@srush_nlpOct 19
Late to the party, but this blog post is excellent. Arguably the way people should first learn neural networks.
Alex Wang
@heyyalexwangOct 16
how much can models remember? there’s a neat perspective on neural networks that deserves to be more widely known here’s how you could have invented every layer of the transformer from memory (1/n) from my new blog post!
10
68
1,169
82,796
Tim Vieira retweeted
Sasha Rush
@srush_nlpOct 15
The Token Game (srush.github.io/the-token-ga…) It's BPE the mobile game? No reason for this to exist. I'm absolutely terrible at it.
9
7
111
10,039
Tim Vieira retweeted
David Hall@dlwhOct 15
I've been thinking about where the data for the next generation of open models will come from. The internet gave us a remarkable commons for pretraining. Posttraining is different: it consumes tasks, not documents. And that commons doesn't exist. Not yet. openathena.ai/blog/there-is-…

There is no posttraining data commons

The LLM revolution ran on a pretraining commons that humanity built for other reasons. Posttraining consumes tasks, not documents, and no such commons exists. If the open models community wants...

openathena.ai
9
34
6
206
30,553
My dream had always been to prove beautiful theorems as the AI agents unloaded my dishwasher. Well, it went the wrong way... 😂
40
311
38
2,875
135,422
Tim Vieira retweeted
Percy Liang
@percyliangOct 6
🚢Marin 535B-A23B training has crossed the halfway mark! Some observations and learnings so far:
23
99
10
1,073
71,503
Tim Vieira retweeted
Sasha Rush
@srush_nlpOct 1
Sampling to Reinforce (srush.github.io/sampling-to-…) An elementary tutorial on sampling w/o randomness with a focus on gaining intuition for variance reduction as the key challenge.
20
46
667
36,488
Tim Vieira retweeted
Sasha Rush
@srush_nlpSep 15
Lean Verified Transformers (srush.github.io/lean-transfo…) In which we prove a bunch of Transformer invariants from scratch in Lean, and speculate about how hard it would be to do that for the rest of the world's code.
27
126
11
1,035
66,571
Tim Vieira retweeted
Geoffrey Litt
@geoffreylittJul 2
Hot take: I think it's still important to understand the code that our agents write! In this mega thread (based on my AIE talk today), I will explain why that's the case, and show some ideas for how to efficiently understand code. Alright, let's dive in. 1/
105
472
139
3,303
489,486
Slides from my ICML tutorial "Is numerical optimization theory irrelevant to machine learning practice in 2026?": cs.ubc.ca/~schmidtm/Document… (Updated to fix some typos, incorporate feedback, and add some things I did not have time for. Will link to the video here when posted.)
5
85
5
539
45,851
Tim Vieira retweeted
Sasha Rush
@srush_nlpJul 21
Fantastic slides from the ICML tutorial of @MarkSchmidtUBC (“is opt theory relevant in 2026”) cs.ubc.ca/~schmidtm/Document…
7
61
2
628
56,072
Replying to @ReviewAcl
@ReviewAcl Why does every author have to fill out the registration form if they already have an OpenReview profile?
2
3
1
24
2,517
Tim Vieira retweeted
Rico Angell
@rico_angellMay 14
It’s deployment time! You’ve done the pre-deployment evals. You THINK your model is safe, so you ship it 🚀 🚨 After deployment, reports of misbehavior start trickling in What happened?? How could you have caught it?? 🤔 @icmlconf 2026 Spotlight! 🧵
3
22
3
108
20,828
Our paper "Autoregressive Language Models are Secretly Energy-Based Models: Insights into the Lookahead Capabilities of Next-Token Prediction" was accepted for publication at #ICML2026 arxiv.org/abs/2512.15605

Autoregressive Language Models are Secretly Energy-Based Models:...

Autoregressive models (ARMs) currently constitute the dominant paradigm for large language models (LLMs). Energy-based models (EBMs) represent another class of models, which have historically been...

arxiv.org
2
26
183
14,599
Tim Vieira retweeted
Yahya Emara
@yemara22Apr 20
We made LLM inference a lot faster. 1/n Speculative decoding throws away draft tokens when they don't match. What if you re-weighted your samples instead using importance sampling? We introduce SMC-SD, sequential Monte Carlo speculative decoding, an approximate sampler that speeds up inference by 2.36× over SOTA speculative decoding, while producing a controllable Pareto frontier of speed and approximation accuracy. ArXiv: arxiv.org/abs/2604.15672 minimal repo release: github.com/abdelfattah-lab/s… Joint work with Mauricio Barba da Costa (@mauricio_b25181), Chi-Chih Chang (@CCCCC1009CCC ), Cameron Freer, Tim Vieira (@xtimv), Ryan Cotterell, and Mohamed Abdelfattah (@mohsaied).
7
23
7
149
23,115
Claude Code is really bad at f-strings. I can't tell you how many times a day I see some version of this error message a day. Please fix!
1
4
592
Some more vibe-coding fun - every math major's favorite party trick: the wobbly table theorem as an interactive 3D visualization. timvieira.github.io/table-th…
1
8
475
Ok, AI-assisted coding is pretty rad. Here's something I built to explore the stuff I've written (blog posts and publications). It's still a bit of work in progress, but it's pretty fun, especially the "semantic" tab + sliders. I'd love some feedback! timvieira.github.io/experime…
2
10
557
I built an interactive JavaScript thingy to study the two faces of KL divergence. timvieira.github.io/blog/int… I have wanted this since 2009. Thank you, Claude Code, for helping me get there!
3
9
2
52
8,090
It's pretty addictive to play with. You can drag the bumps around and +/- them. It currently only supports single-model approximators.
1
3
492
Congratulations to our researchers (including @leoduw and alum @xtimv) on winning an Outstanding Paper Award at the most recent @COLM_conf! Learn more about their award-winning paper here: cs.jhu.edu/news/hopkins-comp…
COLM logo.

ALT COLM logo.

4
3
612
Load more