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.
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.
FinSkillBench is a new benchmark for testing AI agents on portfolio construction, risk management, and fundamental analysis. Across nine models, curated procedural skills raised mean performance from 0.366 to 0.528, while self-generated skills delivered little benefit.
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.
This paper frames context acquisition as active inference, enabling agents to choose among clarifying questions, retrieval, tool calls, task actions, or stopping. It introduces an Optimal Question Asking benchmark and evaluates language models on tasks involving 25 to 300 candidates, as well as clarification before generation and prompt optimization.
An independent builder pretrained a 1.02-billion-parameter Kimi-K3 replica on 5 billion decontaminated tokens for roughly $250. The model achieved 33.4% on HellaSwag, exceeding the reported 28% score of GPT-2 124M despite receiving no instruction tuning.
Moonshot and ZAI are reportedly approaching the performance of models from OpenAI and Anthropic. The models can reportedly build passable websites at a cost 75% lower than Claude, highlighting potential advantages in cost-sensitive coding workflows.
Claude is expanding into protein design, adding a scientific research capability to its existing AI portfolio. The update suggests potential use in computational biology and biotech workflows, though the article excerpt does not provide implementation or benchmark details.
This position paper argues that AI leaderboards underserve the Global South because of weak governance, conflicts of interest, and inflexible metrics—not a lack of data. Using India as a case study, it recommends independently governed regional leaderboards that incorporate benchmarks such as IndicSUPERB, MILU, and LAHAJA.