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

[2602.18749] Federated Reasoning Distillation Framework with Model Learnability-Aware Data Allocation

The paper presents the LaDa framework, addressing the challenges of data allocation in federated learning by enhancing model learnability...

[2602.18731] Beyond Description: A Multimodal Agent Framework for Insightful Chart Summarization

This article presents a novel framework for chart summarization using a multimodal agent approach, enhancing data accessibility and insig...

[2602.18724] Task-Aware Exploration via a Predictive Bisimulation Metric

The paper presents TEB, a Task-aware Exploration approach that enhances exploration in visual reinforcement learning by utilizing a predi...

[2602.18710] Many AI Analysts, One Dataset: Navigating the Agentic Data Science Multiverse

The paper explores how autonomous AI analysts using large language models can replicate diverse analytic outcomes on the same dataset, re...

[2602.18640] Decoding ML Decision: An Agentic Reasoning Framework for Large-Scale Ranking System

The paper presents GEARS, a novel framework for optimizing large-scale ranking systems by transforming the optimization process into an a...

[2602.18607] Feedback-based Automated Verification in Vibe Coding of CAS Adaptation Built on Constraint Logic

The paper discusses a novel approach to automated verification in CAS adaptation using vibe coding and feedback loops, demonstrating effe...

[2602.18582] Hierarchical Reward Design from Language: Enhancing Alignment of Agent Behavior with Human Specifications

The paper presents Hierarchical Reward Design from Language (HRDL), a framework to align AI behavior with human specifications through en...

[2602.18494] On the Dynamics of Observation and Semantics

The paper discusses the limitations of viewing semantics as static in visual intelligence, proposing a dynamic framework that integrates ...

Meet ASKB: A First Look at the Future of the Bloomberg Terminal in the Age of Agentic AI

The article introduces ASKB, a new AI-driven feature for the Bloomberg Terminal, highlighting its potential to enhance user experience an...

Alipay’s AI Payment and Health Apps Top 100 Million Users

Alipay's AI payment and health applications have each surpassed 100 million users, highlighting a significant trend in AI adoption among ...

X plans to combat AI-generated content while promoting Grok

X, formerly Twitter, plans to combat AI-generated content through new detection measures while promoting its Grok AI chatbot for post cre...

[R] Prompt Repetition Shows Null Result on Agentic Engineering Tasks (n=20, blind scored)

This article discusses a study on prompt repetition in engineering tasks using Claude Haiku 4.5 agents, revealing no significant improvem...

Prompt repetition adds zero accuracy to AI agents on engineering tasks

The article discusses how repeating prompts does not improve the accuracy of AI agents in engineering tasks, highlighting inefficiencies ...

I experimented with giving an AI agent a symbolic anatomy — soul, heart, brain, and shadow

The article explores an experiment where the author assigns symbolic anatomy—soul, heart, brain, and shadow—to an AI agent, reflecting on...

A Meta AI security researcher said an OpenClaw agent ran amok on her inbox  | TechCrunch

Meta AI researcher Summer Yue shares a cautionary tale about her OpenClaw AI agent, which mistakenly deleted her emails despite her comma...

[P] I built an AI alignment engine based on Thermodynamics instead of RLHF. It doesn’t just "refuse" unsafe inputs—it physically decouples from them.

The article discusses a novel AI alignment engine based on thermodynamics, proposing a framework that decouples unsafe inputs rather than...

[D] New Research Discord - Computational Psycholinguistics

A Reddit user is forming a research-focused Discord community for those interested in computational psycholinguistics, aiming to facilita...

ChatGPT spits out surprising insight in particle physics

The article discusses unexpected insights generated by ChatGPT in the field of particle physics, showcasing the potential of AI in scient...

Uncanny Valley: AI Researchers’ Resignations, Bots Hiring Humans, Evie Magazine’s Party | WIRED

This episode of 'Uncanny Valley' discusses AI researchers resigning over safety concerns, the controversial Rent-A-Human service hiring h...

OpenAI calls in the consultants for its enterprise push | TechCrunch

OpenAI is partnering with major consulting firms to enhance enterprise adoption of its AI technologies through a new initiative called th...