Hui Huang

LLMs can achieve better semantic similarity scores with lower-dimensional representations by dynamically selecting relevant semantic components, rather than relying on fixed, high-dimensional hidden states.

LLMs struggle to predict prosecutorial decisions, highlighting a critical blind spot in legal AI's ability to assess criminal liability beyond formally indicted cases.

Domain shifts and novel classes at test time can be tamed by nudging features back towards the source distribution, even for out-of-distribution examples.

Transformer-based architectures can now outperform CNNs in multi-view crowd tracking, especially in large, complex real-world scenes, thanks to a novel view-ground interaction mechanism.

Ditch the sparsity hyperparameter search: sparsemax attention in autoencoders automatically adapts neuron sparsity, boosting reconstruction and concept quality.

Style transfer can now capture the essence of artistic abstraction, not just surface-level appearance, by explicitly reinterpreting image structure.

LLM-based judges, widely used for automated evaluation, are riddled with diverse biases that can be significantly reduced through bias-aware training using RL and contrastive learning.

Cycle-consistent training unlocks robust layered image decomposition in diffusion models, even with complex interactions like shading and reflections.

By explicitly modeling specular effects with view-dependent opacity, this augmented Gaussian Splatting method leapfrogs NeRFs in rendering performance and parameter efficiency.