Kling-Keye (Yuzhe Chen, Muxi Diao, Lele Yang, Wenbin Xu, Xuyang Liu, Qile Su, Changyi Liu, Jiankang Chen) · Agent Benchmark
VibePrevisBench
VibePrevisBench is a benchmark for agentic coding in 4D scene reconstruction, where an agent turns a brief into an executable Blender scene and renders its proxy video. It has 242 tasks across four input kinds (scripts, photos, continuations and reference videos), scored by a rule-based check against hidden ground truth in nine dimensions and a VLM judge of the rendered proxy video. The authors evaluated 11 frontier coding agents and report that multimodal understanding of space and camera motion separates them, and that agents inspect their renders but fail to correct errors.
Open the source ↗Similar Research & Safety releases
- Nemotron RL Litmus-Bench V2.0
NVIDIA · Science Post-Training Dataset · Week 48, 2026
- NovaAtom-Lite
Geodesic Intelligence · All-atom structure prediction · Week 47, 2026
- XYBench
Ai2 (allenai), with researchers at Carnegie Mellon, NVIDIA and Johns Hopkins · XY problem benchmark · Week 47, 2026
- One Table Benchmark
Yuanrui Wang, Kaifang Li, Xuanyue Li, Ziyun Li, Yan Lu, Xingxuan Zhang, Peng Cui (Stable AI) · Tabular Benchmark · Week 47, 2026
- Embodied First Exam
MLL Lab (Kangrui Wang, Qineng Wang, Changhe Chen et al.) · Robot Control Benchmark · Week 47, 2026
- MMDM: Measuring Massive Multitask Decision Making
Junhao Shen, Qing Zong, Yue Guo, Zhanqiu Zhang, Yiwen Guo, Hong Cheng (The Chinese University of Hong Kong, Database Group) · Decision model benchmark · Week 47, 2026
Published by Quanta AI Labs, an AI agency in India. Browse every week in AI News Weekly.