Latest

Graphical user interface (GUI) grounding, the ability to map natural language instructions to specific actions on graphical user interfaces, remains a critical bottleneck in computer use agent development. Current benchmarks oversimplify grounding tasks as short referring expressions, failing to...

Knowledge Base Question Answering (KBQA) aims to answer natural language questions using structured knowledge from KBs. While LLM-only approaches offer generalization, they suffer from outdated knowledge, hallucinations, and lack of transparency. Chain-based KG-RAG methods address these issues by...

Embodied Question Answering (EQA) requires agents to autonomously explore and understand the environment to answer context-dependent questions. Existing frameworks typically center around the planner, which guides the stopping module, memory module, and answering module for reasoning. In this paper,...

We introduce STAR (Stream Transduction with Anchor Representations), a novel Transformer-based model designed for efficient sequence-to-sequence transduction over streams. STAR dynamically segments input streams to create compressed anchor representations, achieving nearly lossless compression (12x)...

This paper introduces a comprehensive end-to-end pipeline for Optical Character Recognition (OCR) on Urdu newspapers. In our approach, we address the unique challenges of complex multi-column layouts, low-resolution archival scans, and diverse font styles. Our process decomposes the OCR task into...

Assessing Large Language Models (LLMs)’ underlying value differences enables comprehensive comparison of their misalignment, cultural adaptability, and biases. Nevertheless, current value measurement datasets face the informativeness challenge: with often outdated, contaminated, or generic test...

Recent advances in Large Reasoning Models (LRMs) have shown impressive capabilities in mathematical and logical reasoning. However, current LRMs rarely admit ignorance or respond with "I don’t know". Instead, they often produce incorrect answers while showing undue confidence, raising concerns about...

Reasoning-enhanced large language models (LLMs) explicitly generate intermediate reasoning steps prior to generating final answers, helping the model excel in complex problem-solving. In this paper, we demonstrate that this emerging generation framework offers a unique opportunity for more...

Steering methods have emerged as effective and targeted tools for guiding large language models’ (LLMs) behavior without modifying their parameters. Multimodal large language models (MLLMs), however, do not currently enjoy the same suite of techniques, due in part to their recency and architectural...

We introduce ${\rm C{\small LEVER}}$, a high-quality, curated benchmark of 161 problems for end-to-end verified code generation in Lean. Each problem consists of (1) the task of generating a specification that matches a held-out ground-truth specification, and (2) the task of generating a Lean...

Just-in-time Information Recommendation (JIR) is a service designed to deliver the most relevant information precisely when users need it, , addressing their knowledge gaps with minimal effort and boosting decision-making and efficiency in daily life. Advances in device-efficient deployment of...

Automatic prompt generation plays a crucial role in enabling general-purpose multi-agent systems to perform diverse tasks autonomously. Existing methods typically evaluate prompts based on their immediate task performance, overlooking the intrinsic qualities that determine their reliability. This...

Information extraction (IE) from Visually Rich Documents (VRDs) containing layout features along with text is a critical and well-studied task. Specialized non-LLM NLP-based solutions typically involve training models using both textual and geometric information to label sequences/tokens as named...

Textbook question answering (TQA) is a complex task, requiring the interpretation of complex multimodal context. Although recent advances have improved overall performance, they often encounter difficulties in educational settings where accurate semantic alignment and task-specific document retrieval...

Large Language Models (LLMs) show potential for complex reasoning, yet their capacity for emergent coordination in Multi-Agent Systems (MAS) when operating under strict swarm-like constraints-limited local perception and communication-remains largely unexplored. Existing benchmarks often do not fully...

Retrieval-Augmented Generation (RAG) has shown substantial promise in improving factual accuracy by grounding model responses with external knowledge relevant to queries. However, most existing RAG approaches are limited to a text-only corpus, and while recent efforts have extended RAG to other...

Chain-of-thought (CoT) reasoning boosts large language models’ (LLMs) performance on complex tasks but faces two key limitations: a lack of reliability when solely relying on LLM-generated reasoning chains and interference from natural language reasoning steps with the models’ inference process, also...

With the rapid advancement of global digitalization, users from different countries increasingly rely on social media for information exchange. In this context, multilingual multi-label emotion detection has emerged as a critical research area. This study addresses SemEval-2025 Task 11: Bridging the...

Language models encode and subsequently perpetuate harmful gendered stereotypes. Research has succeeded in mitigating some of these harms, e.g. by dissociating non-gendered terms such as occupations from gendered terms such as ’woman’ and ’man’. This approach, however, remains superficial given that...

Large language model (LLM) based agents have shown great potential in following human instructions and automatically completing various tasks. To complete a task, the agent needs to decompose it into easily executed steps by planning. Existing studies mainly conduct the planning by inferring what...

Google announced during Google I/O 2025 that it’s rolling out Project Mariner, the company’s experimental AI agent that browses and uses websites, to more users and developers. Google also says it’s significantly updated how Project Mariner works, allowing the agent to take on nearly a dozen tasks at...

Android Studio, the integrated development environment (IDE) for Android app developers, is getting an AI upgrade, Google announced at its developer conference, Google I/O 2025, on Tuesday. In addition to the rollout of the latest Gemini 2.5 Pro model, Android Studio is gaining a new “agentic AI”...

Last Friday, OpenAI introduced a new coding system called Codex, designed to perform complex programming tasks from natural language commands. Codex moves OpenAI into a new cohort of agentic coding tools that is just beginning to take shape. From GitHub’s early Copilot to contemporary tools like...

The AI models keeps the user in the loop by providing detailed source tracking, ensuring transparency, Microsoft said. Read more: Microsoft Discovery aims to power scientists with agentic AI

Can AI speed up aspects of the scientific process? Microsoft appears to think so. At the company’s Build 2025 conference on Monday, Microsoft announced Microsoft Discovery, a platform that taps agentic AI to “transform the [scientific] discovery process,” according to a press release provided to...

OpenAI has announced the launch of a research preview of Codex, a cloud-based coding AI agent for software engineers.

Y Combinator startup Firecrawl is back on the hunt for AI agent employees, after a previous attempt to hire one didn’t go as planned.

If you want to know where AI is headed, this year’s Google I/O has you covered. The company’s annual showcase of next-gen products, which kicked off yesterday, has all of the pomp and pizzazz, the sizzle reels and celebrity walk-ons, that you’d expect from a multimillion-dollar marketing event. But...

In her new book Empire of AI, journalist Karen Hao chronicles the anxieties around the OpenAI office in its early days.

Popular battle royale game Fortnite has finally returned to the U.S. App Store amid game maker Epic Games’ lengthy legal skirmish with Apple. As of Tuesday, Fortnite is also available on the Epic Games Store and AltStore in the EU. It’ll show up in App Store searches soon, Epic said in a post on X....

Magentic-UI, new from Microsoft Research, is an open-source research prototype of a human-centered AI agent, designed to work with people to complete complex, web-based tasks in real time over a web browser. The post Magentic-UI, an experimental human-centered web agent appeared first on Microsoft...

Mind map for Google I/O 2025 with various topics branching from it.

In this discussion, Matthew Lungren, Jonathan Carlson, Smitha Saligrama, Will Guyman, and Cameron Runde explore how teams across Microsoft are working together to generate advanced AI capabilities and solutions for developers and clinicians around the globe. The post Collaborators: Healthcare...

Introducing our novel standardized framework for evaluating enterprise AI assistants across both text and voice.

At Google I/O ‘25, we introduced a new Google Cloud and NVIDIA developer community, a dedicated forum to connect with experts from both companies.In the coming weeks, co…

We’re doing cutting-edge research to build the most helpful AI that’s more intelligent, agentic and personalized.

A dark background with rainbow ombre at the top, with text saying Google AI Ultra subscription, and tiles with text including Gemini app and Flow. Underneath that, neon jellyfish float over Earth.

Found the perfect outfit but you’re not quite sure how it’ll look on you? Google now lets you virtually try on garments using your own photo in four simple steps:Opt in …

New Gemini updates built directly into tools you use every day — like Gmail, Meet, Vids and Docs — are coming to Google Workspace.Gmail is getting personalized smart rep…

Today at I/O, we announced that AI Overviews in Google Search are launching in new countries and languages, bringing the helpfulness of AI to more people around the worl…