E-commerce live streaming requires omni-modal understanding of noisy, temporally extended streams, where product facts are distributed across speech, video frames, product images, overlaid text, and user queries. We present TLive-Omni, an omni-modal understanding model tailored to live-commerce scenarios. It maps image...
A production-oriented extension of X-Dub that combines task-adaptive post-training with task-aware few-step distillation that improves lip-sync consistency, identity consistency, and visual quality over X-Dub and largely retains the Teacher's generation quality and audiovisual synchronization.
LLM-based full-duplex voice services allow users to speak while the assistant is responding. Because servers can generate output and advance dialogue state faster than clients can play it, subsequent user speech may be interpreted based on content the user never heard. We call this failure Generative Context Mis-anchor...
Shi-Bo Wang, Zicheng Zhang, Libo Wang et al.· 0 citations
Large language models deployed as personalized assistants must reason over long, evolving interaction histories. However, in long-term dialogue reasoning, relevant evidence is scattered across sessions, preferences may be revised over time, and standard long-context training fails to address these challenges under data...
Na-En Xu, Wan-Qing Cui, Yi-Bo Hu et al.· 0 citations
Decoupled-Live-ProsodyJudge (D-LPJ), a cost-effective pairwise evaluator distilled from Gemini into Qwen3-Omni, and a novel span-local GRPO strategy that applies normalized advantages strictly to their corresponding rationale spans.
Zi-Fan Guan, Long-Yu Lu, Ju-Nan Zhang et al.· 0 citations
Harness-Aware Training (HAT), which trains compact models to adapt to changing Harnesses through reinforcement learning in augmented environments, is proposed and deployed in Taobao Live's digital-avatar service.
T. Sun, Wen-Hao Lin, Yong-Dong Luo et al.· 1 citation
This work proposes AptAvatar, a 14B-parameter long-form audio-driven avatar generation framework that delivers fast and expressive inference, and introduces Endpoint-Anchored Distribution Distillation, a dedicated Anchor Score Estimator trained on the trajectory-endpoint distribution defined from a frozen pretrained 4-...
Hengyuan Zhang, Jing-Na Sun, Mei-Guang Jin et al.· arXiv.org· 1 citation
Results show that \method preserves stable appearance across prompt-conditioned segments and strong audio-visual synchronization under autoregressive generation, accelerating autoregressive inference without pipeline-specific retraining.
Qijun Gan, Chen-Wei Zhang, Mei-Guang Jin et al.· arXiv.org· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
