Papers by Xiangchi Yuan
MTMCS-Bench: Evaluating Contextual Safety of Multimodal Large Language Models in Multi-Turn Dialogues (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing contextual safety benchmarks are mostly single-turn and miss how malicious intent can emerge gradually or how the same scene can support both benign and exploitative goals. |
| Approach: | They propose a benchmark that evaluates contextual safety in multimodal large language models . they observe persistent trade-offs between contextual safety and utility . |
| Outcome: | The proposed model combines multi-turn and multi-switch scenarios to evaluate safety in multimodal large language models. |
Modality-Aware Neuron Pruning for Unlearning in Multimodal Large Language Models (2025.acl-long)
Copied to clipboard
| Challenge: | Large Language Models and Multimodal Large Language Modells can memorize sensitive information, raising ethical and privacy concerns. |
| Approach: | They propose a novel unlearning framework that selectively clips neurons based on their relative importance to the targeted forget data. |
| Outcome: | The proposed framework selectively clips neurons based on their relative importance to the targeted forget data, curated for different modalities. |
Superficial Self-Improved Reasoners Benefit from Model Merging (2025.emnlp-main)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) rely heavily on large-scale reasoning data, but as data becomes scarce, model self-improvement offers a promising alternative. |
| Approach: | They propose to merge the weights of original and self-improved LLMs to mitigate model collapse and improve generalized reasoning capability. |
| Outcome: | The proposed model merge mitigates model collapse and improves generalized reasoning capability. |
What Makes a Good Curriculum? Disentangling the Effects of Data Ordering on LLM Mathematical Reasoning (2026.acl-long)
Copied to clipboard
| Challenge: | Curriculum learning (CL) orders data corpus by difficulty, but prior work employs disparate difficulty metrics and training setups. |
| Approach: | They propose a framework that decomposes curriculum difficulty into five dimensions: Problem Difficulty, Model Surprisal, Confidence Margin, Predictive Uncertainty and Decision Variability. |
| Outcome: | The proposed framework decomposes curriculum difficulty into five dimensions . the results show that no curriculum strategy dominates universally . |
Growing Through Experience: Scaling Episodic Grounding in Language Models (2025.acl-long)
Copied to clipboard
| Challenge: | Language models (LMs) require effective episodic grounding to perform well at physical planning tasks due to their limited ability to learn from and apply past experiences. |
| Approach: | They propose a weak-to-strong episodic learning framework that integrates episodic memory into hierarchical representations and pre-trained knowledge to unlock larger LMs' potential for grounding. |
| Outcome: | The proposed framework outperforms top proprietary LMs by 3.45% across diverse planning and question-answering tasks. |
Behavior Knowledge Merge in Reinforced Agentic Models (2026.acl-long)
Copied to clipboard
| Challenge: | Existing methods for supervised fine-tuning (SFT) are suboptimal to preserve task-specific capabilities on RL-trained agentic models. |
| Approach: | They propose a distribution-aware merging framework specifically designed for RL-trained agentic models that disentangles shared and task-specific unique parameter updates while selectively preserving and rescaling unique ones. |
| Outcome: | Experiments across multiple agent domains and model architectures show that the proposed framework surpasses baselines and unlocks synergistic potential among agents. |