Data-centric debugging for teams training neural nets [P]

<!-- SC_OFF --><div class="md"><p>Disclosure: I built LabelSets (labelsets.ai). Sharing the technical approach behind how we score dataset quality.</p> <p>THE PROBLEM</p> <p>Most

<!-- SC_OFF --><div class="md"><p>Been working on a weight divergence trajectory curvature approach to detecting neural network training instability. Treats weight updates as

<!-- SC_OFF --><div class="md"><p>We have been exploring a project around post-training infrastructure, a minimalist tool that does one thing really well:<br /> Make post-training

<table> <tr><td> <a href="https://www.reddit.com/r/MachineLearning/comments/1t71y36/what_should_a_pytorch_training_endofrun/"> <img alt="What should a PyTorch training end-of-run

<!-- SC_OFF --><div class="md"><p>Every time one of our agents made a bad call in production, we had the same problem. Logs showed what happened. Nothing showed why it decided to

<table> <tr><td> <a href="https://www.reddit.com/r/MachineLearning/comments/1uga687/a_debugger_for_rl_reward_functions_that_detects/"> <img alt="A debugger for RL reward functions

<!-- SC_OFF --><div class="md"><p>We have gates for code, infrastructure, deployment and model performance.</p> <p>But when it comes to the actual training artifact, the decision

<!-- SC_OFF --><div class="md"><p>The BDH (Dragon Hatchling) paper (arXiv:2509.26507) describes a Hebbian synaptic plasticity mechanism where model weights update during