🏢 Chinese University of Hong Kong

VFM-6D: a novel framework achieving generalizable object pose estimation for unseen categories by leveraging vision-language models.

Reasoning Boundary Framework (RBF) quantitatively assesses and optimizes chain-of-thought (CoT) in LLMs, offering novel metrics and optimization strategies validated across various models and tasks.

UNICORN: a unified framework reveals that existing offline meta-reinforcement learning algorithms optimize variations of mutual information, leading to improved generalization.

FKEA: a novel, scalable method for reference-free evaluation of generative models’ diversity using random Fourier features, overcoming computational limitations of existing entropy-based scores.

VideoStreaming, a novel vision-language model, enables efficient and accurate understanding of arbitrarily long videos using a constant number of tokens via streaming encoding and adaptive memory sele…

Bias-free performative prediction is achieved using a novel lazy deployment scheme with SGD, handling non-convex loss functions.

Quantum algorithms achieve speedups in non-smooth, non-convex optimization, outperforming classical methods by a factor of ε⁻²/³ in query complexity for finding (δ,ε)-Goldstein stationary points.

This paper proposes ’natural counterfactuals’ for more realistic counterfactual reasoning in AI, using backtracking to minimize deviations from observed data while ensuring feasibility.

Multi-Stage Predict+Optimize tackles optimization problems where parameters are revealed sequentially, improving predictions and decisions through stage-wise updates.

MR-Ben: A new benchmark reveals LLMs’ meta-reasoning flaws, pushing the boundaries of AI evaluation beyond simple accuracy.

Strategic players can manipulate reward signals, but simple threshold policies still achieve a surprisingly good approximation to the optimal prophet value, even in this more realistic setting.

InterControl: Zero-shot multi-person interaction generation by precisely controlling every joint using only single-person data.

HORSE: A novel attention-based neural network significantly improves large-scale neural subset selection by up to 20%, addressing limitations in existing methods.

GREAT Score: A novel framework using generative models for efficiently and accurately evaluating the global robustness of machine learning models against adversarial attacks.

Gradient Cuff: A novel defense mechanism against LLM jailbreaks, leveraging refusal loss landscapes for improved malicious query rejection without harming model performance on benign inputs.

A-SAM: Turning SAM’s inherent ambiguity into an advantage for controllable, diverse, and convincing ambiguous object segmentation.

Score-based diffusion models are improved by a novel coefficient design, enabling efficient adaptation to unknown low-dimensional data structures and achieving a convergence rate of O(k²/√T).