Latest stories
Qwen-Planner-Agent: A Closed-Loop AI-for-AI Framework for Mobile Planner Agents
A new research paper introduces Qwen-Planner-Agent, a closed-loop AI-for-AI framework that connects data production, model trainin...
Hugging Face Paper Proposes Agent-Editing World Model to Curb Task-State Contamination in LLM Agents
The Agent-Editing World Model (AEWM) shifts world modeling for LLM agents away from predicting tool responses and toward modeling...
OmniEcho Brings Spatial Audio Understanding to Embodied AI Agents
Hugging Face researchers introduce OmniEchoBench, a unified benchmark for spatial audio-visual perception and navigation, alongsid...
Hugging Face Paper Introduces WROP, a 150-Task Dataset for Teaching Object Permanence to World Models
A new research paper presents WROP, a cognitive-science-inspired data infrastructure of 150 tasks and a 1.5M-sample corpus designe...
Google DeepMind Unveils Gemini 3.8 Live with Live Avatar for Enterprises
Google DeepMind adds a real-time visual persona to its live dialogue models, pairing low-latency video with speech for enterprise...
Ringg's AI Agents Resolve Up to 65% of Customer Calls Using OpenAI's GPT-5.6
Ringg is using OpenAI's GPT-5.6 to power multilingual AI agents across voice, chat, WhatsApp, and web, resolving up to 65% of cust...
Shanghai AI Lab Unveils InternW0, a Physical World Model for Real-World Robot Interaction
Shanghai AI Laboratory introduces InternW0, the first model in its InternW physical world model series, combining an asymmetric vi...
Hugging Face Paper Argues x0-Prediction Beats Velocity Matching for High-Dimensional Latents
A new paper on Hugging Face examines how finetuning visual encoders for reconstruction shrinks the effective dimensionality of the...
Hugging Face Paper Targets Cross-Attention Imbalance in Joint Video Generation
A new paper on Hugging Face identifies an asymmetry in how joint multimodal diffusion transformers link video with companion modal...
WhatWorkedBench Tests Whether AI Agents Truly Understand Their Experiments
Hugging Face researchers introduce WhatWorkedBench, a benchmark measuring whether AI research agents can predict how changing a co...
VHD-Play Generates 3,300 Agentic RL Environments from Solved Math Models
A new Hugging Face paper, VHD-Play, flips environment generation on its head by solving a mathematical model first, then rendering...
Hunyuan-A13B: Tencent's Open MoE Model Activates Only 13B of 80B Parameters
Hugging Face has published the technical report for Hunyuan-A13B, an open-source Mixture-of-Experts language model with 80 billion...