Daily Research Brief 2026-06-13

📊 Token usage: estimated from retrieval and writing scale.

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

Editor’s Note

The paper thread today: MemRefine compresses long-term agent memory, MaxProof scales math proof via generative-verifier RL, epiphenomenal CoT probing, and OpenMedReason for medical VLMs.

1. Latest arXiv Papers

1. Cross-Modal Masked Compositional Concept Modeling for Enhancing Visio-Linguistic Compositionality — https://arxiv.org/abs/2606.13288

2. MemRefine: LLM-Guided Compression for Long-Term Agent Memory — https://arxiv.org/abs/2606.13177

3. Beyond the Commitment Boundary: Probing Epiphenomenal Chain-of-Thought in Large Reasoning Models — https://arxiv.org/abs/2606.13603

4. ReSET: Accurate Latency-Critical NVFP4 Reasoning via Step-Aware Temperature Scaling — https://arxiv.org/abs/2606.13233

5. SkMTEB: Slovak Massive Text Embedding Benchmark and Model Adaptation — https://arxiv.org/abs/2606.13647

6. MaxProof: Scaling Mathematical Proof with Generative-Verifier RL and Population-Level Test-Time Scaling — https://arxiv.org/abs/2606.13473

7. OpenMedReason: Scientific Reasoning Supervision for Medical Vision-Language Models — https://arxiv.org/abs/2606.12169

8. From Prompts to Tokens: Internalizing Causal Supervision in Vision-Language Model for Multi-Image Causal Reasoning — https://arxiv.org/abs/2606.11745