Ivor Tsang

Principal Researcher@A*STAR·新加坡·OpenReview

研究方向

Transfer Learning · Learning from Noisy Supervision · Deep Generative Models · Kernel methods · Large-scale optimization · Reinforcement learning · Multi-label prediction

Ivor Tsang

Sample Reward Soups: Query-efficient Multi-Reward Guidance for Text-to-Image Diffusion Models

Mitigating Mismatch within Reference-based Preference Optimization

Why Settle for One? Text-to-ImageSet Generation and Evaluation

MermaidFlow: Redefining Agentic Workflow Generation via Safety-Constrained Evolutionary Programming

FZOO: Fast Zeroth-Order Optimizer for Fine‑Tuning Large Language Models towards Adam‑Scale Speed

FlowSearcher: Synthesizing Memory-Guided Agentic Workflows for Web Information Seeking

Dataset Color Quantization: A Training-Oriented Framework for Dataset-Level Compression

Personalized Vision via Visual In-Context Learning

Efficient Stochastic Algorithms for Continual Finite-Sum Minimax Optimization

OSCAR: Orthogonal Stochastic Control for Alignment-Respecting Diversity in Flow Matching

DSA: Efficient Inference For Video Generation Models via Distributed Sparse Attention

EvoCF: Multi-agent Collaboration with Memory-guided Evolutionary Counterfactual Planning

Verification and Co-Alignment via Heterogeneous Consistency for Preference-Aligned LLM Annotations

Stochastic Gaussian Zeroth-Order Optimization: Improved Convergence Analysis under Skewed Hessian Spectra

Model-Agnostic Text Condensation with Coherence Awareness

A Diffusion-based Foundation Model for Irregular Spatio-Temporal Trajectories

Causal Scaffolding for Physical Reasoning: A Benchmark for Causally-Informed Physical World Understanding in VLMs

FM-IRL: Flow-Matching for Reward Modeling and Policy Regularization in Reinforcement Learning

Lang-PINN: From Language to Physics-Informed Neural Networks via a Multi-Agent Framework

Domain Bridging: Enabling Adaptation without Peeking at Target Data

Value-Alignment via Safe Semantic Manifold-Constrained Latent Diffusion

HyperSyn: Synthesizing Instance-wise Model by Fusing Blackbox Expert via Hypernetwork

UniErase: Towards Balanced and Precise Unlearning in Language Models

Numerical Sensitivity and Robustness: Exploring the Flaws of Mathematical Reasoning in Large Language Models

InstructFlow: Adaptive Symbolic Constraint-Guided Code Generation for Long-Horizon Planning

Analytical Construction on Geometric Architectures: Transitioning from Static to Temporal Link Prediction

DepthVanish: Optimizing Adversarial Interval Structures for Stereo-Depth-Invisible Patches

Fast Direct: Query-Efficient Online Black-box Guidance for Diffusion-model Target Generation

Sharpness-Aware Black-Box Optimization

AngleRoCL: Angle-Robust Concept Learning for Physically View-Invariant Adversarial Patches

Diversifying Policy Behaviors with Extrinsic Behavioral Curiosity

Training-Free Dataset Pruning for Instance Segmentation

Beyond-Expert Performance with Limited Demonstrations: Efficient Imitation Learning with Double Exploration

Plasticity from Structured Sparsity: Mastering Continual Reinforcement Learning through Fine-grained Network Allocation and Dormant Neuron Exploration

Second-Order Fine-Tuning without Pain for LLMs: A Hessian Informed Zeroth-Order Optimizer

Personalized Vision via Visual In-Context Learning

Sanitizing LLMs: Retrospective Learning for Self-Correction of Inconsistent Samples via User Preferences

Beautifying Diffusion Models: Learning Context-Aware Filters for Robust Dense Prediction on Test-Time Corrupted Images

ProAdvPrompter: A Two-Stage Journey to Effective Adversarial Prompting for LLMs

A Computation and Communication Efficient Projection-free Algorithm for Decentralized Constrained Optimization

Quality Diversity Imitation Learning

Predicting User Behaviors with Scene via Dual Sequence Networks

CamoVid60K: A Large-Scale Video Dataset for Moving Camouflaged Animals Understanding

Time-variant Duo-image Inpainting via Interactive Distribution Transition Estimation

The Advancement in Stochastic Zeroth-Order Optimization: Mechanism of Accelerated Convergence of Gaussian Direction on Objectives with Skewed Hessian Eigenvalues

Multi-view Consistent Image Generation through Self-calibrated Latent Refinement

CRVR: Continuous Representation-Driven Video Frame Modulation Against rPPG Heart Rate Measurement

Memorisable Prompting: Preventing LLMs Forgetting False Positive Alarm