15 papers
Selecting Diverse SFT Traces Improves Post-RL Generalization
Dylan Zhang, Mingyuan Wu, Jinning Li
Verified solutions are not equally useful for preparing reasoning models for reinforcement learning (RL). We present a comprehensive study of route diversity, the variation in the…
SLIDEFORGE: An LLM Agent for Controllable Editing of Slides as Structured Artifacts
Haozhen Zheng, Fulin Wang, Tianhu Xiong +6
Current AI agents compellingly describe slides. However, AI-assisted slide editing requires more than understanding: the output must retain layout, style, component structure, and…
SVR-R1: Bootstrapping Multi-modal Reasoning with Self-verification in Reinforcement Learning
Mingyuan Wu, Jingcheng Yang, Shengyi Qian +11
We introduce Self-Verified Reasoner (SVR-R1), a multi-turn RL framework that turns a model's own verification into a learning signal for multimodal reasoning. For each query, the m…
Spreadsheet-RL: Advancing Large Language Model Agents on Realistic Spreadsheet Tasks via Reinforcement Learning
Banghao Chi, Yining Xie, Mingyuan Wu +9
Spreadsheet systems (e.g., Microsoft Excel, Google Sheets) play a central role in modern data-centric workflows. As AI agents grow increasingly capable of automating complex tasks,…
QoS-QoE Translation with Large Language Model
Yingjie Yu, Mingyuan Wu, Ahmadreza Eslaminia +3
QoS-QoE translation is a fundamental problem in multimedia systems because it characterizes how measurable system and network conditions affect user-perceived experience. Although…
Circuit Tracing in Vision-Language Models: Understanding the Internal Mechanisms of Multimodal Thinking
Jingcheng Yang, Tianhu Xiong, Shengyi Qian +2
Vision-language models (VLMs) are powerful but remain opaque black boxes. We introduce the first framework for transparent circuit tracing in VLMs to systematically analyze multimo…
AquaVLM: Improving Underwater Situation Awareness with Mobile Vision Language Models
Beitong Tian, Lingzhi Zhao, Bo Chen +5
Underwater activities like scuba diving enable millions annually to explore marine environments for recreation and scientific research. Maintaining situational awareness and effect…
Spatio-Temporal LLM: Reasoning about Environments and Actions
Haozhen Zheng, Beitong Tian, Mingyuan Wu +3
Despite significant recent progress of Multimodal Large Language Models (MLLMs), current MLLMs are challenged by "spatio-temporal" prompts, i.e., prompts that refer to 1) the entir…
Aha Moment Revisited: Are VLMs Truly Capable of Self Verification in Inference-time Scaling?
Mingyuan Wu, Meitang Li, Jingcheng Yang +6
Inference time techniques such as decoding time scaling and self refinement have been shown to substantially improve mathematical reasoning in large language models (LLMs), largely…
VTool-R1: VLMs Learn to Think with Images via Reinforcement Learning on Multimodal Tool Use
Mingyuan Wu, Jingcheng Yang, Jize Jiang +6
Reinforcement Learning Finetuning (RFT) has significantly advanced the reasoning capabilities of large language models (LLMs) by enabling long chains of thought, self-correction, a…
AquaScope: Reliable Underwater Image Transmission on Mobile Devices
Beitong Tian, Lingzhi Zhao, Bo Chen +5
Underwater communication is essential for both recreational and scientific activities, such as scuba diving. However, existing methods remain highly constrained by environmental ch…
Cache-of-Thought: Master-Apprentice Framework for Cost-Effective Vision Language Model Reasoning
Mingyuan Wu, Jize Jiang, Haozhen Zheng +8
Vision Language Models (VLMs) have achieved remarkable success in a wide range of vision applications of increasing complexity and scales, yet choosing the right VLM model size inv…
UOUO: Uncontextualized Uncommon Objects for Measuring Knowledge Horizons of Vision Language Models
Xinyu Pi, Mingyuan Wu, Jize Jiang +5
Smaller-scale Vision-Langauge Models (VLMs) often claim to perform on par with larger models in general-domain visual grounding and question-answering benchmarks while offering adv…
TraceNet: Segment one thing efficiently
Mingyuan Wu, Zichuan Liu, Haozhen Zheng +4
Efficient single instance segmentation is essential for unlocking features in the mobile imaging applications, such as capture or editing. Existing on-the-fly mobile imaging applic…
360TripleView: 360-Degree Video View Management System Driven by Convergence Value of Viewing Preferences
Qian Zhou, Michael Zink, Ramesh Sitaraman +1
360-degree video has become increasingly popular in content consumption. However, finding the viewing direction for important content within each frame poses a significant challeng…