Showing 2026Show all
2 papers · 1 filter
cs.AI2026
Do Dynamic Routers Need Memory? HeRo: History-Aware Routing for Efficient LLM Inference
Hongjin Lin, Wentao Wan, Keze Wang
Dynamic layer routing reduces the inference cost of Large Language Models (LLMs) by learning to skip layers for individual tokens. Existing methods, however, treat each routing dec…
cs.AI2026
ORACLE: Optimizing Reasoning Abilities of Large Language Models via Constraint-Led Synthetic Data Elicitation
Zhuojie Yang, Wentao Wan, Keze Wang
Training large language models (LLMs) with synthetic reasoning data has become a popular approach to enhancing their reasoning capabilities, while a key factor influencing the effe…