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
- Datasetjmatarese/atp-data
ATP data for the tennis-hof project The pinned data source for jmatarese/tennis-hof, an objective ranking of ATP players for the International Tennis Hall of Fame. All da
- Datasetphanerozoic/orbitlight
orbitlight Planets, their atmospheres and the companions of their stars, read out of archived high-resolution spectra of the host stars. A transiting planet's atmosphere,
- DatasetJonesLin/vla-arena-offline-cot-10k
VLA-Arena Offline CoT 10k — password-encrypted tar 10,000 verified simulated trajectories with CoT, L0/L1/L2, 2,428,135 frames. Only AES-256 encrypted data is published.
- Dataset
- Datasetabo003516/flutter-agentic-dataset
🚀 Flutter AI Agentic Trajectories & Apps Dataset Generated on AMD Instinct™ MI300X (192GB VRAM) using Qwen/Qwen2.5-Coder-32B-Instruct via vLLM with compiler-verified sel
- Datasetraiansar/universal-remote-code-list-versions
Universal remote code list versions Per-brand code counts for the code lists behind GE, UltraPro and Philips universal remotes (Jasco CL3, CL4, CL5, CL6) and 35 RCA remot
- Datasetraiansar/bra-size-conversion-chart
Bra size conversion (UK, US, EU, FR, AU/NZ, JP) Lookup tables for converting a bra size label between six systems, built for size normalisation in product catalogues and
- Datasetraiansar/bed-mattress-sizes
Bed and mattress sizes (US, UK, France) 22 mattress sizes across three markets, each with width and length in both inches and centimetres and the published source it was
- Datasetstoragestuffnobody/stories-embedding
stories-embedding Public full-text embeddings of storagestuffnobody/stories. Status: queued. Full-text source rows committed: 0. Chunk vectors saved: 0; previously comput
- Datasetstoragestuffnobody/stack-v3.1-subsampled-embeddings
stack-v3.1-subsampled-embeddings Public full-text embeddings of storagestuffnobody/stack-v3.1-subsampled. Status: running. Full-text source rows committed: 80,000. Chunk
- Datasetstoragestuffnobody/finepdf-subsampled-embeddings
finepdf-subsampled-embeddings Public full-text embeddings of storagestuffnobody/finepdf-subsampled. Status: running. Full-text source rows committed: 130,000. Chunk vecto
- DatasetTommysha/octopus-train
octopus-train 一个数据集，两条分支。gpu 上如果 huggingface.co 被重置，先设镜像再下载。 export HF_ENDPOINT=https://hf-mirror.com huggingface-cli download Tommysha/octopus-train --repo-type dataset
- Dataset
- Dataset
- Dataset
- DatasetPMurod/uzbek-SAP-llm
uzbek-SAP-llm SAP dasturi bo‘yicha o‘zbek tilidagi savol-javob (instruct) ma’lumotlar to‘plami. Til modellarini SAP mavzusida fine-tuning qilish uchun mo‘ljallangan. Tark
- Datasetcliecy/DeepSeek-V4.1-Flash-1017-pruning-eval
DeepSeek-V4.1-Flash-1017 专家剪枝 / 极限量化测试 目标：让 DeepSeek-V4.1-Flash-1017（748.5B 参数，40 层 × 384 路由专家，每 token 激活 6 个）跑在 2×RTX 4090 48G + 61G 内存 + NVMe 上。办法是剪掉一部分路由专家，再把剩下的专家压到 I
- DatasetSandeep026/ci-repair-bench
CI-REPAIR-BENCH Overview CI-REPAIR-BENCH is a benchmark dataset for research on Continuous Integration (CI) failures and automated repair in Python repositories. The data
- DatasetSOTAagi2030/Northlight-Rights-Windows-458
Rights-window register Controlled territorial and date-window terms for the digitization release queue.
- DatasetSOTAagi2030/Northlight-Digitization-Queue-458
Digitization release queue Controlled fixture for scheduling public-domain scan releases against recorded rights windows.
- Datasetattacker243/cellxss1021
cell xss probe
- DatasetThuraAung1601/disco-dafny-corpus-hint-gen
Disco Dafny corpus: hint infill (DafnyBench-style), deduplicated RLVR tasks built from both Formal Disco releases of LLM-generated, verified Dafny programs, merged and de
- Datasetai4ce/wanderland-preview
Wanderland Preview Compressed, browser-ready previews of the 42 Wanderland V2 showcase scenes, used by the interactive explorer. These previews are not for research use.
- Dataset
- Dataset
- Datasetxyliu6/koniq10k-1024x768
KonIQ-10k 1024x768 — unmodified mirror This repository mirrors koniq10k_1024x768.zip from the KonIQ-10k IQA Database exactly as published on DaRUS, for reproducible build
- Datasetjizhuochen/record3d-mapping-jump-20261021
Record3D mapping sequence with a tracked-pose jump Original indoor landscape RGB-D recording 2026-10-21--16-22-34.r3d, captured on 21 October 2026. This sequence is share
- DatasetPureOne/SYRAETH-II-Sharp-Field-Gaps
SYRÆTH II Sharp Field Gaps in Normalized Spectral Completion Research manuscript · Version 2.0 · 21 October 2026Eve · Mathematical research prepared for Maciej Nowicki Re
- DatasetJasmina123/llm-blind-spot-demand-forecasting
Limited-History Sales Forecasting: A Blind-Spot Stress Test for Qwen3-4B 1. Motivation and the blind spot My background is in computer science engineering, AI, and data s
- Dataset
- Datasetjackyk02/embed-nemotron-qwen3.6-35b-a3b-fp8-embeddings-256k
embed-nemotron retrieval pairs, state/action embeddings with mined hard negatives: Qwen3.6-35B-A3B-FP8, 256k context Every distinct query of the 14 subsets of nvidia/embe
- Datasetfood-ai-nexus/wheat-deoxynivalenol-nir-austria-france
Deoxynivalenol in Wheat by NIR Spectroscopy, Austria and France Original data: Nadeak, Anlanger, Rennhofer, Sulyok and Freitag (2026), Zenodo, doi:10.5281/zenodo.20931697
- Datasetfood-ai-nexus/listeria-qac-tolerance-wgs
Listeria monocytogenes QAC Disinfectant Tolerance from WGS Original data: Gmeiner, Ivanova, Njage, Hansen, Chindelevitch and Leekitcharoenphon (2025), supplementary infor
- Datasetjackyk02/nemotron-sft-agentic-v2-qwen3.6-35b-a3b-fp8-embeddings-256k
Nemotron-SFT-Agentic-v2 interactive_agent state/action embeddings: Qwen3.6-35B-A3B-FP8, 256k context Every assistant turn of the interactive_agent conversations of nvidia
- Datasetjfrick/python-math-500m-sweep-stopped-20261021
Stopped Python/math distillation sweep Private archive of Qwen3-8B generated problems, code and explanations. Producer tar files preserve all accepted chunk JSONL.gz file
- Datasetbiunlp/beir-he-hotpotqa
BeIR-HE: HotpotQA (Hebrew) Hebrew translation of the HotpotQA BeIR benchmark dataset. Translated from English to Hebrew using gemini-3.1-flash-lite (Vertex AI batch) with
- Datasetjackyk02/nemotron-agentic-v1-qwen3.6-35b-a3b-fp8-embeddings-256k
Nemotron-Agentic-v1 state/action embeddings: Qwen3.6-35B-A3B-FP8, 256k context Every assistant turn of every conversation in nvidia/Nemotron-Agentic-v1 (revision 650d590:
- Datasetjackyk02/acemath-rm-qwen3.6-35b-a3b-fp8-embeddings-256k
AceMath-RM state/action embeddings with hard negatives: Qwen3.6-35B-A3B-FP8, 256k context Research use only (CC BY-NC 4.0, non-commercial; must not feed a commercial mode
- Dataset
- Datasetjackyk02/nemotron-abstention-safety-qwen3.6-35b-a3b-fp8-embeddings-256k
Abstention and safety decisions (QA-Abstention, Safety-v1, Aegis 2.0), state/action embeddings with hard negatives: Qwen3.6-35B-A3B-FP8, 256k context Three NVIDIA release
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.