2 papers
cs.CL2026
Conceptizer: Next-Token Prediction Is Not All You Need
Haoran Zheng, Chen Shani
Next-token prediction (NTP) rewards only the token that occurs next, penalizing every alternative. But natural language often offers many ways to express the same meaning. We intro…
cs.AI2026
When to Think Fast and Slow? AMOR: Adaptive Entropy Gate for Hybrid Models
Haoran Zheng, Chen Shani
Recurrent-attention hybrids aim to combine the efficiency of recurrence with the contextual recall of attention, but existing approaches typically apply attention uniformly across…