Claude Fable 5 does not permit zero-data-retention contracts and retains all prompts and outputs for 30 days for security purposes, even where organizations have ZDR agreements with older Claude models.
Aligning router rows with the principal singular directions of their associated expert matrices improves the efficiency and stability of Mixture-of-Experts models.
Anthropic calls for an aviation-like regulatory authority or commissioned private auditors to examine AI models for critical risks before their release.
The Claw-SWE-Bench framework demonstrates that adapter design is critical for code agents: with a minimal adapter, OpenClaw achieves 19.1% Pass@1, with a complete adapter 73.4%.
Arbor enables AI-driven research through systematic hypothesis management and achieved an average of 2.5x higher improvements than existing code models on six test tasks.
Arbor coordinates autonomous AI agents via persistent hypothesis trees and achieved 2.5× better results than Codex and Claude Code on six research tasks.
Bebop uses rejection sampling and TV loss optimization to maintain stable MTP acceptance rates during RL training and accelerates rollouts by up to 1.8x.
