Transforms source code into verifiable, reusable agent skills by extracting atomic operations, workflows, and recurring patterns and validating them via source-body-blind reconstruction. Produces CodeSkillBank (1,006,822 accepted records from 19,769 GitHub repos) and yields ~11.7% average downstream improvement.
Provides layered code pretraining corpora (L2 ~400B tokens, L3 ~150B tokens) across 11 languages by filtering ~192M public GitHub repositories into standardized files, algorithmically relevant selections, and implementation-grounded programming exercises. Includes per-file metadata (role, algo relevance, quality) and serialized task records; released under Apache-2.0.
Provides a bilingual Chinese–English corpus for LLM training covering pretraining, capability-oriented midtraining (16K–256K long contexts), and supervised fine-tuning. Includes ~4.2T pretrain tokens, ~600B midtrain tokens, and ~4.57M SFT samples; sources span web, PDFs/OCR, code, math, QA, and agentic trajectories under mixed upstream licenses.
Provides ~86K verifiable-reward RL training samples across Math, Knowledge (STEM), Long-Context, and Code for post-training LLMs; each sample includes a ground-truth and verifier-friendly JSONL format for stable reward signals. Note: code tasks require an external sandbox to execute tests.
Selected ~245B-token corpus of algorithmically relevant code across 11 languages from The Stack v3, produced by file-level language-adaptive selection (UltraData-Code-L2-Classifier). Concentrates problem-solving implementations for LLM code pretraining while preserving broad language coverage; released under ODC-By.
Snapshot delivery of arXiv metadata, submission files and rendered documents in multiple Parquet configs (metadata, paper_text, latex, source, pdf, ps). Includes ~3.15M papers, full-text TeX assemblies and indexes to fetch large assets for training, retrieval and analysis.
Simulation-ready home dataset for embodied AI: CAD-based household scenes with configured physical properties and metadata, plus 1,000 robot trajectory episodes (RGB-D, HDF5/USDZ) for simulation training and evaluation under CC BY-NC-SA 4.0.
Provides 1,274 hours of head-mounted egocentric video paired with seven-point IMU arm tracking (24 Hz orientation; raw accel/gyro/mag on a subset), packaged for embodied-AI and egocentric-vision research. Key features: torso-relative pose via chest reference, separate Parquet IMU repo for efficient joins, CC-BY-4.0 license; heavy class skew and limited contributor diversity are important constraints.
Provides 22 curated Three.js-based cases for code-guided video generation across eight application categories. Each case bundles an editable scene (Three.js program), a 960×544 proxy video, a 1536×864 reference image, and a plain-text prompt for controlled video-generation evaluation.
Benchmark for typed probabilistic decisions: given one shared state, a model answers multiple typed questions at once and returns full probability distributions (noul/choice/score). Contains four workflows with train/test splits and soft gold labels from teacher samples, designed to evaluate accuracy, calibration, and latency trade-offs.
Provides transcripts, translations, and multi-task text labels for speech across 48 languages and about 463,000 hours. The release contains manifests only; audio must be obtained from the original data providers.
Provides a complete benchmark and training release for native audio‑visual dialogue: 2,800 synthesized audio‑visual dialogues, tiered rubrics, scoring code, ~28GB of 1080p media, and an RL reward recipe to evaluate and train omni models that take video+audio and return text.