Every paper, note, and technical report the lab has published. Most are long; many include code and reproductions. Sorted by date.

The Unreasonable Efficiency of a Single Gradient Step

When does distillation help reinforcement learning?

Can a Language Model Learn Facts Continually in Its Weights?

Post-Training Science for Supervised Fine-Tuning

Still: Amortized KV Cache Compaction in a Single Forward Pass

Towards infinite context windows: neural KV cache compaction