MG-Verilog: Multi-grained Dataset Towards Enhanced LLM-assisted Verilog Generation
Received the Best Paper Award at the inaugural IEEE LAD 2024 workshop on LLM-Aided Design.
News & Highlights
Recent milestones, publications, and recognitions.
Unveiling and Harnessing Hidden Attention Sinks
Our attention calibration framework for LLMs has been accepted to ICML 2024.
EDGE-LLM
Layer-wise compression and adaptive tuning techniques for on-device LLM adaptation accepted by DAC 2024.
GPT4AIGChip
Our LLM-driven accelerator design framework has been accepted by ICCAD 2023.
Gen-NeRF Demo
Won 2nd place in the University Demo Best Demonstration Award at DAC 2023.
Master-ASR
Modularized multilingual ASR system accepted by ICML 2023.
Hint-Aug
Few-shot ViT tuning framework accepted by CVPR 2023.
NetBooster
Efficiency boosting framework for tiny neural networks accepted by DAC 2023.
Research Focus
I study how to make large language models adaptable, efficient, and dependable in real-world settings. My research spans three pillars:
- Inference Calibration: Elevating LLM outputs with lightweight attention refinements and dynamic evaluation strategies.
- Adaptive Tuning: Designing compression-aware fine-tuning and voting schemes so edge devices can host powerful LLM experiences.
- LLM-for-Hardware: Co-designing silicon and software by steering LLMs to generate and verify hardware designs with human-centered workflows.
Previously, I earned my Ph.D. in Computer Science from Georgia Tech, advised by Prof. Yingyan (Celine) Lin. I also hold an M.S. from Columbia University and a B.Eng. from Zhejiang University, and I have collaborated with MIT-IBM Watson AI Lab.
Selected Publications
* denotes equal contribution
Unveiling and Harnessing Hidden Attention Sinks: Enhancing Large Language Models without Training through Attention Calibration
A systematic study of attention sink behaviors in LLMs paired with an inference-time calibration method.
Paper
EDGE-LLM: Enabling Efficient Large Language Model Adaptation on Edge Devices via Layerwise Unified Compression and Adaptive Layer Tuning and Voting
Compression-aware fine-tuning that unlocks on-device LLM adaptation with minimal hardware overhead.
Paper
MG-Verilog: Multi-grained Dataset Towards Enhanced LLM-assisted Verilog Generation
The first multi-granularity Verilog dataset enabling precise LLM-guided hardware code generation.
Paper
GPT4AIGChip: Towards Next-Generation AI Accelerator Design Automation via Large Language Models
Human-in-the-loop workflows that translate natural language into accelerator design artifacts.
Paper
Master-ASR: Achieving Multilingual Scalability and Low-Resource Adaptation in ASR with Modularized Learning
A modular ASR architecture that balances multilingual performance and low-resource specialization.
Paper
Hint-Aug: Drawing Hints from Vision Foundation Models towards Boosted Few-shot Parameter-Efficient ViT Tuning
Integrating attention-aware data augmentation to amplify few-shot ViT tuning effectiveness.
Paper
NetBooster: Empowering Tiny Deep Learning By Standing on the Shoulders of Deep Giants
A blueprint for uplifting compact neural networks via expansion-then-contraction strategies.
Paper
EyeCoD: Eye Tracking System Acceleration via FlatCam-based Algorithm/Hardware Co-Design
A compact lensless eye-tracking system achieving high throughput through algorithm-hardware co-design.
Paper
AugViT: Improving Vision Transformer Training by Marrying Attention and Data Augmentation
An attention-aware augmentation framework that consistently lifts ViT accuracy across hardware tiers.
Paper
MIA-Former: Efficient and Robust Vision Transformers via Multi-grained Input-Adaptation
A multi-grained input-adaptive ViT that dynamically adjusts depth, heads, and tokens.
Paper
LDP: Learnable Dynamic Precision for Efficient Deep Neural Network Training and Inference
Temporal and spatial precision scheduling for DNN training that optimizes accuracy-efficiency trade-offs.
PaperAcademic Service
Conference Reviewer
ICLR 2023, NeurIPS 2022–2023, ICML 2023, CVPR 2023, AAAI 2022, AICAS 2022.
Artifact Evaluation
MICRO 2023 Artifact Evaluation Committee.