13 papers
RACER: Retrieval-Augmented Contextual Rapid Speculative Decoding
Zihong Zhang, Zuchao Li, Lefei Zhang +2
Autoregressive decoding in Large Language Models (LLMs) generates one token per step, causing high inference latency. Speculative decoding (SD) mitigates this through a guess-and-v…
TrigReason: Trigger-Based Collaboration between Small and Large Reasoning Models
Yi Zhao, Yajuan Peng, Cam-Tu Nguyen +4
Large Reasoning Models (LRMs) achieve strong performance on complex tasks through extended chains of thought but suffer from high inference latency due to autoregressive reasoning.…
SmallKV: Small Model Assisted Compensation of KV Cache Compression for Efficient LLM Inference
Yi Zhao, Yajuan Peng, Cam-Tu Nguyen +4
KV cache eviction has emerged as an effective solution to alleviate resource constraints faced by LLMs in long-context scenarios. However, existing token-level eviction methods oft…
Scaling LLM Speculative Decoding: Non-Autoregressive Forecasting in Large-Batch Scenarios
Luohe Shi, Zuchao Li, Lefei Zhang +3
Speculative decoding accelerates LLM inference by utilizing otherwise idle computational resources during memory-to-chip data transfer. Current speculative decoding methods typical…
ToM: Leveraging Tree-oriented MapReduce for Long-Context Reasoning in Large Language Models
Jiani Guo, Zuchao Li, Jie Wu +5
Large Language Models (LLMs), constrained by limited context windows, often face significant performance degradation when reasoning over long contexts. To address this, Retrieval-A…
XQuant: Achieving Ultra-Low Bit KV Cache Quantization with Cross-Layer Compression
Haoqi Yang, Yao Yao, Zuchao Li +3
Large Language Models (LLMs) have demonstrated remarkable capabilities across diverse natural language processing tasks. However, their extensive memory requirements, particularly…