Event
ConfigWM: A Configurable Perception World Model for Task-Conditioned State Extraction
1 reports · 1 sources
The ConfigWM authors propose a configurable perception mechanism that uses the natural-language description of a task as a configuration signal, guiding visual embeddings to be mapped into a task-specific state space, selectively retaining task-relevant information and suppressing distractors. In evaluation environments with progressively increasing distractors, the planning success rates of existing methods drop significantly, whereas ConfigWM remains relatively stable; the authors report that it is on average 22.3% higher than the strongest competing method. Physical-quantity probing (Pearson r +0.365) and retrieval from distractor to clean environments (+75.5%) also show consistent improvements.
Report timeline
- ConfigWM: Configurable-Perception World Models via Task-Conditioned State Extraction
ConfigWM proposes a configurable perception mechanism that uses task language descriptions as configuration signals to guide the mapping of visual embeddings into task-specific state spaces, selectively retaining task-relevant information and suppressing distractions. In environments with progressively increasing distraction, existing methods show a marked decline in planning success rate, while ConfigWM remains relatively stable, averaging 22.3% higher than the strongest competing method. Physical-quantity probing (Pearson r +0.365) and distracted-to-clean retrieval (+75.5%) also show consistent improvements.
HuggingFace Daily Papers ·