WBench
PulseAugur coverage of WBench — every cluster mentioning WBench across labs, papers, and developer communities, ranked by signal.
2 day(s) with sentiment data
-
Mix Forcing Pre-Training Enhances Interactive Video World Models
Researchers have introduced Mix Forcing, a novel causal pre-training method designed to improve interactive video world models. This technique, implemented with Mixture-of-Bidirectional-and-Autoregressive (MoBA) attenti…
-
World-in-Context framework unifies video controls for interactive modeling
Researchers have introduced World-in-Context (WiC), a novel framework for interactive video world modeling that unifies view, action, and camera controls. Unlike previous methods that handle these controls implicitly or…
-
AI research tackles long-video memory with novel weight-internalization techniques · 7 sources tracked
Researchers are developing novel methods to enhance the memory capabilities of AI models, particularly for processing long video streams. Several papers introduce techniques that internalize historical data into model w…
-
HiDream.ai launches embodied world model, tops RoboColiseum benchmark
HiDream.ai has launched its embodied world model, HiDream-O1-Embodied, which focuses on enhancing robots' physical perception and dynamic prediction capabilities for advanced physical interaction. The model achieved the…
-
HiDream.ai unveils native fully multimodal world model HiDream-O1-World
At the World Robot Conference 2026, HiDream.ai founder and CEO Mei Tao unveiled the HiDream-O1-World, a native, fully multimodal world model. This model integrates text, image, and interactive control, aiming to bridge …
-
JoyAI-Echo-1.5 system unifies long-form video and interactive world generation
Researchers have introduced JoyAI-Echo-1.5, a unified system for generating long-form audio-visual content, including persistent stories and interactive worlds. The system utilizes cross-shot memory to maintain characte…
-
ZhiXiang Future launches interactive world model HiDream-O1-World
ZhiXiang Future has released HiDream-O1-World, an interactive world model that supports multimodal inputs like text and images. The model excels in generating temporally and physically consistent worlds, a significant a…
-
Alaya-EVOKE introduces interactive world model for open-ended video generation
Researchers have introduced Alaya-EVOKE, an interactive world model designed for open-ended video generation. The model utilizes external persistent memory and a novel long-horizon teacher to achieve responsive generati…
-
New frameworks and benchmarks accelerate interactive video world models
Researchers have developed new methods to accelerate interactive video world models, which generate video content based on user camera movements. "Light Interaction" offers a training-free approach by adaptively managin…