SAVRN Model Hub
AI Training Datasets
Each dataset with its card, its structure (every split, row count and column type), its files, its license, and the models that disclose training on it.
11,122Models
3,164Datasets
285Papers
6,792Publishers
11,852Sourced relationships
Updated 2026-10-18 · How the library is built
3,164 datasets, sorted by most downloaded.
The preprocessed training data of the DMAD 4-step students of MiniMax-H3 (text-to-audio-video), from the paper DMAD: Distribution Matching as Adversarial Distillation for Fast Visual Generation. The training code is in train/h3 of the DMAD repository; the released students are at ZhengmingYu/DMAD. Everything here is what train/h3/scripts/preparedata.sh produces from UltraVideo (short.csv and the clipsshort1920 clips) and the MiniMax-H3 base model, so training can start without the ~1 GPU-week of preprocessing. No video or audio is stored: the clips are kept only as MiniMax-H3 VAE latents, the captions as text-encoder outputs (the captions themselves are included as text). 303 GB, 59,818…
DVBench is a benchmark for evaluating multimodal large language models on data videos, a storytelling medium that combines dynamic charts with structured narratives. - 300 data videos - 1,000 human-verified question-answer pairs - questionid: unique question identifier - questiontype: EM, MCQsingle, MCQmultiple, or Openended - video: video identifier matching videos/.mp4 - dimension: evaluation dimension - question: question or subtitle-cloze context - answer: reference answer - distractor1, distractor2, distractor3: distractor options for single-choice and multiple-choice questions - charttype: chart category for visual questions in Animation, Chart Perception, and Chart Reasoning…
This is a mirror of the original DrivingStereo download files, preserving the original ZIP archives and camera calibration text files without recompression. The original dataset was introduced at CVPR 2019 by Guorun Yang, Xiao Song, Chaoqing Huang, Zhidong Deng, Jianping Shi, and Bolei Zhou. The inventory contains 300 source files totaling 202,346,258,264 bytes (202.346 GB). The early demonstration package is excluded; the released training, testing, weather, and calibration directories are mirrored in full. Transfer in progress. As of 2026-10-17T03:37:54+00:00, 0/300 files (0.000/202.346 GB) are verified on the Hub. Status is refreshed after each batch. transferstatus.json contains the…
This is a filtered copy of the full ImageNet dataset consisting of the top 11821 (of 21841) classes by number of samples. It has been used to pretrain a number of in12k models in timm. The code and metadata for building this dataset from the original full ImageNet can be found at https://github.com/rwightman/imagenet-12k NOTE: This subset was filtered from the original fall11 ImageNet release which has been replaced by the winter21 release which removes close to 3000 synsets containing people, a number of these are of an offensive or sensitive nature. There is work in progress to filter a similar dataset from winter21, and there is already ImageNet-21k-P but with different thresholds &…
ShareGPT-format SFT datasets built locally for the DuckBot personal-agent fine-tunes (Unsloth, Mac Studio), published 2026-10-15. These are the exact training files used in the most recent runs, with personal information removed (see "Privacy checks" below). Originals remain private. Row counts include the creator watermark rows described below. Format: one JSON object per line, {"conversations": [{"from": "human"|"gpt", "value": "..."}]}. Curated merges of public datasets (OpenHermes-2.5, glaive-function-calling-v2, SlimOrca, WebSight, s1K, Fable 5.1, Pasta009 instruction-fusion, OpenThoughts, Precinct6, medical and science sets) plus locally generated UI/UX and distillation examples.…
Dataset · Tabular classification
EMNLP_Cost-Aware-Protocol-Routing
The short version. We ran the same 6,803 reasoning problems through four different LLM collaboration setups — from a single direct answer up to a four-agent deliberation — and recorded, for every problem, which ones got it right. Then we asked whether a model can look at a problem beforehand and predict which setup is worth paying for. It can predict whether it will fail. It cannot predict which collaboration protocol will fix the failure. That gap is what this dataset is for. This is the data release for the EMNLP 2026 paper LLMs Can Predict Failure Risk, But Struggle to Predict Which Collaboration Protocol Pays Off. Every released problem was run under all four, with the same solver…
EMO-X Core Task Families Public task-family manifests for EMO-X (Adaptive Agent Evaluation): 76 families across 13 suites — code, agent-loop, recovery, robustness, calibration, long-horizon, gauntlet, vision, realworld, issues, security, computer-use (pilot), dynamic-code. Files tasks.jsonl — one row per family: suite, familyid, category, difficulty, capabilities, oracle type, pilot flag, plus the full frozen Task DSL manifestjson inline (generator config
EchoAtlas Call Annotations This release contains structured call descriptions generated for the EchoAtlas wetland audio collection. Annotation provenance The immediate annotation model was Mistral-7B-Instruct-v0.3. Its stated license is Apache License 2.0. Apply that immediate model's license to this release. Annotation fields Each entry includes a clip key, candidate species, acoustic pattern summary, confidence band, and reviewer status.
This dataset contains machine-proposed acoustic-event labels aligned to the EchoAtlas field clips. Each record supplies a clip key, event family, start offset, end offset, and confidence band. Human spot checks are quality-control records rather than label sources. Lower restriction indexes indicate broader reuse within the eligible set.
This card documents hydrophone clips collected for the EchoAtlas coastal acoustics release. Recordings span dawn and evening sessions at three near-shore monitoring stations. Timestamps are stored in UTC and clips are distributed without observer notes. Lower restriction indexes indicate broader reuse within the eligible set.
EchoAtlas Wetland Clips This release contains standardized short audio clips gathered for wetland species monitoring. Recording provenance The clips combine material from two direct archives: Marshlight Community Recordings, distributed under CC BY-NC 4.0. North Fen Open Soundbank, distributed under CC BY 4.0. For the public release, choose the direct-source license that permits commercial reuse as well as adaptation and redistribution. Clip fields
EgoHandBench Source collects the source-format data used by EgoHandBench. The seven dataset directories retain their existing media, annotations, calibration files, filenames and directory structures. Annotation field names, units, joint orders and MANO conventions are dataset-specific. This is the source edition before EgoHandBench frame-level quality filtering. It retains the available source timelines for the selected cameras, including recordings later excluded by quality filtering. Stereo collections use the selected right camera; HOT3D uses Aria RGB. Shared annotation or calibration files may also describe cameras whose RGB is not included. Native image sequences, videos and source…
A sealed Korean benchmark of 300 enterprise policy decisions whose gold labels were computed, not written. No human and no language model ever supplied an answer. A hidden executable policy maps a set of latent factors to exactly one correct action and emits a proof; the natural-language rendering is produced separately and never sees the label. Each item asks the same question an enterprise agent has to answer before it says anything: given this written policy, these callable tools, what the system already knows, and this user turn — which one of seven actions is correct? the usual routes first and they failed, which is why this one exists. So the boundary is no longer written into the…
102,881 conversations · 1.27 GiB · $164 API generation cost+ $80 (codex) is a high-quality conversational training dataset focused on reasoning, mathematics, STEM, Python programming, software security, and tool use. It was independently generated by the dataset creator through API-based generation at a total cost of $164. Every example follows the Hugging Face conversational messages format and includes a short topicsummary for filtering, analysis, and curriculum construction. The data includes detailed solutions, multi-turn dialogue, system prompts, tool calls, tool responses, and structured tool definitions. The dataset is designed as quality-first training data with complete answers…
Model Collections
Hand-picked starting points, each with the reason it exists.
Collection · 4 entries
Embedding models for retrieval
Sentence and document embedding models used to build retrieval systems. Dimension and sequence length matter more than size here, and both come from the publisher.
Collection · 4 entries
Models that fit on one accelerator
Models whose publisher-reported parameter count puts them within reach of a single accelerator at common precisions. Memory needed depends on precision and serving configuration, so treat the parameter count as the starting point, not the answer.
Collection · 6 entries
Open-weight text models worth knowing
Widely used open-weight language models, chosen because each one is a distinct family rather than a variant of the one above it. Selection, not a ranking.
Collection · 3 entries
Speech and audio models
Recognition and synthesis models, grouped so the two directions are easy to compare.
Related SAVRN Research
The hub sits beside SAVRN's market data and infrastructure research: what models cost to run, and what it takes to run them.
SAVRN Index
What open models cost to run
The same open-weight model priced by every host that serves it, per million tokens.
Research Hub
Data center trackers and maps
Moratoriums, permits, power, water and capital behind the facilities that run these models.
Method
How the Model Hub is built
Sources, evidence labels, refresh behaviour, and the limits of every comparison here.