🏢 Computer Science Department, TU Darmstadt
Neural Concept Binder
·3025 words·15 mins·
loading
·
loading
Computer Vision
Visual Question Answering
🏢 Computer Science Department, TU Darmstadt
The Neural Concept Binder (NCB) framework learns expressive, inspectable, and revisable visual concepts unsupervised, integrating both continuous and discrete representations for seamless use in neura…
Interpretable Concept Bottlenecks to Align Reinforcement Learning Agents
·2372 words·12 mins·
loading
·
loading
Machine Learning
Reinforcement Learning
🏢 Computer Science Department, TU Darmstadt
Successive Concept Bottleneck Agents (SCoBots) improve reinforcement learning by integrating interpretable layers, enabling concept-level inspection and human-in-the-loop revisions to fix misalignment…