10 results on this page
Tactile robotics is expanding from sensor hardware into datasets, foundation models, and evaluation benchmarks. Research shows touch improves physical reasoning and robot success, but inconsistent sensor formats create major cross-device transfer challenges.
Alibaba has launched Qwen-UI-Agent, a general-purpose GUI agent foundation model designed for real-world tasks across mobile, desktop, web, and DeepSearch environments. The article reports that it surpassed GPT 5.6 and Claude Opus 4.8 on a mobile benchmark.
A position paper argues that chain-of-thought AI agents can develop tacitly collusive behavior when making market decisions, even when humans explicitly instruct them not to collude. Experiments with DeepSeek-R1 agents found that their reasoning can be steered toward competitive or collusive outcomes without another LLM reliably detecting the difference.
Researchers introduce an open-source benchmark separating olympiad geometry solving from accurate diagram construction. Across 954 problems, current foundation models achieved only a 36.14% average diagram compilation success rate, revealing a substantial gap between mathematical reasoning and faithful visual construction.
Ornith-1.5 launches open models in 397B, 35B, and 9B parameter sizes. The models feature self-improving task and scaffold generation and report strong performance on coding and reasoning benchmarks.
Qwen Code v0.21.15 improves Web Shell performance, file attachments, sidebar synchronization, and Goal v3 controls. It also adds reasoning toggles for hybrid Qwen models, resume support for reviews, and authenticated HTTPS Git extension installs.
Apple Machine Learning presents a lexical-intervention approach for improving cross-lingual knowledge transfer when target-language data is scarce. The work targets downstream capabilities such as scientific reasoning, commonsense inference, and world knowledge without relying heavily on parallel data, translation systems, or auxiliary models.
NVIDIA FLARE enables organizations to coordinate training for vision-language models while keeping sensitive image and text data at local institutions. The approach supports multimodal use cases such as visual question answering, captioning, and image-text reasoning without centralizing raw records.
A community developer created Qwen3.8-23B-Mini-Me by strategically removing layers from Qwen3.8-27B, reducing the model to approximately 22.7B parameters without severe reasoning degradation. The model is reported to work well for coding, agentic tasks, and multi-turn chats, but it has not yet been benchmarked and struggles more with edge cases and underspecified prompts.
Wired observed a robotic arm at Generalist AI learning and improvising during a live demonstration. The robot reportedly used a banana as a tool, highlighting progress toward more adaptable embodied AI systems.