uberdavid

4 ms·

1.
by uberdavid 5mo ago
Author here. DeepSeek-V4 replaced multi-domain RL with a specialist-then-distill pipeline: train domain experts independently, merge through on-policy distillation. This post connects that production decision to three recent papers (Neural
2.
Random Machines: Why the Optimizer Is the Least Important Part of Deep Learning(sotaverified.org)
1 points by uberdavid 6mo ago|1 comments
3.
by uberdavid 6mo ago
Author here. The core idea is that when you train the same model with different random seeds, both reach the same accuracy but disagree on ~10% of predictions. The reason connects three well-established results (loss landscape geometry, the
4.
by uberdavid 6mo ago
The system card directly compares to Opus 4.6 and other frontier models on the same evals. Cybench went from ~75% to 100%, Firefox exploitation from 1 bug unreliably to 4 bugs reliably. It's true there are many capable coding models ou
5.
How RL Reward Hacking Made Claude Mythos a Zero-Day Hunter(uberdavid.substack.com)
2 points by uberdavid 6mo ago|2 comments
6.
The Dark Factory Harness: From Autonomous Hill-Climbing to Autonomous Research(sotaverified.org)
2 points by uberdavid 6mo ago|1 comments
7.
by uberdavid 6mo ago
Author here. This is a synthesis of Karpathy's autoresearch (the experiment loop) and OpenAI's harness engineering post (the environment design) applied to ML research with 5 practical design principles. The core idea is that afte
8.
by uberdavid 6mo ago
Thank you! The progress on research agents is exciting, but understanding what papers are reproducible on different datasets and architectures is often the bottleneck.
9.
SOTAVerified the open verification layer for ML research(sotaverified.org)
2 points by uberdavid 6mo ago|3 comments
10.
by uberdavid 6mo ago
Hi HN, I'm David, an ML researcher at Meta. I built SOTAVerified as an independent project after Papers with Code shut down last year and took 575k papers worth of benchmark data with it. SOTAVerified inherits that dataset (658k papers