#rl

The latest AI and developer news about #rl, with the original source and a short summary.

Magic Announces Pretraining Research Achieving Performance Similar to DeepSeek V4 Pro Base with Approximately 50x Fewer FLOPs

Magic announced that a new pretraining recipe reached performance similar to DeepSeek V4 Pro Base with approximately 50x fewer FLOPs, and that training scaled up by approximately 10x subsequently outperformed public base models in perplexity evaluations.

·2026.09.08 09:00

Miles v0.1: Production-Level Post-Training (20-minute read)

RadixArk released Miles v0.1, a full-stack system for frontier-scale RL training.

·1·2026.08.19 09:00

SWE-1.7: Frontier-Level Intelligence at a Much Lower Cost

Cognition unveiled SWE-1.7, a frontier-level intelligence model that maximizes cost efficiency through improvements to its RL pipeline.

·2026.07.09 02:00

Should You Train Your Own Model: The Pros and Cons of Post-Training for Application Companies

This analyzes why AI application companies should adopt post-training based on open-weight models to cut costs and optimize performance.

·2026.04.28 09:00

What We Learned This Week (20 min read)

Frontier models are harder to prevent **distillation** of than expected, and large-scale pretraining can easily collapse from small **bias** and **causality** violations.

·1·2026.04.17 09:00

Medical Model Trained with RL

A 32B medical LLM built with SeedRL was released, with thinking mode set as the default.

·2026.04.15 15:00