Chengru Song

发表论文 26 篇 · 总被引 304 次 · h-index 6

代表论文

• Video-LaVIT: Unified Video-Language Pre-training with Decoupled Visual-Motional Tokenization (2024 · International Conference on Machine Learning · 被引 103)
• Kwai Keye-VL Technical Report (2025 · 被引 32)
• Accelerating the Training of Large Language Models using Efficient Activation Rematerialization and Optimal Hybrid Parallelism (2024 · USENIX Annual Technical Conference · 被引 30)
• VidCapBench: A Comprehensive Benchmark of Video Captioning for Controllable Text-to-Video Generation (2025 · Annual Meeting of the Association for Computational Linguistics · 被引 23)
• Privacy-Preserving DNN Training with Prefetched Meta-Keys on Heterogeneous Neural Network Accelerators (2023 · Design Automation Conference · 被引 17)
• SlimPipe: Memory-Thrifty and Efficient Pipeline Parallelism for Long-Context LLM Training (2025 · International Conference on Software Composition · 被引 9)