Field Notes

Dispatches from the work of building continual learning. Research, customer stories, and other artifacts.

Continual Grading

Field Report

Latest

Continual Grading

A few hours of expert examples teach a metagrader how the expert grades, and it writes a grader for every new task.

Author: Neil Kale
Published: November 2, 2026

What’s on the bookshelf

Field Report
Continual Grading
Neil Kale · November 2, 2026
Essay
Own Your Value Function: Introducing the Trajectory Platform
The Trajectory Team · November 2, 2026
Field Report
You Only Rollout Once
Nikhil Vemuri and Dian Ang Yap · November 2, 2026
Field Report
Enabling Frontier-Scale Training for the GLM-5.3 Family
Hersh Godse and Chuck Tang · October 2, 2026
Field Report
Intelligence Density
Dian Ang Yap · September 24, 2026
Field Report
Choosing the Trainable Geometry for Agentic RL
Neil Kale and Chuck Tang · September 15, 2026
Field Report
Continual Learning Cheap Enough to Meter
Dian Ang Yap · August 11, 2026
Field Report
Post-Training NVIDIA Nemotron 3 Ultra on Harvey LAB in Under 24 Hours
Jerry Chan · June 9, 2026
Field Report
Scaling SDPO
Neil Kale · June 2, 2026
Essay
Continual Learning: End of Frozen Software
Arjun Karanam · June 1, 2026
Field Report
Multi-LoRA Training for Continual Learning
Chuck Tang · May 30, 2026
Field Report
The Sovereign Path for Continual Learning: Early Results on Harvey LAB with NVIDIA Nemotron
Jerry Chan · May 29, 2026
Essay
The Pioneers of Continual Learning
Arjun Karanam · May 28, 2026