Publications
Published papers, preprints, and submissions. Venue and status labels distinguish accepted work from manuscripts and earlier versions.
Co-author Institutions
Universities represented by co-authors on my papers.
21 entries
SCOPE: Semantic Bounded Credit Optimization for Prompt Engineering in Text-to-image Generation
arXiv preprint arXiv:2610.22264
RoTA-Bench: Evaluating 3D Spatial Reasoning through Interactive Projection Alignment
A benchmark of interactive 3D spatial reasoning: models use visual feedback to rotate objects until their projections match target silhouettes, across 674 tasks and a fixed 100-task test set.
Duplex-MPE: Benchmarking Multi-Party Interaction in Full-Duplex Dialogue
Continuous-audio evaluation of when an assistant should answer, stay silent, or stop speaking in multi-party full-duplex dialogue.
Diagnosing and Mitigating Perception-Decision Misalignment in Omni-LLMs via Modality Subspace Activation
Diagnoses whether decisions depend on multimodal evidence and rebalances modality contributions during inference without additional training.
WaveZip: Wavelet-Driven Space-Time Decoupling for Video Token Condensation
Training-free video token compression using wavelets to separate temporal redundancy from spatial detail.
Unleashing MLLMs on the Edge: A Unified Framework for Cross-Modal ReID via Adaptive SVD Distillation
A unified model for matching identities across images, infrared, sketches, and text, with knowledge distillation for edge deployment.
FOLD: Fast Correct Speculative Decoding
Earlier speculative-decoding manuscript indexed by Google Scholar. The later submission, Correct While Verifying, is listed separately above.
Training-Free Multimodal Large Language Model Orchestration
A training-free orchestration framework for interactive omni-modal assistants, composing off-the-shelf modality experts through explicit LLM routing, cross-modal memory, and interruption-aware streaming interaction.
A2RBench: An Automatic Paradigm for Formally Verifiable Abstract Reasoning Benchmark Generation
An automatic paradigm for generating abstract reasoning benchmarks with formal verification and scalable answer checking.
Correct While Verifying: Intermediate Target Predictions for Speculative Draft Correction
Uses intermediate target-model states to proactively correct speculative drafts during verification, achieving up to 4.53x speedup over autoregressive decoding.
Event-Anchored Frame Selection for Efficient Long-Video Understanding
A training-free, event-aware frame selection pipeline for efficient long-video understanding.
Mitigating Gradient Interference for Efficient Sparse Fine-Tuning of Large Language Models
Theoretical and practical analysis of gradient interference in sparse PEFT, with an iterative sparse fine-tuning strategy for large language models.
SocialOmni: Benchmarking Audio-Visual Social Interactivity in Omni Models
An offline benchmark for speaker identification, turn-entry decisions, and response generation, using query-time-bounded audio-visual evidence. The current release evaluates 11 omni models on 2,000 perception items and 200 interaction-generation items.
Wavelet-based Frame Selection by Detecting Semantic Boundary for Long Video Understanding
Wavelet-based semantic boundary detection for efficient long-video frame selection.
Event-Anchored Frame Selection for Effective Long-Video Understanding
Event-anchored frame selection for improving the efficiency and coverage of long-video understanding.
CrystaL: Spontaneous Emergence of Visual Latents in MLLMs
Studies the spontaneous emergence of visual latents in multimodal large language models.
An Attention Mechanism-Improved YOLOv7 Object Detection Algorithm for Hemp Duck Count Estimation
Attention-enhanced YOLOv7 for dense hemp duck counting and intelligent livestock monitoring. Included in the full publication list, but not selected as a representative homepage paper.
A Multitask Model for Realtime Fish Detection and Segmentation Based on YOLOv5
A YOLOv5-based multitask model for real-time fish detection and segmentation.
GAM-YOLOv7-tiny and Soft-NMS-AlexNet: Improved Lightweight Sheep Body Object Detection and Pose Estimation Network
Lightweight sheep body detection and pose estimation for practical livestock monitoring.
SDSCNet: An Instance Segmentation Network for Efficient Monitoring of Goose Breeding Conditions
Instance segmentation for efficient monitoring and analysis of goose breeding conditions.
DuckSegmentation: A Segmentation Model Based on the AnYue Hemp Duck Dataset
A duck detection and segmentation dataset/model pipeline for intelligent hemp duck farming.