Papers by Zekai Lin

7 papers
English as Defense Proxy: Mitigating Multilingual Jailbreak via Eliciting English Safety Knowledge (2025.findings-emnlp)

Copied to clipboard

Challenge: Large language models excel in many tasks, but their safety guarantees vary by language.
Approach: They propose a unified approach that leverages English as a universal safety anchor.
Outcome: The proposed approach leverages English as defense proxy (E-Proxy) to transfer safety knowledge across languages.
Parameter Importance is Not Static: Evolving Parameter Isolation for Supervised Fine-Tuning (2026.acl-long)

Copied to clipboard

Challenge: Recent approaches to fine-tuning of large language models suffer from task interference and catastrophic forgetting.
Approach: They propose a fine-tuning framework that adapts isolation decisions based on online estimates of parameter importance.
Outcome: The proposed framework reduces interference and forgetting while releasing outdated parameters to recover plasticity.
BanditMTL: Bandit-based Multi-task Learning for Text Classification (2021.acl-long)

Copied to clipboard

Challenge: Existing methods to regularize task variance are unexplored in multi-task text classification.
Approach: They propose a multi-task learning method based on adversarial multi-armed bandit to regularize the task variance by means of a mirror gradient ascent-descent algorithm.
Outcome: The proposed method achieves state-of-the-art in multi-task text classification.
MASPO: Unifying Gradient Utilization, Probability Mass, and Signal Reliability for Robust and Sample-Efficient LLM Reasoning (2026.acl-long)

Copied to clipboard

Challenge: Existing RLVR algorithms rely on rigid, uniform, and symmetric trust region mechanisms . current algorithms lack robustness, asymmetric signal reliability and inefficient gradient utilization .
Approach: They propose a framework to harmonize three dimensions of RLVR algorithms, a paper argues . a binary cutoff is used to discard valuable reinforcement signals, they argue .
Outcome: The proposed framework outperforms baselines in evaluating a robust RLVR solution.
From log 𝜋 to 𝜋: Taming Divergence in Soft Clipping via Bilateral Decoupled Decay of Probability Gradient Weight (2026.acl-long)

Copied to clipboard

Challenge: Standard algorithms for Large Language Models (LLMs) enforce stability via "hard clipping" but relying on log-probability gradient yields divergent weights as probabilities vanish, destabilizing LLM training.
Approach: They propose a decoupled gradient policy optimization that uses a decay mechanism to decouple the probability of a boundary token.
Outcome: The proposed algorithm outperforms baselines on various mathematical benchmarks.
Read it in Two Steps: Translating Extremely Low-Resource Languages with Code-Augmented Grammar Books (2025.acl-long)

Copied to clipboard

Challenge: Using code rules improves rule retrieval and application of grammar books in low-resource languages.
Approach: They propose to decompose a grammar rule retrieval and application step into two steps . they propose to represent grammar rules as code functions to facilitate LLM reasoning .
Outcome: The proposed model significantly boosts rule retrieval and application, resulting in 13.1% BLEU improvement.
MetaWeighting: Learning to Weight Tasks in Multi-Task Learning (2022.findings-acl)

Copied to clipboard

Challenge: Existing task weighting methods assign weights only based on training losses, while ignoring the gap between the training loss and generalization loss.
Approach: They propose a task weighting algorithm which automatically weights the tasks via a learning-to-learn paradigm and a multi-task text classification paradigm.
Outcome: Extensive experiments show that the proposed method outperforms existing methods in multi-task text classification.

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