Category

Multi-Benchmark LLM Agent Traces

Provides 1,781 OpenTelemetry execution traces of LLM-powered agents across six benchmarks, including full conversations, token usage, timing, tool calls and model metadata—useful for performance analysis, agent-behavior research, and inference debugging.

Stera-10M

Open egocentric multimodal dataset for embodied AI and robot learning captured on commodity iPhone Pro: ~200 hours and ~10M RGB frames with LiDAR depth, ARKit 6‑DoF poses, IMU, two‑hand MANO mocap, room meshes, and hierarchical action captions.

Carbon Pretraining Corpus

Provides 173M DNA/RNA sequences (≈1.1 trillion nucleotides) assembled specifically for pretraining genomic foundation models. Includes eukaryote, prokaryote, and mRNA configs plus a 10B‑token eukaryote subset for faster experiments; formatted for streaming and tokenized with Carbon's 6‑mer setup.

OpenCS2 - POV Renders

Provides tick-aligned Counter-Strike 2 player POV video clips with per-tick inputs and world-state sidecars — near-lossless 1280×720@32fps video, per-player stereo audio, and parquet indexes for event/kill/round filtering; suited for RL, video classification and clip mining.

DeepSeek v4 Pro Agent Traces

Contains 4,006 newline-delimited JSONL agent-session traces recording assistant responses and tool calls from deepseek/deepseek-v4-pro — includes a training-ready tools schema snapshot and helpers for conversion to SFT/distillation workflows.

5CD-AI/Viet-Handwriting-OCR-v2

Labeled Vietnamese handwritten line images paired with text transcriptions for training and evaluating OCR/text-recognition models. Stored in Parquet (optimized) with a dataset size in the 10K–100K sample range, suitable for model training and benchmarking.

K12-KGraph: A Curriculum-Aligned Knowledge Graph for Benchmarking and Training Educational LLMs

Provides a curriculum-aligned knowledge graph extracted from Chinese K–12 textbooks and accompanying benchmarks and training data to evaluate and train educational LLMs. Releases a 23,640-question multi-select benchmark and a 7,335-sample graph-guided training corpus with multimodal VQA pairs and the full construction pipeline.

Open-MM-RL

Multimodal STEM problem set for verifiable, answer-supervised training and RL: contains single-image, multi-panel, and multi-image PhD-level questions across physics, math, chemistry and biology. Each example has a deterministic ground-truth answer, enabling reward modeling and automated evaluation.

TiniX Vietnam OCR Annual Financial Statements

OCR-extracted Vietnamese annual financial reports (2015–2025) from 18,231 filings across 1,491 tickers — plain-text OCR outputs for document-QA, information extraction, VLM/RAG development. Contains only TXT OCR files; CC BY-NC 4.0 license.

IndustryBench

Multilingual benchmark for evaluating LLMs' industrial domain knowledge via 2,049 expert-curated QA pairs spanning 10 product verticals and four languages, with each item grounded to industry or national standards and an LLM-as-judge evaluation pipeline.

i1-captions (zlab-princeton)

Provides the full caption corpus used to train and ablate the i1 text-to-image model: 12 curated subsets with multiple caption variants (long/short, VLM-generated, rendered text) to enable reproducible training and captioning experiments.

CUA-Gym

Pairs natural-language instructions with executable setup artifacts and Python reward functions to create verifiable computer-use agent tasks. Provides a Parquet task table for fast filtering plus a compressed archive of runnable task bundles; several web task endpoints are placeholders that require a local CUA-Gym-Hub deployment.