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…
Friday 23 October
- DatasetWoojongKim/eval_smolvla_policy_omy_env1_task1_test_20261023_125607
This dataset was created using LeRobot. Dataset Structure meta/info.json: { "codebase_version": "v3.0", "robot_type": "omy_follower", "total_episodes": 2, "total_frames":
- DatasetMarshootLab/beam-verified-reasoning-traces
Beam Verified Reasoning Traces by Marshoot Lab 7,632 reasoning traces from Reflection's Beam-1.0-501B-A23B on 3,595 procedurally generated problems across 13 task familie
- Dataset
- Dataset
- Dataset
- Datasetjuiceb0xc0de/qwen2.5-0.5b-cgs-atlas
juiceb0xc0de/qwen2.5-0.5b-cgs-atlas A brain atlas for Qwen/Qwen2.5-0.5B, a 494M-parameter Qwen2ForCausalLM with 24 layers. This is not a chat dataset or a benchmark. It i
- Dataset
- Datasetsaurabhb2412/sdbench-mai-dxo
MAI-DxO diagnostic agent for SDBench A 5-persona "Chain of Debate" diagnostic agent (Dr. Hypothesis, Dr. Test-Chooser, Dr. Challenger, Dr. Stewardship, and Dr. Checklist
- Datasetbx6d/r2s2r-blue-cube-to-bowl-stage1
xArm6 blue cube to bowl: stage 1 Frozen stage 1 dataset: 400 training + 40 validation episodes, no test split. This is the original retained visual setup, not the separat
- Datasetyuiseki/ksj-p11-bus-stop-jp
ksj-p11-bus-stop-jp The bus stops of Japan as the Ministry of Land, Infrastructure, Transport and Tourism's 国土数値情報 バス停留所データ (P11) gives them in its 2022 edition (第3.0版, 令
- DatasetKelvin000010191/Aggresive-Unfiltered-DPO
What's inside dataset.json Each entry is exactly: { "prompt": "...", "chosen": "...", "rejected": "..." } Design (for maximum refusal-rate impact): 25 categories × 40 ent
- Datasetlaiyingxin/TouchReason
TouchReason Physical evidence for tactile-grounded reasoning Paper · Project Page · GitHub · Touch-R1 Model TouchReason is the dataset release hub for Touch-R1: Reinforci
- Datasetattacker43543/dspt5
dspt5 archive extraction (zip-slip) probe
- Dataset
- Datasetbx6d/n1m19-test1018-memory-tool-sweep
N1M19 (test_1018) — memory × tool sweep over 6 vision-language models Full trajectories of vision-language-model agents on N1M19, an embodied search task in AI2-THOR. Eac
- Dataset
- Dataset
- DatasetVaidhyaMegha/dbms-research-kg
DBMS Research Open-Problems Knowledge Graph 20,211 nodes. 53,721 edges. The 1,053 open problems of database systems research as a graph: for each problem, its topic and s
- Datasethotocooo/lfm25-gsq-replication-diagnosis
Low-bit weight quantization of LFM2.5-2.6B (via DavidAU's Turbo-Brilliance BF16 GGUF): replicating GSQ and a trellis-coded alternative Research report, interim. Last upda
- Datasetwayrise/RoboCOIN
RoboCOIN Robot pretraining data prepared for the OpenWAM project.
- DatasetOpenWAM/Droid
Droid Robot pretraining data prepared for the OpenWAM project.
- DatasetOpenWAM/InternData-A1
InternData-A1 Robot pretraining data prepared for the OpenWAM project.
- DatasetOpenWAM/AgiBotWorld-Beta
AgiBotWorld-Beta Robot pretraining data prepared for the OpenWAM project.
- DatasetChloeing/STOU
STOU STOU is a benchmark for evaluating the safety of multimodal assistants on egocentric (first-person) video. Each instance pairs a short egocentric video clip with a h
- DatasetBeifengdo/Haifeng-Club-Training
version https://git-lfs.github.com/spec/v1 oid sha256:557ab88fd6fa67d7c50e9601947e80cfad18c2e384aff750a095ea2cf7ce83e4 size 442
- DatasetBeifengdo/Dianhang-Club-Training
version https://git-lfs.github.com/spec/v1 oid sha256:8daaa113e543586a5a3d1b51dd49dbd02b54932917e662e1cc8f114c44437809 size 442
- DatasetAlayaLab/MeshAxiom-Assets
MeshAxiom Project Assets Media and downloadable meshes used by the official MeshAxiom homepage and asset library. Inference code and documentation MeshAxiom-Base model we
- Dataset
- DatasetVaidhyaMegha/maths-research-kg
Maths Research Open-Problems Knowledge Graph 34,073 nodes. 78,162 edges. 1,161 open and hard problems in mathematics, in 10 fields, as a graph: what is proved toward each
- Datasetqinghangliu/PAL_realworld
PAL Real-World Evaluation Videos This dataset publishes real-robot experiment videos for Ours, FrameSamp-Modul, and MemER on PatternLock, PickXtimes, VideoPlaceOrder, and
- Datasetatmem/atmem-2.3.10b1-dolphinbench
AtMem 2.3.10b1 — DolphinBench, 600 tasks AtMem with Hermes and GPT-5.6 Luna passed 390 of 600 tasks (65.0%) on the public DolphinBench release. This dataset provides all
- DatasetSeNa1Naka2/so101-pick-place-ver1_20261023_122831
This dataset was created using LeRobot. Dataset Structure meta/info.json: { "codebase_version": "v3.0", "fps": 30, "features": { "action": { "dtype": "float32", "names":
- Dataset
- Datasetcodemaivanngu/nemo-rl-v070-b200-runtime
NeMo RL v0.7.0 runtime for B200 Runtime for the Llama-3.1-8B-Instruct student, with a separate Qwen3-235B-A22B teacher node. Each project node has 8 B200 GPUs. This packa
- DatasetSOTAagi2030/Public-Mural-Conservation-Sequence
Public Mural Conservation Sequence This release turns approved conservation actions into a dependency-safe field sequence for spring 2027. Audit totals Physical revision
- DatasetDSACZX1231EDAS/Annoy-PyEdu-Rs-Raw
Annoy: This should be a paper Title 📑 Paper | 🌐 Project Page | 💾 Released Resources | 📦
- DatasetDSACZX1231EDAS/Annoy-PyEdu-Rs
Annoy: This should be a paper Title 📑 Paper | 🌐 Project Page | 💾 Released Resources | 📦
- DatasetOliverLee520/WeatherForecast
WeatherForecast English · 简体中文 Hourly ERA5-derived weather fields for lower-cost forecasting experiments. Both subsets store unnormalized float32 values and continuous se
- DatasetJamie-nus/CROSS-benchmark-runs
CROSS benchmark: per-frame data Per-frame data of the CROSS benchmark results, plotted by the results page: for every result cell the top-down trajectory (traj_gt, traj_e
- Dataset
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.