CogVideoX
PulseAugur coverage of CogVideoX — every cluster mentioning CogVideoX across labs, papers, and developer communities, ranked by signal.
- 2026-07-19 research_milestone A physics-consistency checker identified a failure mode in the CogVideoX-5b-I2V model where generated objects failed to fall as expected. source
3 day(s) with sentiment data
-
New LFV method enables faster spectral editing in video-VAE latents
Researchers have developed a new method called Latent-Frequency Validity (LFV) to enable faster and more precise spectral editing within video-VAE latents. This technique allows for control over noise, flicker, and smoo…
-
New T2VAttack method reveals vulnerabilities in text-to-video diffusion models
Researchers have developed T2VAttack, a new method to probe the vulnerabilities of text-to-video diffusion models. The attack focuses on both semantic and temporal aspects of video generation, aiming to degrade the alig…
-
CachedSearch accelerates video diffusion model search with novel caching
Researchers have developed CachedSearch, a novel training-free method to accelerate test-time search for video diffusion models. This technique significantly reduces the computational cost of generating high-quality vid…
-
AI video model CogVideoX-5b-I2V fails physics consistency test
An independent physics-consistency checker, Haga, has identified a failure mode in the CogVideoX-5b-I2V model. When prompted with a "ball and block fall" scenario, the model generated videos where the object remained ai…
-
New SIRUS framework enables concept removal in text-to-video models
Researchers have developed SIRUS, a novel framework designed to remove specific concepts from text-to-video (T2V) models at inference time without requiring model retraining. This method localizes and suppresses target …
-
Stable Diffusion user seeks model recommendations for new 16GB GPU
A user on Reddit is seeking recommendations for Stable Diffusion models to utilize with a new graphics card, an RX 6800 XT with 16GB of VRAM. They are upgrading from an RX 6600 and are interested in exploring various im…
-
ACID method accelerates video generation with dynamic caching
Researchers have developed ACID, a novel adaptive caching method designed to accelerate video generation from diffusion models. Unlike existing methods that use a fixed threshold for caching, ACID dynamically adjusts th…
-
OrbitQuant enables data-agnostic quantization for diffusion transformers
Researchers have developed OrbitQuant, a novel method for post-training quantization of diffusion transformers (DiTs). This technique allows for efficient inference by quantizing in a normalized, rotated basis, eliminat…
-
New frameworks enhance physical realism in AI video generation
Researchers have developed two new frameworks, Proprio and LaMo, aimed at improving the physical realism of AI-generated videos. Proprio, a training-free method, enables existing video generators to self-assess and refi…
-
New methods boost video diffusion model efficiency and quality
Researchers are developing new methods to improve the efficiency and quality of video diffusion models. Several papers introduce techniques to optimize attention mechanisms, such as sparse attention (LVSA, Veda) and lin…
-
AI researchers develop latent space probing for adult content detection in video models
Researchers have developed a new method for detecting adult content in videos generated by AI models by analyzing the model's internal latent space representations. This approach, applied to the CogVideoX diffusion mode…