Meta's agentic image model lands at No. 2 — and agents get smarter verifiers
A busy midweek: Meta Superintelligence Labs ships its first media models with full agentic plumbing, while the research side keeps tightening the screws on how agents verify, remember, and learn from the world. Here's what moved.
Meta Superintelligence Labs launches Muse Image and previews Muse Video with agentic generation capabilities
Muse Image debuting at No. 2 Arena Elo with tool use and self-refinement signals that agentic generation — not just better diffusion — is where media models are heading.
LLM-as-a-Verifier: Training-free verification framework scales along granularity, repetition, and criteria decomposition
Training-free verification that hits SOTA across four hard benchmarks reframes verification as a scaling axis in its own right — with RL feedback benefits on top.
CompactionRL trains long-horizon agents with context compaction via reinforcement learning
Jointly learning to summarize context while executing tasks is a practical path to long-horizon agents — and a 7-point SWE-bench gain already earned it a spot in GLM-5.2's training pipeline.
EdgeBench finds log-sigmoid scaling laws for agent learning from real-world environments
A log-sigmoid scaling law for agent environment learning — with speed doubling every three months — is the kind of empirical throughline that reshapes how we think about post-deployment progress.
Nemotron-Labs Audex-30B: Unified audio-text MoE LLM with no text regression
A single decoder handling ASR, TTS, translation, and speech-to-speech with minimal text regression shows the unified audio-text architecture is maturing fast.
First multiplayer world model for dynamic environments using latent diffusion on Rocket League gameplay
Real-time four-player world model generation at 20 FPS with stable long-horizon rollouts is a meaningful step for multi-agent coherence under complex dynamics.
Get the next edition in your inbox
Midweek and Sunday. Primary-sourced AI news paired with the guides that explain it.