Digest
This week in open AI: flow fixes, world models, and export unlocks
Researchers stabilize flow-model RL and world-model hallucinations, diffusion LMs triple throughput, export controls lift on Claude Mythos 5, and uncensored Qwen fine-tunes flood HuggingFace.
ByAlex Sokoloff·
Training stability took center stage this week as teams tackled image-quality drift in flow models and hallucination modes in world simulators. Diffusion language models pushed token throughput past nine per pass, while real-time world simulation hit consumer GPUs. Meanwhile, export policy shifted: the US cleared Claude Mythos 5 for broad distribution even as Anthropic tightened filters on Fable 5, and the open-weight community released a wave of uncensored Qwen variants within 24 hours.
In this digest
- NormGuard preserves image quality in flow-model RL fine-tuning by capping velocity inflation — A simple velocity-norm penalty stops RL fine-tuning from wrecking image quality in flow generators—no architecture changes required.