Jian Liu

MECT, a speaker verification model that integrates the Mixture-of-Experts (MoE) mechanism into a CNN-Transformer backbone with optimized block structure and stacking scheme, achieves state-of-the-art performance on VoxCeleb1 and delivers strong results on CN-Celeb, demonstrating its effectiveness across diverse datasets.

GeoAAC is proposed, a geometry-based adaptive action chunking method for flow-based VLA policies that adjusts the action horizon according to the reliability of the current action prediction, with geometric variation across action prefixes remaining positively correlated with predictive uncertainty.

Realtime-Venus, a proactive full-duplex interaction system with two separately trained 9B models: Realtime-Venus-Omni for audio-visual interaction and Realtime-Venus-Audio for spoken interaction, which leads the compared models on MMAU, Llama Questions, and Speech CMMLU.

Real-time emotion predictions can be made more reliable by selectively invoking multimodal reasoning only when necessary, improving accuracy without sacrificing speed.

Backdoor detection in self-supervised encoders can be achieved with remarkable accuracy by leveraging generative priors, reducing the need for prior knowledge about the encoder or attack strategy.

Current OmniLLMs stumble when processing real-world, long-form audio-visual content, achieving only ~35-65% accuracy on a new benchmark designed to test long-term memory and fine-grained understanding.