Daily Research Brief 2026-05-28

📊 Token usage: estimated from retrieval and writing scale.

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

Editor’s Note

HackerNews-led day: YouTube will auto-label AI-generated videos; Wall Street pays $25,000/day for AI trainers; OpenAI reportedly offered to buy Chrome amid Google antitrust; Micron became the first memory maker to top a $1T market cap; a Chinese humanoid robot launched at $12,500.

1. Latest arXiv Papers

1. CORE: Contrastive Reflection Enables Rapid Improvements in Reasoning — https://arxiv.org/abs/2605.28742

2. Entropy-Aware Masking for Masked Language Modeling — https://arxiv.org/abs/2605.28526

3. A Conflict-Aware Penalty and Statistical Loss Framework for LLM Alignment — https://arxiv.org/abs/2605.28575

4. Calibrating Conservatism for Scalable Oversight — https://arxiv.org/abs/2605.28807

5. A Matter of TASTE: Improving Coverage and Difficulty of Synthetic Data — https://arxiv.org/abs/2605.28556

6. B³D-RWKV: Triplet-Block Diffusion RWKV — https://arxiv.org/abs/2605.25969

7. Hera: Learning Long-Horizon Coordination for Device-Cloud Agents — https://arxiv.org/abs/2605.24598

8. CoD: Diffusion Foundation Model for Image Compression — https://arxiv.org/abs/2511.18706

9. Automated Benchmark Audit: An Agentic Framework for Evaluation Integrity — https://arxiv.org/abs/2605.08083

10. DynaSchedBench: Calibrated Dynamic Scheduling Benchmark — https://arxiv.org/abs/2605.27642