Lewman
PulseAugur coverage of Lewman — every cluster mentioning Lewman across labs, papers, and developer communities, ranked by signal.
5 day(s) with sentiment data
-
New LpWM model uses sparse representations for improved AI planning
Researchers have introduced LpWorldModel (LpWM), a new JEPA model that utilizes Rectified Distribution Matching Regularization (RDMReg) to encourage sparse representations. This approach contrasts with traditional metho…
-
New methods enhance AI planning with latent world models · 4 sources tracked
Researchers have developed new methods to improve planning in latent world models, which are systems that predict outcomes of action sequences. One approach, Reinforced Planning (RP1), learns to improve multi-step plans…
-
New SCALE method enhances AI planning by improving latent space geometry
Researchers have developed SCALE (State-Calibrated Latent Embeddings), a new method to improve planning in joint-embedding predictive world models. SCALE enhances the geometric properties of latent representations, simi…
-
New AI model Traj-LeWM improves planning with path-aware trajectory costs
Researchers have introduced Traj-LeWM, an enhanced visual world model designed to improve planning capabilities in AI agents. This new model addresses limitations in its predecessor, LeWM, by incorporating trajectory-le…
-
Object-centric world models show improved planning and robustness
Researchers have conducted a study on object-centric world models (OCWMs) for visual model-predictive control, investigating the impact of representation quality and robustness under distribution shifts. The study found…
-
QQWorld enhances latent world model regularization with quantile-quantile matching
Researchers have introduced QQWorld, a novel method for regularizing latent world models. This new approach addresses limitations in existing Epps-Pulley (EP) objectives, which can struggle to control heavy-tailed devia…
-
INTACT system enables search-free intent-to-action learning for world models
Researchers have developed INTACT, a novel approach to intent-to-action learning for world models. This end-to-end system bypasses the need for expensive test-time search by directly mapping latent motion intent to acti…
-
Temporal-Distance JEPA enhances world model predictive control
Researchers have introduced Temporal-Distance JEPA (TD-JEPA), a novel approach to representation learning for latent world model predictive control. This method enhances Joint-Embedding Predictive Architectures (JEPAs) …
-
Qantara JEPA model enables multi-paradigm control from single checkpoint
Researchers have introduced Qantara, a novel Joint-Embedding Predictive Architecture (JEPA) that enables a single model checkpoint to support multiple inference paradigms for control from raw pixels. Unlike previous JEP…
-
Qantara JEPA enables multi-paradigm control from pixels
Researchers have introduced Qantara, a novel Joint-Embedding Predictive Architecture (JEPA) that enables multi-paradigm control from raw pixels. Unlike previous JEPAs that commit to a single inference method at training…
-
New JEPA method disentangles task progression from content
Researchers have developed a new method called Subspace-Decomposed JEPAs (SD-JEPA) to improve latent world models. This approach disentangles task progression from content within the model's latent space, using separate…