LancoPKU

Lanco Lab website

Code&Data for the paper "Watch Out for Your Agents! Investigating Backdoor Threats to LLM-Based Agents" [NeurIPS 2024]

Adaptive and Momental Bounds for Adaptive Learning Rate Methods.

Towards Multimodal Video Paragraph Captioning Models Robust to Missing Modality

Repository for Label Words are Anchors: An Information Flow Perspective for Understanding In-Context Learning

Repository for Towards Codable Watermarking for Large Language Models

[Findings of ACL 2023] Communication Efficient Federated Learning for Multilingual Machine Translation with Adapter

A simple module consistently outperforms self-attention and Transformer model on main NMT datasets with SoTA performance.

code for Explicit Sparse Transformer