Researchers present VideoX-Qwen, an integrated framework combining scalable data construction with model training to advance general-purpose, instruction-driven video editing. The system utilizes a production pipeline that organizes specialized models to generate directional editing records, resulting in over 1.2 million high-quality samples across addition, removal, replacement, and attribute tasks.
A computational linguist and Hausa NLP specialist has introduced a new dataset designed for evaluating Large Language Models in the context of healthcare and safety within the Hausa language.
The FlyEye project has published the first compact neuron-level dataset alongside a browser and runtime developer kit. This release includes the Tubban/flyeye-escape-neuron-v1 dataset hosted on Hugging Face.
Adaption Labs has introduced "Invent a Dataset," a feature that generates structured, training-ready datasets directly from natural language task descriptions without requiring a seed corpus, predefined schema, or labeling guide. The tool is available via the Adaption app, Python SDK, and REST API, allowing users to download generated rows in JSONL, JSON, CSV, or Parquet formats.
The authors present TrialGPT 2.0, an AI-assisted clinical trial recommendation system designed for real-world deployment that assesses which trials warrant consideration based on patient needs and workflow priorities.
A new tutorial documents the step-by-step process of building and training a byte-level BPE tokenizer from first principles. The guide covers key implementation details such as Unicode-aware pretokenization, incremental pair-frequency updates, and using a max heap with lazy deletion.
The Open ASR Leaderboard has introduced its first Global South languages with two new evaluation sets: Monsoon en-IN (Indian English) and Monsoon hi-IN (Hindi). These additions address the lack of demographic diversity in current benchmarks, which have historically focused on European languages.
The author has released Calibra, a tool designed to identify poor demonstrations in robotics imitation learning datasets before they consume GPU resources.
Harmonic Frontier has published a preview of the Celtic Constellation Reference Sessions dataset, featuring aligned multitrack recordings of traditional and contemporary Celtic ensemble arrangements. The architecture consists of aligned pairs where each piece includes dry isolated stems, the same performance processed through a production chain, and the finished mix.
A complete personal archive of Konstantin Tsiolkovsky, comprising 51,008 sheets and 2,019 archival files, has been released as a CC0 dataset. The collection spans fifty years of work by the self-taught rocket theorist and includes handwritten manuscripts alongside typed copies.