New
Everything new in AI as it appears: launches announced by the labs, brand-new models, datasets, papers and open-source tools. Checked every minute or two. Launches feed · Trending · News
Live now
- Connecting…
Yesterday
- Datasetkuvvi/gt-tool-trace-pilot
GT Tool-Trace Pilot: reference solutions with exact tool calls and tool-output images Summary This repository holds 471 records in 4 visual tasks: rotated text reading, m
- DatasetSOTAagi2030/Gallery-Artwork-Rights-472
Gallery artwork rights register This controlled dataset records dated clearance revisions for artworks in the exhibit queue.
- DatasetSOTAagi2030/Gallery-Exhibit-Queue-472
Gallery exhibit queue This controlled dataset contains dated exhibition queues. Reconcile every artwork occurrence against the companion dated rights register.
- Datasetdererererek/cover_orange_8_varycup_orangeat37_20261021_155750
This dataset was created using LeRobot. Dataset Structure meta/info.json: { "codebase_version": "v3.0", "fps": 30, "features": { "action": { "dtype": "float32", "names":
- Datasetfrachristian/cs224n-nlp-qa
NLP QA Image Audio Data Notes Dataset summary This repository contains a preparation pipeline and a small metadata sample for NLP QA work with Image Audio inputs. It does
- Datasetvineetk13/rowhammer_analysis
RowHammer Analysis Pipeline GPU-accelerated spatial-statistical analysis of DRAM RowHammer bitflip data, producing cached results in a SQLite database for downstream pape
- Datasetdererererek/cover_orange_8_varycup_orangeat59_20261021_155243
This dataset was created using LeRobot. Dataset Structure meta/info.json: { "codebase_version": "v3.0", "fps": 30, "features": { "action": { "dtype": "float32", "names":
- Datasetnakas/gpttrace-text-train
GPTTrace text-detection training data The human and machine text we created or collected to train GPTTrace's AI-text detectors, beyond the public RAID and MAGE benchmarks
- DatasetSOTAagi2030/Estuary-Recording-Grants-462
Estuary recording grant registry This registry is the controlled rights authority for the delivery exercise. Territory tokens use semicolons.
- DatasetSOTAagi2030/Estuary-Delivery-Orders-462
Estuary sound delivery orders Controlled delivery requests for the Estuary acoustic archive. Use SOUND_DELIVERY_ORDERS.tsv together with the separately supplied grant reg
- Dataset
- Datasetdererererek/cover_orange_8_varycup_orangeat159_20261021_154900
This dataset was created using LeRobot. Dataset Structure meta/info.json: { "codebase_version": "v3.0", "fps": 30, "features": { "action": { "dtype": "float32", "names":
- Datasetdererererek/cover_orange_10_varycup_1_20261021_154702
This dataset was created using LeRobot. Dataset Structure meta/info.json: { "codebase_version": "v3.0", "fps": 30, "features": { "action": { "dtype": "float32", "names":
- Dataset
- DatasetArkonyx/assistant-axis-reward-hacking
Assistant Axis reward-hacking observations 824 observations from GPT-OSS-120B layer 17 (hidden size 2880), six cache shards, and the original Assistant Axis tensor. Label
- Datasetanne-voigt/bcc-pdac-kg
BCC PDAC literature knowledge graph — Phase 0 An evidence-bearing graph of the pancreatic ductal adenocarcinoma (PDAC) literature: 62,835 papers published from 2021-06-01
- Dataset
- Datasethausmer/ua-ru-tg-channels-clean
Ukrainian/Russian Telegram channels: news, official, tech, jobs (PII-clean) The non-humour part of the Truha training corpus: Ukrainian and Russian news, regional news, o
- Datasethausmer/truha-eval-packets
Truha blind-test packets Generated samples from the tiny «Truha» LMs (42M/70M from-scratch, Qwen3-1.7B LoRA arms) used in blind A/B and coherence tests, with the unblindi
- Datasethausmer/ru-pd-humor
Russian public-domain humour 1,869 stories/feuilletons, 15,397,680 chars: Аверченко 944, early Chekhov / Антоша Чехонте (535, pre-1888 only), Тэффи 390. Texts from ru.wik
- Datasethausmer/truha-jokes-ru
Russian jokes (anekdoty), cleaned 216,318 Russian jokes (~33M chars) with a visible punchline and without dated Soviet/90s or political material, used as the jokes tag of
- Dataset
- Datasetdivinity-library/03-open-theology-ocr
Document OCR using GLM-OCR This dataset contains OCR results from images in divinity-library/open-theology-pages using GLM-OCR, a compact 0.9B OCR model achieving SOTA pe
- Datasetzhennan1/LongZero
LongZero 🌐 Github | 📖 Paper | 🤗 Data Introduction LongZero lets an LLM improve its own long-context reasoning without any external data. The model first writes the que
- Dataset
- Datasetliveatlasearth/most-active-volcanoes-since-1900
Most Active Volcanoes Since 1900 486 Volcanoes and Their 4,454 Eruptions Since 1900, Ranked, With VEI and Maps Every volcano that has started at least one eruption since
- Dataset
- DatasetSOTAagi2030/Field-Sample-Policies-473
Sample policy history This controlled dataset is the authoritative dated policy history for field samples.
- DatasetSOTAagi2030/Field-Dispatches-473
Field dispatches This controlled dataset contains the dispatch list for the field-sample access register.
- Datasetliveatlasearth/us-tornado-season-centroid
US Tornado Season Centroid by Day, 1950-2025 Daily Center of US Tornado Activity From 73,378 NOAA Tornadoes, 1950-2025. Stack every US tornado from 1950 to 2025 onto a si
- DatasetDeepakSinghworks/video-text-samples
Notes on Social Video Text data Overview Preparation notes and schema examples for Social tasks using Video Text data. Full source material is intentionally not bundled,
- Datasetmazakone/telegram-trading-bots
Crypto Trading Bots & AI Agents — Expert Knowledge Base v5 27 Tools · Telegram Bots · Web Terminals · CEX · Arbitrage · AI Agents · Prediction Markets · 14 Chains The mos
- Datasetrbtrprjkt/philips-screw-fasten_2
This dataset was created using LeRobot. Dataset Structure meta/info.json: { "codebase_version": "v3.0", "fps": 30, "features": { "action": { "dtype": "float32", "names":
- Datasetannaschwarz/textile-samples
Textile Video Text Data Notes Dataset summary Preparation notes and schema examples for Textile tasks using Video Text data. Full source material is intentionally not bun
- Dataset
- Dataset
- DatasetSOTAagi2030/HarborArchive-Accession-Register
Harbor Archive public accession register This release contains 4 digitized accessions. 2 resolved accessions were withheld under the release rule. Source events received
- Datasetmadokalif/calvin-abc-stop0-then-h30
CALVIN ABC 합성 데이터셋 (LeRobot v2.1) data 와 videos 는 bundle.tar.part-* 로 묶여 있다. 파일 하나씩 올리면 HF 의 API 한도 (5 분에 1,000 번) 에 걸려 한 데이터셋에 다섯 시간이 넘게 걸린다. 받은 뒤 푸는 법이다. bash calvin/un
- Datasetanagrammio/anagram-statistics
English first, version française complète plus bas. Kurzfassung auf Deutsch, resumen en español e sintesi in italiano at the end. How Often Words Have Anagrams, Language
- Datasetanagrammio/anagram-benchmark-llm
English first, version française complète plus bas. Kurzfassung auf Deutsch, resumen en español e sintesi in italiano at the end. A Multilingual Anagram Benchmark For Lan
Headlines and links come from the labs' own blogs, Hacker News and the GDELT news index, and from the sources our index already follows. We show a short excerpt and always link to the original. A launch reported by a news site rather than the lab is marked “Reported” until the lab itself announces it.