Daily Research Brief 2026-06-09

📊 Token usage: estimated from retrieval and writing scale.

Covers the latest AI research, open source and industry moves, updated daily.

Editor’s Note

The skill layer explodes on GitHub — andrej-karpathy-skills, mattpocock/skills and hermes-agent lead; Huawei Cloud launched Agentic AI RLaaS; Microsoft Build debuted 7 fully in-house MAI models; OpenAI confirmed Codex merges into ChatGPT.

1. Latest arXiv Papers

1. Token-Level LLM Collaboration via FusionRoute — https://arxiv.org/abs/2606.08106

2. PACE: Anytime-Valid Acceptance Tests for Self-Evolving Agents — https://arxiv.org/abs/2606.08098

3. When Does Delegation Beat Majority? A Delegation-Based Aggregator for Multi-Sample LLM Inference — https://arxiv.org/abs/2606.08093

4. A Multi-Modal Agentic Co-pilot for Evidence-Grounded Computational Pathology

5. PhysX-Anything: Single Image to Physically Plausible 3D Assets

6. Token-Level Routing for Heterogeneous LLM Collaboration

7. Self-Correcting Multi-Agent System for Complex Reasoning

8. Multimodal Understanding for Long-Form Video Content