11 papers
Sparse-to-Dense: A Free Lunch for Lossless Acceleration of Video Understanding in LLMs
Xuan Zhang, Cunxiao Du, Sicheng Yu +4
Due to the auto-regressive nature of current video large language models (Video-LLMs), the inference latency increases as the input sequence length grows, posing challenges for the…
LightTransfer: Your Long-Context LLM is Secretly a Hybrid Model with Effortless Adaptation
Xuan Zhang, Fengzhuo Zhang, Cunxiao Du +4
Scaling language models to handle longer contexts introduces substantial memory challenges due to the growing cost of key-value (KV) caches. Motivated by the efficiency gains of hy…
LongSpec: Long-Context Lossless Speculative Decoding with Efficient Drafting and Verification
Penghui Yang, Cunxiao Du, Fengzhuo Zhang +4
As Large Language Models (LLMs) can now process extremely long contexts, efficient inference over these extended inputs has become increasingly important, especially for emerging a…
Demystifying the Slash Pattern in Attention: The Role of RoPE
Yuan Cheng, Fengzhuo Zhang, Yunlong Hou +5
Large Language Models (LLMs) often exhibit slash attention patterns, where attention scores concentrate along the -th sub-diagonal for some offset . These patterns play a k…
Annealed Relaxation of Speculative Decoding for Faster Autoregressive Image Generation
Xingyao Li, Fengzhuo Zhang, Cunxiao Du +1
Despite significant progress in autoregressive image generation, inference remains slow due to the sequential nature of AR models and the ambiguity of image tokens, even when using…
BanditSpec: Adaptive Speculative Decoding via Bandit Algorithms
Yunlong Hou, Fengzhuo Zhang, Cunxiao Du +6
Speculative decoding has emerged as a popular method to accelerate the inference of Large Language Models (LLMs) while retaining their superior text generation performance. Previou…