Researchers propose a new AI framework that integrates visual perception and task planning within a single computational graph, maintaining a continuous soft symbolic state and allowing gradients from the planning objective to update perception parameters.

A new framework for selecting pre-trained vision-language models for specific downstream tasks is proposed, using layer-wise conductance and directional conductance divergence to improve performance and outperform state-of-the-art baselines. This development is relevant to people building and operating AI agents as it addresses the challenge of selecting the optimal model for a given task, which is crucial for efficient and effective AI deployment.

A proposed AI architecture for multi-domain decision-making in the Brazilian Armed Forces is presented, aiming to support autonomous and auditable decision-making across the Navy, Army, and Air Force. The architecture integrates data sources, tools, and execution, with a focus on security and permission safeguards for responsible AI employment.

Researchers propose SAIGE, a lightweight mechanism for multi-agent collaboration that achieves a favorable trade-off between context efficiency and task performance. Experiments show that SAIGE is effective in long-horizon, complex tasks, but scaling the agent pool does not consistently improve outcomes, suggesting that more agents do not necessarily make a system more intelligent.

Researchers introduced Generalist-Specialist-Mixture-of-Experts (GS-MoE) architecture for multimodal medical imaging, balancing modality-specific specialization with cross-modal shared representations. GS-MoE performed well on a dataset of 1.35M images, recovering detection of rare pathologies with per-class gains up to +0.60 F1. This may inform the development of AI agents for medical imaging tasks.

A new benchmark, Blindspot, is introduced for evaluating the safety and refusal calibration of long-horizon tool-using AI agents. It assesses agent behavior through adaptive adversarial interaction and stateful tool execution, providing a live-simulation framework for evaluating 13 LLMs and revealing substantial differences in safety-utility calibration across models.