Tuesday, July 14, 2026

1. MM-ToolSandBox: A Unified Framework for Evaluating Visual Tool-Calling Agents
   T2arXiv AI/CL/LG（论文）

   MM-ToolSandBox is a new benchmark and evaluation framework from Apple for visually grounded tool-calling agents. It provides a stateful execution environment with 500+ tools across 16 domains, supporting multi-image, multi-turn tasks with realistic conversational phenomena such as goal revisions and error corrections. An automated pipeline generated 258 human-verified nominal scenarios and 50 UI variants. Evaluating 12 models (from 4B open-weight to frontier proprietary systems), the best model achieves under 50% success rate, with 53% of failures stemming from visual information extraction errors rather than planning. Small models falter in decision-making while large models struggle with perception, indicating different research directions for models at different scales.

2. Finterm.ai: A Bloomberg-Style Financial Data Terminal for Claude Code
   T2Hacker News 热门（Algolia · 中文化）

   Finterm.ai is a financial data interface designed for AI agents like Claude Code, providing timestamped ticker data bundles, SEC filing diffs, insider and institutional flows, and multi-source deep research—all in a single call, aiming to strengthen the factual foundation for trading decisions.

3. LoRA-Based Cascaded Multimodal Fusion for Action Recognition in Medical Training Environments
   T2arXiv AI/CL/LG（论文）

   This paper proposes a cascaded Low-Rank Adaptation (LoRA)-based multimodal fusion framework for action and activity recognition in healthcare-oriented training environments. The architecture combines parameter-efficient modality-specific adaptation with sequential fusion, enabling modalities to be integrated in stages without retraining previously learned components. It first integrates more closely related modalities and then incorporates additional heterogeneous modalities, supporting scalable adaptation across datasets with different modality sets. Evaluated on the NurViD and Nurse Training datasets, preliminary results show that the cascaded fusion strategy improves over individual modality models and provides competitive performance relative to previously reported baselines. The findings indicate that cascaded LoRA-based fusion is a promising parameter-efficient approach for integrating heterogeneous modalities in medical training action recognition tasks.

—

1. New York State Halts Approval of All New Data Centers, First State to Do So
   T2TechCrunch：AI（RSS）

   New York Governor Kathy Hochul has halted approvals for large data centers, arguing that the AI-driven construction boom should not come at the expense of higher electricity costs, water supplies, or local control. The moratorium applies to new facilities above a certain scale and energy threshold, while existing projects continue. This could set a precedent for other states.

1. Inside the Unfair Judge: A Mechanistic Interpretability Account of LLM-as-Judge Bias
   T2arXiv AI/CL/LG（论文）

   This paper proposes that scoring biases in LLM-as-judge are not only present at the input-output level but can also be explained at the representation level in the hidden state. Across 7 judges, 7 bias types, and 9 benchmarks, they find that biased inputs are displaced along a low-dimensional subspace in activation space; causal steering along this subspace bidirectionally modulates scores; and a simple linear projection onto the subspace predicts judge failures on unseen benchmarks, substantially outperforming text-based methods.

—