Datasets

AIAI-generated summary Programmers uploaded a newly released male fruit fly connectome dataset to an artificial intelligence model to train it for tasks like playing video games. Gerald M. Rubin from the Howard Hughes Medical Institute notes that this unexpected application draws public interest to the research.

AIAI-generated summary The platform has introduced Memories Upload as a capture network that converts first-person video of everyday work into training data for robots. The company states that the network adds new footage every month to provide robotics labs with a growing supply of real-world data.

AIAI-generated summary Ant Group embodied AI company Robbyant has open-sourced LingBot-Touch, a tactile foundation model designed to give robots advanced touch perception using a dataset of nearly 5,000 hours of real-world tactile data. Robbyant says the model supports magnetic tactile, visual tactile, and piezoresistive sensing, which could improve fine manipulation tasks such as turning pages, unscrewing bottle caps, and identifying Mahjong tiles.

AIAI-generated summary Memories.ai is paying gig workers to supply first-person video data to help train robots to perform everyday human tasks. The startup uses Nvidia technology to process and label the video footage for robotics developers and artificial intelligence labs.

AIAI-generated summary LangChain explains eval-driven development as a process using a consistent set of metrics to measure AI agent performance and guide decision making. LangChain says the approach borrows from test-driven development and breaks down into defining agent behavior, building a dataset, and writing an evaluator.

AIAI-generated summary Elith and NVIDIA have released Nemotron-Personas-Finance-Japan as an open-source persona dataset paired with official Japanese statistics and financial data. NVIDIA says the dataset was built using NeMo Data Designer to expand the environment for developing and evaluating AI tailored to local requirements.

AIAI-generated summary Unite.AI explains that class imbalance occurs when a target class is much rarer than another, allowing standard training and aggregate metrics to neglect important cases. The guide outlines operational boundaries and metrics to prevent high accuracy from concealing the systematic neglect of minority classes in statistical learning.

AIAI-generated summary AI labs such as Anthropic and OpenAI are reportedly paying media companies fees ranging from $350K to $1.8M for work records to train better models. Micro1 CEO Ali Ansari says that labs, enterprises, and robotics companies currently pay about $40 to $50 billion annually to acquire data.

AIAI-generated summary Hugging Face is introducing FoundationWorlds as a framework for creating controlled, simulated environments to study robotics. The first foundation world is ToyBlocks, an open simulation framework built around a robot arm and blocks on a tabletop.

AIAI-generated summary The National Museum of Korea uses artificial intelligence initiatives and digital technologies to preserve cultural heritage and improve visitor experiences. The digital museum division created a high-resolution dataset of the Denny Taegeukgi and plans further digitization projects.

AIAI-generated summary Egyptian and Nigerian English have joined the Open ASR Leaderboard through a new collaboration with Liva AI. Liva AI says it collects multimodal interaction data through consumer products used by real people.

AIAI-generated summary @huggingface announced that its open-source ML Intern is now available as a Slack bot for user workspaces. @huggingface states that users can evaluate models, build datasets, and fine-tune models on the fly.

AIAI-generated summary Collinear AI releases CWE-bench v2, a public set of 100 tasks designed to test the limits of frontier cyber capabilities. Collinear AI says the tasks cover 91 different CWEs and span a wide range of vulnerability classes.

AIAI-generated summary Artificial Analysis has released AA-Omniscience version 1.1 to update its knowledge and reasoning benchmark with a refreshed public set and ambiguity fixes. Artificial Analysis says the update features a revised 3,000-question dataset and a newly calibrated public set on @huggingface.