All news
10671
Plan-and-Patch: Diffusion LLM Repairs Agent Plans Twice as Often as Qwen3-8B
The diffusion-based DreamReasoner-8B repairs agent plans with a 53.7% success rate versus 27.0% for Qwen3-8B, and after fine-tuning it generates plans 39–46% faster.
arXiv authors call for replacing a single annotation "truth" with a space of interpretations
An arXiv position paper argues for ending the reduction of labels to one answer and modeling the space of acceptable human judgments, from training through…
TGIT Translates Short Drone Commands and Lifts Navigation Success from 11% to 33%
The TGIT instruction translator leaves the frozen OpenFly navigator untouched, yet lifts mission success from 11.33% to 32.51% on real human commands.
Biohub expands Virtual Biology Initiative with $1.8 billion for AI data
Simon Willison releases ttok 0.4: the token counter gets --list-models
Snyk turns its internal support agent into a customer-facing feature, Snyk Assist
Bloomberg Analyst: Questions Over OpenAI's Revenue Accounting Won't Halt Demand for AI Infrastructure
Principal Global Investors Strategist Weighs In on the AI Capex Cycle and Market Sentiment
At the Milken Asia Summit in Singapore, Principal Global Investors' Seema Shah discussed how the AI capex cycle shapes investor sentiment. Here is the context and…
Anthropic Launches Free OSS Scanner to Hunt for Open Source Vulnerabilities
Anthropic has launched OSS Scanner — a free service for automatically finding vulnerabilities in open source projects. It uses Claude Mythos and other powerful…
SpaceX Demands Orbital Coordination After Dangerous Starlink Satellite Close Calls
Midjourney Adds Thinking Mode for V8.2 and Shared Folders in Alpha
Midjourney Tests Thinking Mode: A Rerun (Thinking) Button for Better Text Accuracy
NVIDIA Scales Bit-for-Bit Determinism in Megatron Core to 2,432 GPUs with 2% Overhead
OpenMobile-2 Open-Sources a Mobile Agent: 79.9 on AndroidWorld and 110 Apps
OpenMobile-2 ships with its environment, data and training recipes: 79.9 on AndroidWorld, 50.4 on MobileWorld, and SPA-Bench on real devices up from 31.9 to 59.6.
RealityBench: 26 VLMs Tested on Documenting Physical Work in Blender
A Blender benchmark of 11 environments shows top VLMs read text in frame but struggle with measurements and fine details. At frame acceptance, they match a single…