2 papers
cs.CL2026
Verbalizing Multi-Token Concepts in LLMs
Xijie Gong, Tonghan Wang, Zimeng Huang
Lens methods inspect model computation by mapping intermediate activations to vocabulary tokens. Yet the concepts humans need to read out often span multiple tokens---entities, phr…
cs.CV2026
FineMoLA: Towards Fine-Grained Motion-Language Alignment from Clip-Level Supervision
Tongyan Wang, Zhengyuan Li, Muhan Lin +5
Text-conditioned human motion generation has made rapid progress with the emergence of large-scale motion--language datasets. However, even datasets with rich long-form description…