AuroraASR-1M
950K-param Conformer-CTC English speech-to-text demo
None defined yet.
950K-param Conformer-CTC English speech-to-text demo
Multi-view 3D spatial VQA with dense reward
Calibrated typed decisions from a case + question schema
8-step DMD text-to-image distillation of MOSS-Image-SFT
8-step DMD text-to-image distillation of MOSS-Image-Base
Turkish speech in one designed voice, 6 emotions + laughter
Structured-prompt scaling for Qwen-Image
Spatial reasoning over video using 3D geometry as position
Camera views + instruction → robot action chunks
VLA model predicting robot action chunks from images
Fast-VLA robot policy - think once, act fast
Typed decisions with calibrated probabilities, no decoding
Zero-shot typed extraction with a 0.6B encoder
Typed decisions, with a thought when it is unsure
Swap one character in a clip for a reference character
3 km climate downscaling with ACE2S + HiROv2
Spatial VLM that decides in one forward pass
Time-series Q&A with a dual-view Qwen3-VL-8B
Agentic streaming video QA with memory and replay
Consolidate a whole video into LoRA memory weights, then ask
Complete dense 3D mesh motion from a few vertices
Arabic speech recognition for 19 dialects + code-switching
Zero-shot probabilistic time series forecasting
Zero-shot probabilistic time series forecasting