9 papers
On-Policy Parameter Update Direction Underlies Generalization in LLM Post-Training
Shufan Shen, Zhongni Hou, Junshu Sun +5
The strong generalization performance of on-policy post-training paradigms has motivated studies of their parameter update behaviors. However, these studies treat the observed beha…
VCN-Bench: A Video-Contextualized Navigation Benchmark for Spatial Reasoning over Prior Visual Experience
Siqi Zhang, Meng Wei, Chenyang Wan +6
Spatial reasoning is fundamental to embodied agents, yet it remains unclear whether spatial understanding can be carried forward to guide sequential interactions. Existing spatial-…
ActiveScope: Actively Seeking and Correcting Perception for MLLMs
Yajing Wang, Chao Bi, Junshu Sun +4
Multimodal Large Language Models (MLLMs) have demonstrated impressive vision-language understanding, yet still struggle with fine-grained perception in high-resolution images. Whil…
Locate-then-Sparsify: Attribution Guided Sparse Strategy for Visual Hallucination Mitigation
Tiantian Dang, Chao Bi, Shufan Shen +3
Despite the significant advancements in Large Vision-Language Models (LVLMs), their tendency to generate hallucinations undermines reliability and restricts broader practical deplo…
Enhancing Pre-trained Representation Classifiability can Boost its Interpretability
Shufan Shen, Zhaobo Qi, Junshu Sun +3
The visual representation of a pre-trained model prioritizes the classifiability on downstream tasks, while the widespread applications for pre-trained visual models have posed new…
Kernelized Sparse Fine-Tuning with Bi-level Parameter Competition for Vision Models
Shufan Shen, Junshu Sun, Shuhui Wang +1
Parameter-efficient fine-tuning (PEFT) aims to adapt pre-trained vision models to downstream tasks. Among PEFT paradigms, sparse tuning achieves remarkable performance by adjusting…
Edit Less, Achieve More: Dynamic Sparse Neuron Masking for Lifelong Knowledge Editing in LLMs
Jinzhe Liu, Junshu Sun, Shufan Shen +2
Lifelong knowledge editing enables continuous, precise updates to outdated knowledge in large language models (LLMs) without computationally expensive full retraining. However, exi…
VL-SAE: Interpreting and Enhancing Vision-Language Alignment with a Unified Concept Set
Shufan Shen, Junshu Sun, Qingming Huang +1
The alignment of vision-language representations endows current Vision-Language Models (VLMs) with strong multi-modal reasoning capabilities. However, the interpretability of the a…
Expanding Sparse Tuning for Low Memory Usage
Shufan Shen, Junshu Sun, Xiangyang Ji +2
Parameter-efficient fine-tuning (PEFT) is an effective method for adapting pre-trained vision models to downstream tasks by tuning a small subset of parameters. Among PEFT methods,…