AI Agents

Deepmind's 'AI Agent Traps' Paper Maps How Hackers Could Weaponize AI Agents Against Users

Google Deepmind's "AI Agent Traps" paper maps 6 attack types targeting autonomous AI agents, with exploit rates reaching 86% in tests.

Agentic AI in Beauty: How ChatGPT Is Reshaping Discovery, Trust, and Conversion

Agentic AI is transforming beauty shopping, shifting discovery from search to intent-driven recommendations where relevance, trust, and c...

Claude, OpenClaw and the new reality: AI agents are here — and so is the chaos

[2602.18600] MapTab: Can MLLMs Master Constrained Route Planning?

The paper introduces MapTab, a benchmark for evaluating Multimodal Large Language Models (MLLMs) on constrained route planning tasks, hig...

[2602.18985] InfEngine: A Self-Verifying and Self-Optimizing Intelligent Engine for Infrared Radiation Computing

InfEngine is an innovative autonomous engine designed to enhance infrared radiation computing by automating workflows, achieving a 92.7% ...

[2602.18584] GIST: Targeted Data Selection for Instruction Tuning via Coupled Optimization Geometry

The paper presents GIST, a method for targeted data selection in instruction tuning, improving efficiency by aligning training gradients ...

[2602.18981] How Far Can We Go with Pixels Alone? A Pilot Study on Screen-Only Navigation in Commercial 3D ARPGs

This study explores the effectiveness of screen-only navigation in 3D ARPGs, demonstrating how visual affordances can guide gameplay, whi...

[2602.18581] Learning Beyond Optimization: Stress-Gated Dynamical Regime Regulation in Autonomous Systems

The paper explores a novel framework for autonomous systems that enables learning without explicit objectives, focusing on self-regulatio...

[2602.18968] Robust and Efficient Tool Orchestration via Layered Execution Structures with Reflective Correction

This article presents a novel approach to tool orchestration in agentic systems, emphasizing a layered execution structure that enhances ...

[2602.18531] Deep Reinforcement Learning for Optimizing Energy Consumption in Smart Grid Systems

This paper explores the use of Deep Reinforcement Learning (RL) combined with Physics-Informed Neural Networks (PINNs) to optimize energy...

[2602.18960] Modularity is the Bedrock of Natural and Artificial Intelligence

The paper discusses the importance of modularity in both natural and artificial intelligence, highlighting its role in efficient learning...

[2602.18956] INDUCTION: Finite-Structure Concept Synthesis in First-Order Logic

The paper introduces INDUCTION, a benchmark for finite structure concept synthesis in first-order logic, focusing on generating logical f...

[2602.18528] Audio-Visual Continual Test-Time Adaptation without Forgetting

The paper presents a novel method, AV-CTTA, for audio-visual continual test-time adaptation that minimizes catastrophic forgetting while ...

[2602.18947] (Perlin) Noise as AI coordinator

The paper explores using Perlin noise as a coordinator for AI in large-scale game environments, addressing challenges in balancing behavi...

[2602.18523] The Geometry of Multi-Task Grokking: Transverse Instability, Superposition, and Weight Decay Phase Structure

This article explores the geometric analysis of multi-task grokking in machine learning, detailing five key phenomena observed during tra...

[2602.18943] High Dimensional Procedural Content Generation

The paper introduces High-Dimensional Procedural Content Generation (HDPCG), a framework that enhances gameplay mechanics by treating non...

[2602.18940] DREAM: Deep Research Evaluation with Agentic Metrics

The paper presents DREAM, a framework for evaluating Deep Research Agents, addressing challenges in assessing research quality through ag...

[2602.18884] TPRU: Advancing Temporal and Procedural Understanding in Large Multimodal Models

The paper introduces TPRU, a dataset aimed at improving temporal and procedural understanding in Multimodal Large Language Models (MLLMs)...

[2602.18843] ABD: Default Exception Abduction in Finite First Order Worlds

The paper introduces ABD, a benchmark for default-exception abduction in finite first-order worlds, evaluating LLMs on their ability to d...

[2602.18493] Learning to Remember: End-to-End Training of Memory Agents for Long-Context Reasoning

The paper presents the Unified Memory Agent (UMA), an end-to-end reinforcement learning framework designed for long-context reasoning, en...

[2602.18812] GenPlanner: From Noise to Plans -- Emergent Reasoning in Flow Matching and Diffusion Models

The paper presents GenPlanner, a novel approach to path planning in complex environments using generative models, specifically diffusion ...

[2602.18773] LAMMI-Pathology: A Tool-Centric Bottom-Up LVLM-Agent Framework for Molecularly Informed Medical Intelligence in Pathology

The LAMMI-Pathology framework proposes a novel tool-centric approach for enhancing molecularly informed medical intelligence in pathology...

[2602.18764] The Convergence of Schema-Guided Dialogue Systems and the Model Context Protocol

This paper discusses the convergence of Schema-Guided Dialogue Systems and the Model Context Protocol, proposing five foundational princi...