Papers by Ziyu Lyu
Please refuse to answer me! Mitigating Over-Refusal in Large Language Models via Adaptive Contrastive Decoding (2026.acl-long)
Copied to clipboard
| Challenge: | Existing methods for mitigating over-refusal can't maintain low refusal ratio for harmless queries while keeping high for malicious queries. |
| Approach: | They propose a model-agnostic approach to mitigate over-refusal in large language models . they propose an adaptive contrastive decoding strategy that incorporates or removes the refusal token distribution . |
| Outcome: | The proposed approach reduces the refusal ratio for over-refusal queries by 10.35% while increasing the refusal rate for malicious queries by 0.13%. |
World Models with Hints of Large Language Models for Goal Achieving (2025.naacl-long)
Copied to clipboard
| Challenge: | Existing methods address this by adding intrinsic rewards, but they fail to provide meaningful guidance in long-horizon decision-making tasks with large state and action spaces lacking purposeful exploration. |
| Approach: | They propose a multi-modal model-based RL approach that integrates the proposed hinting subgoals into the model rollouts to encourage goal discovery and reaching in challenging tasks. |
| Outcome: | The proposed model outperforms existing methods in challenging, sparse-reward environments such as HomeGrid, Crafter, and Minecraft by 41.8%, 21.1%, and 9.9%. |
MidPO: Dual Preference Optimization for Safety and Helpfulness in Large Language Models via a Mixture of Experts Framework (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Recent studies address safety-constrained online and offline preferences optimizations, but offline methods perform poorly in adaptively balancing safety and helpfulness. |
| Approach: | They propose a mixture of experts framework for safety-helpfulness dual Preference Optimization . they combine a single-preference enhanced direct preference optimization approach with a dynamic routing mechanism . |
| Outcome: | The proposed framework outperforms state-of-the-art methods in safety and helpfulness. |