3 papers
cs.LG2026
Structural Instability of Feature Composition
Yunpeng Zhou
Sparse Autoencoders (SAEs) have emerged as a powerful paradigm for disentangling feature superposition in transformer-based architectures, enabling precise control via activation s…
cs.CV2026
Pixelis: Reasoning in Pixels, from Seeing to Acting
Yunpeng Zhou
Most vision-language systems are static observers: they describe pixels, do not act, and cannot safely improve under shift. This passivity limits generalizable, physically grounded…
cs.AI2026
Sparse Visual Thought Circuits in Vision-Language Models
Yunpeng Zhou
Sparse autoencoders (SAEs) improve interpretability in multimodal models, but it remains unclear whether SAE features form modular, composable units for reasoning-an assumption und…