8 papers
MMLongEmbed: Benchmarking Multimodal Embedding Models in Long-Context Scenarios
Haitian Wang, Ruoxi Sun, Quantong Qiu +5
Recent advancements have significantly expanded the theoretical context windows of Multimodal Embedding Models (MEMs). However, larger context windows do not necessarily translate…
Mechanistic Insights into Functional Sparsity in Multimodal LLMs via CoRe Heads
Ruoxi Sun, Quantong Qiu, Juntao Li +3
While Multimodal Large Language Models (MLLMs) demonstrate remarkable proficiency on complex vision-language tasks, the mechanisms by which they extract query-relevant visual featu…
Flux Attention: Context-Aware Hybrid Attention for Efficient LLMs Inference
Quantong Qiu, Zhiyi Hong, Yi Yang +5
The quadratic computational complexity of standard attention mechanisms presents a severe scalability bottleneck for LLMs in long-context scenarios. While hybrid attention mechanis…
Elastic Attention: Test-time Adaptive Sparsity Ratios for Efficient Transformers
Zecheng Tang, Quantong Qiu, Yi Yang +6
The quadratic complexity of standard attention mechanisms poses a significant scalability bottleneck for large language models (LLMs) in long-context scenarios. While hybrid attent…
LongRM: Revealing and Unlocking the Context Boundary of Reward Modeling
Zecheng Tang, Baibei Ji, Quantong Qiu +4
Reward model (RM) plays a pivotal role in aligning large language model (LLM) with human preferences. As real-world applications increasingly involve long history trajectories, e.g…
CaliDrop: KV Cache Compression with Calibration
Yi Su, Quantong Qiu, Yuechi Zhou +6
Large Language Models (LLMs) require substantial computational resources during generation. While the Key-Value (KV) cache significantly accelerates this process by storing attenti…