Xiaojian Ma

MILR: Improving Multimodal Image Generation via Test-Time Latent Reasoning

AdaReP: Plug-and-Play Acceleration for World Model Predictive Control using Adaptive Re-Planning

``Noisier'’ Noise Contrastive Estimation is (Almost) Maximum Likelihood

LEO-VL: Efficient Scene Representation for Scalable 3D Vision-Language Learning

DeepHA: Scaling Action Chains Elicits Deep Hierarchical Agents

COIN: Chain Of INteraction Benchmark: When Reasoning meets Embodied interaction

Multi-modal Agent Tuning: Building a VLM-Driven Agent for Efficient Tool Usage

NEP: Autoregressive Image Editing via Next Editing Token Prediction

Iterative Tool Usage Exploration for Multimodal Agents via Step-wise Preference Tuning

Falcon: Fast Visuomotor Policies via Partial Denoising

GROOT-2: Weakly Supervised Multimodal Instruction Following Agents

LongViTU: Instruction Tuning for Long-Form Video Understanding

Task-oriented Sequential Grounding in 3D Scenes

Optimizing Inference-Time Reasoning in LLMs via Retrieval-Augmented Reflection