Papers by Tongtong Yang

3 papers
MPO: Multilingual Safety Alignment via Reward Gap Optimization (2025.acl-long)

Copied to clipboard

Challenge: Existing preference learning methods for safety alignment are monolingual and struggle with noisy multilingual data.
Approach: They propose a multilingual reward gaP optimization approach that leverages the well-aligned safety capabilities of the dominant language to improve safety alignment across multiple languages.
Outcome: Extensive experiments on three LLMs, LLaMA-3.1, Gemma-2 and Qwen2.5, validate MPO’s efficacy in multilingual safety alignment without degrading general multilingual utility.
Real-time Ad Retrieval via LLM-generative Commercial Intention for Sponsored Search Advertising (2025.emnlp-main)

Copied to clipboard

Challenge: Existing methods for retrieving documents and ads use one-to-few mappings and time-consuming content extraction.
Approach: They propose a framework that leverages LLM-generated commercial intents as an intermediate semantic representation to directly retrieve ads for queries in real-time.
Outcome: The proposed framework has been implemented in a real-world online system, handling daily search volumes in billions.
LLEOT: A Privacy-Enhancing Offsite Tuning Framework via Loss Landscape Elevation (2026.findings-acl)

Copied to clipboard

Challenge: Existing approaches to fine-tune large language models are infeasible due to privacy regulations.
Approach: They propose an offsite tuning framework that secures data privacy and model parameter and capability privacy.
Outcome: The proposed framework secures data privacy and model parameter and capability privacy while preserving gradient alignment.

What is GenGO?

GenGO is an NLP powered publication search system. It currenctly indexes 30k+ papers from ACL Anthology, and implements multi-aspect summarization, semantic search, and more!

Information

About
Limitations