Search papers, labs, and topics across Lattice.
Shanghai Conservatory of Music
5
0
9
UniVoice achieves competitive performance in both speech and singing voice synthesis by cleverly separating melody control from speech prosody, revolutionizing how we think about unified voice generation.
Forget hand-crafted reward functions: $\text{RLR}^3$ leverages rubrics and LLMs to provide fine-grained, multi-criteria supervision, outperforming standard RLVR in vision-language tasks.
Multi-behavior recommendation models can now effectively filter out noisy auxiliary data and boost performance by contrasting denoised auxiliary and target behavior representations.
Fine-grained rubrics unlock significantly better visual reasoning in preference optimization, rivaling GPT-5.4 with a much smaller model.
Static public transport timetables are misleading: this framework uses real-time location data to build corrected, empirical timetables at national scale.
