trott

889 karmajoined March 3, 2016On HN

Oleg Trott, PhD (Columbia)

* Top 1% and co-winner in the biggest ML competition (Department of Homeland Security, $1,500,000 prize fund)

* Built the most cited molecular docking program (30,000 citations)

* Discovered how to get perfect (0.00000) scores in ML competitions (It's an algorithmic trick)

Details: https://olegtrott.com

Recent submissions

Tiny Model, Big Logic: Large-Model Reasoning Ability in VibeThinker-1.5B arxiv.org
4trott9modiscuss
The flaw in cryonics that could ruin your shot at living forever sciencefocus.com
2trott1y1 comment
Is this the $200K ticket to cheating death?bbc.com
3trott1ydiscuss
ARC Prize side quest: SnakeBench arcprize.org
4trott1ydiscuss
LIMO: Less Is More for Reasoning arxiv.org
389trott1y128 comments
De novo designed proteins neutralize lethal snake venom toxins nature.com
1trott1ydiscuss
Frontier Models are Capable of In-context Scheming arxiv.org
10trott1y1 comment
ARC Prize 2024: Technical Report [pdf]arcprize.org
2trott1ydiscuss
Duct-taped banana artwork sells for $6.2M in NYC bbc.com
7trott1ydiscuss
Test-Time Training on Nearest Neighbors for Large Language Models arxiv.org
1trott1ydiscuss
The surprising effectiveness of test-time training for abstract reasoning [pdf]mit.edu
87trott1y27 comments
AI Mathematical Olympiad – Progress Prize 2 kaggle.com
80trott1y6 comments
LLMs and Diagnostic Reasoning: A Randomized Clinical Vignette Study [pdf]medrxiv.org
3trott1y3 comments
O1 Test-Time Compute Scaling Laws github.com/hughbzhang
1trott1ydiscuss
OpenAI Pitched White House on 5-7 5GW datacenters bloomberg.com
3trott1ydiscuss

All submissions and comments on Hacker News.