一作ACL 2026Poster· Natural Language Processing
A da F use: Adaptive Ensemble Decoding for Large Language Models
Chengming Cui · Tianxin Wei · Ziyi Chen · Ruizhong Qiu · Zhichen Zeng · Zhining Liu · Xuying Ning · Duo Zhou · Jingrui He
摘要
Large language models (LLMs) exhibit complementary strengths arising from differences in pretraining data, model architectures, and decoding behaviors. Inference-time ensembling provides a practical way to combine these capabilities without retraining. However, existing ensemble approaches suffer...