Search papers, labs, and topics across Lattice.
The Hong Kong University of Science and Technology (
7
0
12
VoiceTTA transforms zero-shot TTS by enabling real-time adaptation to unseen speaking styles, achieving remarkable improvements in performance on challenging prompts.
PhaseWin slashes the computational cost of visual attribution from quadratic to linear while preserving high faithfulness, revolutionizing how we interpret model decisions.
Fewer than two edits can enable 76.2% of unsafe images to bypass safety classifiers while retaining their malicious intent, exposing critical vulnerabilities in current moderation systems.
MATE achieves a 4.75% boost in robotic manipulation success rates under data scarcity by leveraging specialized expert routing across heterogeneous inputs.
Forget supervised fine-tuning: RL alone can unlock high-quality chain-of-thought reasoning in audio-language models, even starting from a model with no prior CoT capability.
Generate realistic SAR imagery globally from just coordinates – no more data scarcity bottleneck.
Achieve superior multi-objective molecule generation by treating it as a condition-aware denoising problem, outperforming SOTA methods in binding affinity and drug-likeness.
