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.
This dataset was created using LeRobot.
Every forecast the OpenThomas weather desk read or produced, as-of. Pushed daily from the trading box, so each commit is a timestamped record of what was known when — the leak-free substrate behind the paper's replay experiments (docs/EXPERIMENTS.md in the replay/ holds the frozen decision-time rows (one JSONL per window, digest in the filename) and the experiment results scored on them. A row carries the settlement station and day, the baseline probability at the snapshot, the Kalshi bid/ask at that snapshot, and the outcome — never anything the agent could not have known at decision time. - replay/e3-high9.txt - replay/e3-low1.txt - replay/e3-low2.txt - replay/e3e4.log - replay/e3e4.sh…
Part of GloSL, a set of continuous sign language translation corpora in one format. annotations.csv, card.json, poses/ /.pose. card.json holds the full provenance. annotations.csv columns (same in every GloSL dataset; empty when not applicable): id, split, text, gloss, glosstype, signer, start, end, fps, nframes, videoref, posefile, videofile. rtmpose-wholebody (Uni-Sign), 133 points (BODY133 23, FACE133 68, LEFTHAND133 21, RIGHTHAND133 21), 2D. Coordinates: pixels of the frame the pose refers to (header width x height, depth 0); z raw. Confidence: RTMPose score; missing point = (0,0) with confidence 0. Frame rate: native, per file (also in annotations.csv). Normalization (glosl-norm-v1) is…
Data for World Models with Configurable Perception, covering Push-T, Two-Room, and OGBench-Cube with distraction variants d0 through d8. The dataset supports query-conditioned visual representation learning, action-conditioned world modeling, and online control evaluation. d0 has no distractors; d1 through d8 add one through eight. Click the figure to view the original image. Each environment uses the same three collections: The full dataset is approximately 1.64 TB. To download one environment and variant, run this from the ConfigWM code repository: - --env: pusht, tworoom, or ogbcube. - --task stageb --variant d3: data for training ConfigWM on one variant. - --task stagea: data for…
Public research release of 102 Singapore legal research questions, model responses from 6 systems, and overlapping grades on five dimensions. Headline metrics are overlapping binary flags, not a ranking and not a partition of 100%. — comparison table, category heatmap, per-question comparison, and every answer with its sources and grades. Hallucination is Incorrectness OR Misgroundedness. Incompleteness and Substantial Correctness are independent of Hallucination. A response may be both Substantially Correct and Hallucinated. Hallucination × Substantial Correctness: H yes / SC yes = 269; H yes / SC no = 178; H no / SC yes = 130; H no / SC no = 35. Each cell is count (percentage; Wilson…
Physics videos of everyday objects breaking and being rebuilt, for STRIVE. Objects from the Breaking Bad everyday subset are placed in texverse-cleverer-strive's emergent scenes, simulated in MuJoCo and rendered in Blender. Some break on impact; others are rebuilt piece by piece by a visible robot tool. Every clip has exact per-instance masks and interaction labels: collisions, contacts, fractures, attaches and holds. This repository also holds what the scenes are made from: the collision proxies and the attach plans. Download a folder as its ZIP. The Hub rate-limits requests, and a folder of thousands of files needs one request per file. Each ZIP holds its folder's files under their…
v3 is the only release available on the current main branch. The default configuration is choicev3. The 2026-10-14 v3 data repairs previously documented supervision and evaluation defects. It regenerates semantic candidate pairs, restores full Hermes requests and native call sets, recovers Nemotron schemas, restores legacy validation, and disables unsupported NLU++ Boolean labels. Model training and capability-retention experiments have not been run for this release. legacyreplayv3, semanticv3, and toolcontextv3 are exact subsets of choicev3. Do not concatenate these subsets with the integrated config. toolenvironmentholdoutv3 overlaps the tool validation/test splits. All three native…
O split test é o subconjunto balanceado de ~10 h; train é o restante. Não há segmentos em comum entre os dois. Nos splits train e test, a coluna humanannotedsentence corresponde à coluna sentence anterior. No split test, ela contém a transcrição humana quando disponível: são 1.612 segmentos anotados manualmente; os demais permanecem nulos. As saídas automáticas originais são mantidas separadamente em sttwhisper, sttwav2vec e sttparakeet.
Internal documentation Analysis and results separates evaluation tasks and exact group memberships. Each plot appears above its data table. Current trained MV results cover 12 settings with 99 groups each; published baselines use a separate 99-group cohort. A matched MV vs. fixed-baseline comparison uses the same 65 DeepSWE groups. The full 98-group intersection is documented; baseline episode records are needed to
Kinex Benchmark (Native Control Budget) Current defaults for new evaluations Environment Public action-step budget Control time LIBERO (all four suites) RoboTwin Official per-task 400–1,700 RoboDojo SIMPLE RoboCasa LIBERO action-step usage and limits on the site and in CSV exports exclude five environment initialization steps. Its runtime limit is 605
Research snapshot of official national legislation from Al-Meezan (Qatari Legal Portal / Ministry of Justice). Not legal advice. The official gazette / authentic source prevails over this corpus. - data/laws.parquet — one row per instrument (LAWFLAT, zstd). - data/articles.parquet — article/section rows for the same snapshot. - scrapers/collectqa.py — research collector used to build this snapshot. Official Al-Meezan texts; authentic Official Gazette prevails; not legal advice. This dataset is not legal advice. The official source prevails. - https://www.almeezan.qa/ National laws in Arabic from almeezan.qa. Constitution included. In-force national laws from official Al-Meezan…
Working storage for a from-scratch English language-model pretraining corpus. Each region is one contiguous slice of the source work plan, tokenized and filtered, produced by one parallel worker. This repository is the producer output, not the training set: the regions still have to be merged in work-plan order (global exact-duplicate removal and the token budget are applied there). Derived from nvidia/Nemotron-ClimbMix, revision 5eaa64b9c0c85b7f56af01d7dffdb0795816b12b, which is licensed CC-BY-NC-4.0. This derived dataset inherits that licence: non-commercial use only, attribution to NVIDIA required. No text was rewritten or generated; documents are the source documents, filtered and…
Newly acquired English educational broadcast archives, with original publisher MP3 bytes, derived 16 kHz mono PCM WAV clips, automatic transcripts, word timestamps and per-episode speaker IDs. Collection was initialized on 2026-10-02; source publication dates are preserved separately. The initial pilot recordings were published in 2025; the historical expansion includes older source publication dates. This is an independent corpus made using the processing approach in mit-ccc/audio-scraper, from the group behind RadioTalk. It does not contain recovered original 2018–2019 RadioTalk recordings, and is not an exact replication of its Kaldi-era system, station sampling or talk-radio…
Evaluation sets for the released NCM checkpoints.
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.