Daily Research Brief 2026-06-14

📊 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: calibration attacks on GNNs (The Confidence Trap), on-device assistant query recommendation (ToolRec), TVI-CoT interleaved chain-of-thought, and MDForge for agentic molecular dynamics pipelines.

1. Latest arXiv Papers

1. The Confidence Trap: Calibration Attacks for Graph Neural Networks — https://arxiv.org/abs/2606.08467

2. ToolRec: Calibrated Preference Alignment for Query Recommendation in On-Device Assistants — https://arxiv.org/abs/2606.08466

3. An Empirical Comparison of General Context-Free Parsers — https://arxiv.org/abs/2606.08465

4. TVI-CoT: Text-Visual Interleaved Chain-of-Thought — https://arxiv.org/abs/2606.08464

5. Iterating Toward Better Search: A Two-Agent Simulation Framework for Evaluating Agentic Search Architectures in E-Commerce — https://arxiv.org/abs/2606.12924

6. MDForge: Agentic Molecular Dynamics Pipeline Design under Sparse Simulator Feedback — https://arxiv.org/abs/2606.12916

7. Zero-Source LLM Hallucination Detection with Human-Like Criteria Probing — https://arxiv.org/abs/2606.12900

8. Efficient Transfer Learning for Low-Resource NLP Tasks