Papers by Weilin Xu
Seq1F1B: Efficient Sequence-Level Pipeline Parallelism for Large Language Model Training (2025.naacl-long)
Copied to clipboard
| Challenge: | Current PP methods face severe bottlenecks, including pipeline bubbles and memory footprint. |
| Approach: | They propose a sequence-level one-forward-one-backward (1F1B) PP method for training LLMs on long sequences with high throughput and memory efficiency. |
| Outcome: | The proposed method achieves 1.14X training throughput with half memory footprint compared to baseline methods . it trains an LLM with 30B parameters on sequences up to 64k tokens using 64X NVIDIA A100 GPUs . |
Improving Empathetic Dialogue Generation by Dynamically Infusing Commonsense Knowledge (2023.findings-acl)
Copied to clipboard
| Challenge: | Existing work on generating empathetic responses by utilizing the speaker's emotion has not been successful. |
| Approach: | They propose an approach which incorporates an adaptive module for commonsense knowledge selection to ensure consistency between the generated empathetic responses and the speaker’s situation. |
| Outcome: | The proposed approach outperforms baseline models in both automatic and human evaluations, exhibiting the generation of more coherent and empathetic responses. |
APB: Accelerating Distributed Long-Context Inference by Passing Compressed Context Blocks across GPUs (2025.acl-long)
Copied to clipboard
Yuxiang Huang, Mingye Li, Xu Han, Chaojun Xiao, Weilin Zhao, Sun Ao, Hao Zhou, Jie Zhou, Zhiyuan Liu, Maosong Sun
| Challenge: | Long-context inference is crucial for advancing large language models, but its prefill speed remains a bottleneck. |
| Approach: | They propose an efficient long-context inference framework that leverages multi-host approximate attention to enhance prefill speed. |
| Outcome: | The proposed framework achieves speedups of 9.2, 4.2, and 1.6 without any degradation in performance. |
Ouroboros: Generating Longer Drafts Phrase by Phrase for Faster Speculative Decoding (2024.emnlp-main)
Copied to clipboard
Weilin Zhao, Yuxiang Huang, Xu Han, Wang Xu, Chaojun Xiao, Xinrong Zhang, Yewei Fang, Kaihuo Zhang, Zhiyuan Liu, Maosong Sun
| Challenge: | Speculative decoding is a widely used method that accelerates the generation process of large language models (LLMs) drafting efficiency has become a bottleneck in the final speedup of speculative drafting, therefore generating longer drafts at less cost can lead to better speedup. |
| Approach: | They propose a method that uses existing model to drafting and target LLM to verify draft in a low-cost parallel manner. |
| Outcome: | The proposed method can achieve speedups of up to 2.4 over speculative decoding and 3.9 over vanilla decoding without fine-tuning draft and target models. |
BMInf: An Efficient Toolkit for Big Model Inference and Tuning (2022.acl-demo)
Copied to clipboard
Xu Han, Guoyang Zeng, Weilin Zhao, Zhiyuan Liu, Zhengyan Zhang, Jie Zhou, Jun Zhang, Jia Chao, Maosong Sun
| Challenge: | Recent years, pre-trained language models (PLMs) have achieved promising results on various NLP tasks. |
| Approach: | They propose an open-source toolkit for big model inference and tuning which can support big model tuning at extremely low computation cost. |
| Outcome: | The proposed toolkit can support big model inference and tuning at extremely low computation cost. |
APB-V: Accelerating Long-Video Understanding via Sequence-Parallelism-aware Approximate Attention (2026.acl-long)
Copied to clipboard
Yuxiang Huang, Mingye Li, Xu Han, Chaojun Xiao, Weilin Zhao, Ao Sun, Ziqi Yuan, Hao Zhou, Fandong Meng, Zhiyuan Liu
| Challenge: | Existing methods for long-video inference use compression or sparse attention . existing methods restrict LMMs from handling longer, more complex videos . |
| Approach: | They propose a sequence-parallel framework with optimized attention that accelerates long-video inference across multiple GPUs. |
| Outcome: | The proposed framework delivers speedups of 12.72x, 1.70x, and 1.18x over FlashAttn, ZigZagRing, and APB without significant performance loss. |
FR-Spec: Accelerating Large-Vocabulary Language Models via Frequency-Ranked Speculative Sampling (2025.acl-long)
Copied to clipboard
Weilin Zhao, Tengyu Pan, Xu Han, Yudi Zhang, Sun Ao, Yuxiang Huang, Kaihuo Zhang, Weilun Zhao, Yuxuan Li, Jie Zhou, Hao Zhou, Jianyong Wang, Maosong Sun, Zhiyuan Liu
| Challenge: | Speculative sampling is an efficient way to accelerate the auto-regressive generation process of large language models. |
| Approach: | They propose a frequency-ranked speculative sampling framework that optimizes draft candidate selection through vocabulary space compression. |
| Outcome: | Experiments show that FR-Spec reduces LM Head computation overhead by 75% while ensuring the equivalence of the final output distribution. |
Soft Token Attacks Cannot Reliably Audit Unlearning in Large Language Models (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Recent work shows that soft token attacks can extract unlearned information from large language models. |
| Approach: | They show that soft token attacks can extract unlearned information from LLMs . |
| Outcome: | The proposed attacks can extract unlearned information from large language models . |
Unilaw-R1: A Large Language Model for Legal Reasoning with Reinforcement Learning and Iterative Inference (2025.emnlp-main)
Copied to clipboard
| Challenge: | Reasoning-focused large language models (LLMs) are rapidly evolving across various domains, yet their capabilities in handling complex legal problems remain underexplored. |
| Approach: | They propose a large language model tailored for legal reasoning with a 7-billion parameter scale and a two-stage training strategy combining Supervised Fine-Tuning and Reinforcement Learning. |
| Outcome: | The proposed model outperforms all models of similar scale on authoritative benchmarks and outperformed Qwen-2.5-7B-Instruct (46.6%) by an average margin of 6.6%. |
BMCook: A Task-agnostic Compression Toolkit for Big Models (2022.emnlp-demos)
Copied to clipboard
Zhengyan Zhang, Baitao Gong, Yingfa Chen, Xu Han, Guoyang Zeng, Weilin Zhao, Yanxu Chen, Zhiyuan Liu, Maosong Sun
| Challenge: | Existing efforts to compress medium-sized models for specific tasks have limited results. |
| Approach: | They propose a task-agnostic compression toolkit for big models that implements quantization, pruning, distillation and MoEfication methods. |
| Outcome: | The proposed tool improves performance on a model with 3 billion parameters by 12x . it also outperforms the original model on three typical NLP benchmarks. |
Beyond the Turn-Based Game: Enabling Real-Time Conversations with Duplex Models (2024.emnlp-main)
Copied to clipboard
Xinrong Zhang, Yingfa Chen, Shengding Hu, Xu Han, Zihang Xu, Yuanwei Xu, Weilin Zhao, Maosong Sun, Zhiyuan Liu
| Challenge: | Large language models (LLMs) are increasingly permeating daily lives and require real-time interactions that mirror human conversations. |
| Approach: | They propose to use time-division-multiplexing to process queries and responses pseudo-simultaneously. |
| Outcome: | The proposed model can listen to users while generating output and adjust to provide instant feedback. |