Bowen Zeng — accepted AI-conference papers

2026

HARD-KV: Head-Adaptive Regularization for Decoding-time KV Compression

Yuxuan Yang, Feiyang Ren, Bowen Zeng, Dalin Zhang, Jinpeng Chen, Gang Chen +1

ICML 2026poster

Long-context LLM inference faces a fundamental conflict: head-adaptive compression algorithms (e.g., Top-$p$ nucleus sampling) offer superior accuracy by dynamically fluctuating memory budgets, yet modern inference engines (e.g., vLLM) demand rigid, static memory patterns to leverage CUDA Graphs and…

Cited by 0 Source ↗Scholar ↗

2026

Long-Context Attention Benchmark: From Kernel Efficiency to Distributed Context Parallelism

Tao Bu, Qiangang Wang, Bowen Zeng, Hanwen Sun, Yunpeng Huang, Chun Cao +1

ICLR 2026poster

Transformer-based large language models (LLMs) have achieved remarkable success, yet their standard attention mechanism incurs quadratic computation and memory costs with respect to sequence length, posing a major bottleneck for long-context training. Prior work tackles this challenge along two dire…

Cited by 0 Source ↗code ↗Scholar ↗