Institution: VirtueAI | Authors: Zhaorun Chen, Xun Liu, Haibo Tong, Chengquan Guo, Yuzhou Nie arXiv Links arXiv | PDF AI summary Abstract A comprehensive platform and autonomous agent framework for evaluating and enhancing AI agent security through controlled red-teaming across multiple real-world domains and simulation environments. AI-generated summary 摘要：一个综合平台和自主代理框架，用于通过跨多个现实世界域和模拟环境的受控红队来评估和增强人工智能代理的安全性。 AI 生成的摘要 Abstract AI agents are increasingly deployed across diverse domains to automa...

Google's research team has introduced "A^2RD," a new AI architecture set to revolutionize long-form video generation. This breakthrough aims to fix common AI video issues like lack of consistency and disjointed narratives, paving the way for more natural and cohesive AI-generated videos. It's a significant step towards practical applications for longer content.

Google's research team has developed 'AutoTTS,' a new technology that automatically optimizes the performance of Large Language Models (LLMs). This innovative approach aims to improve AI model accuracy while simultaneously reducing operational costs, presenting a significant step forward in making advanced AI more efficient and accessible for various applications.

OpenAI

Follow

2d

As Large Language Models (LLMs) advance, the effectiveness of 'lexical search' is getting a fresh look. A new search agent, Pi-Serini, demonstrates how this older technique, when paired with powerful LLMs, can outperform modern dense retrieval methods in accuracy and recall. This suggests that simple, established search methods still hold significant potential for enhancing AI's capabilities.

Researchers from the Chinese University of Hong Kong introduced 'SLIM,' a new framework designed to significantly improve how AI agents manage skills for complex tasks. SLIM dynamically optimizes an agent's external skill set by evaluating contributions, retaining high-value skills, and retiring underperforming ones. This leads to more efficient task resolution and an average performance increase of 7.1% over existing methods, suggesting a potential shift in AI agent design.

A new research benchmark called 'InterLV-Search' has been released, aiming to measure AI's capability to search for information by combining text and images. This evaluation reveals the current limitations of AI systems in performing complex, intertwined multimodal searches, showing that even top-performing models are far from proficient.

Anthropic

Follow

4/18/2026

Anthropic just released Claude Opus 4.7, their latest AI model designed to supercharge developer workflows. This upgrade specifically targets advanced coding, high-resolution image processing, and long-term autonomous tasks. It’s not a complete overhaul but a focused boost to critical areas, promising more reliable and capable AI for real-world applications.

Hugging Face just launched 'ml-intern,' an open-source AI agent built on their smolagents framework. This tool automates the entire post-training workflow for large language models, tackling everything from literature reviews and dataset discovery to training execution and iterative evaluation. It's a game-changer for SMBs looking to fine-tune LLMs, potentially slashing development costs and time.

Institution: BAIDU | Authors: Haotian Zhao, Songlin Zhou, Yuxin Zhang, Stephen S. -T. Yau, Wenyu Zhang arXiv Links arXiv | PDF AI summary Abstract A novel supervision-free credit assignment method for reinforcement learning in language model agents that adapts entropy dynamics at the response level to improve exploration-exploitation trade-offs and task performance. AI-generated summary 摘要：一种新颖的无监督信用分配方法，用于语言模型代理中的强化学习，该方法在响应级别上调整熵动力学，以改善探索-利用权衡和任务性能。 AI 生成的摘要 Abstract Reinforcement learning (RL...

OpenAI

Follow

4/22/2026

OpenAI introduced a new paradigm and product today that is likely to have huge implications for enterprises seeking to adopt and control fleets of AI agent workers. Called "Workspace Agents," OpenAI's new offering essentially allows users on its ChatGPT Business ($20 per user per month) and variably priced Enterprise, Edu and Teachers subscription plans to design or select from pre-existing agent templates that can take on work tasks across third-party apps and data sources including Slack, Goog...

Anthropic

Follow

5/5/2026

Just two months ago, researchers at the Data Intelligence Lab at the University of Hong Kong introduced CLI-Anything, a new state-of-the-art tool that analyzes any repo’s source code and generates a structured command line interface (CLI) that AI coding agents can operate with a single command. Claude Code, Codex, OpenClaw, Cursor, and GitHub Copilot CLI are all supported, and since its launch in March, CLI‑Anything has climbed to more than 30,000 GitHub stars. But the same mechanism that makes ...