FineVLA: Fine-Grained Instruction Alignment for Steerable Vision-Language-Action Policies

Xintong Hu^*, Xuhong Huang^*, Jinyu Zhang, Yutong Yao, Yuchong Sun, Qiuyue Wang, Mingsheng Li, Sicheng Xie, Yitao Liu, Junhao Chen, Yixuan Chen, Yingming Zheng, Shuai Bai, Tao Yu

🌐 Project Page Code RoboFine-VLM Benchmark

An open framework for fine-grained VLA supervision, including: (1) FineVLA-Data And Pipeline that unifies 972K trajectories from 10 robot datasets into 47K human-verified fine-grained trajectories; (2) RoboFine-Bench, a 500-video benchmark with 10K+ atomic facts and 1K VQA questions; (3) RoboFine-VLM, a robotics-specialized VLM annotator for scalable trajectory annotation; (4) FineVLA-Policy, a steerable VLA policy achieving 86.8%/82.5% in RoboTwin and 62.7/100 in real-world dual-arm manipulation.

Qwen-VLA: Unifying Vision-Language-Action Modeling across Tasks, Environments, and Robot Embodiments

Qwen Team (Xintong Hu is a core contributor)

Code 🌐 Project Page

• A unified embodied foundation model extending Qwen’s VL stack to action and trajectory generation via a DiT-based decoder, unifying manipulation, navigation, and trajectory prediction.
  
• Achieves 97.9% on LIBERO, 86.1%/87.2% on RoboTwin, 69.0% OSR on R2R, 76.9% OOD success on real-world ALOHA.

MRSAudio: A Large-Scale Multimodal Recorded Spatial Audio Dataset with Refined Annotations

Wenxiang Guo^*, Changhao Pan^*, Zhiyuan Zhu^*, Xintong Hu^*, Yu Zhang^*, Li Tang, Rui Yang, Han Wang, Zongbao Zhang, Yuhan Wang, Yixuan Chen, Hankun Xu, Ke Xu, Pengfei Fan, Zhetao Chen, Yanhao Yu, Qiange Huang, Fei Wu, Zhou Zhao^†

DemoPage Code Dataset

-Datasets: Establish MRSAudio, a 500-hour multimodal spatial audio dataset with ambisonic audio, synchronized video, motion trajectories, and fine-grained annotations (transcripts, lyrics, scores), covering 4 real-world scenarios (daily life/speech/singing/music).
-Benchmark: Unified benchmark for 5 spatial audio tasks (spatialization, text-to-speech, singing synthesis, music generation, sound localization) enabling 3D-aware audio modeling.

Tree of Preferences for Diversified Recommendation

Hanyang Yuan,Ning Tang, Tongya Zheng, Jiarong Xu, Xintong Hu, Renhong Huang, Shunyu Liu, Jiacong Hu, Jiawei Chen, Mingli Song^†

Code

-Abstract: With the help of the agent, complete the user information to improve the diversity of recommendations. Use Agent to help solve the Filter Bubble problem in traditional recommendation algorithms.