Live Assistant: Learning Whether, When, and Whom to Assist in Real-World Live Social Streams

2026-09-22

AI Summary

该研究通过构建liveassistant框架，利用自回归策略处理直播中的视听内容和观众互动，以决定是否、何时及向谁提供帮助。

Abstract

Livestreams are long-lasting interactive environments where audiovisual content, viewer activity, host behavior, and platform signals evolve together, creating assistance needs that emerge from the stream itself. We introduce \liveassistant, a framework for mixed-initiative, role-conditioned assistance that formulates livestream interaction as four coupled decisions: \textbf{whether to act, when to act, whom to address, and what to communicate}. At each 10-second interval, one autoregressive policy consumes native audio and video with synchronized comments, gifts, viewer dynamics, and room metadata, then selects \textsc{OBS}, \textsc{MEM}, or \textsc{ANS}. \textsc{OBS} remains silent, \textsc{MEM} records a private semantic update, and \textsc{ANS} specifies a recipient, task, and grounded message. To support this task, we build a trajectory engine that reconstructs real livestream sessions into structured causal supervision, yielding over 320 hours of optimization trajectories and a human-reviewed benchmark of 275 clips and 13,812 decision intervals. We train the policy with Marker-Aware Multiturn Supervised Fine-Tuning (MA-MSFT), which strengthens sparse structured decisions, followed by Streaming Multiturn GSPO (SM-GSPO), which optimizes self-generated trajectories with turn- and trajectory-level credit. On the held-out benchmark, \liveassistant reaches 71.14 state accuracy, 72.67 recipient accuracy, and 58.41 task accuracy, with consistent gains over representative streaming and general multimodal baselines. Together, the formulation, benchmark, and training framework establish livestream assistance as selective participation in a shared social stream.

Problem

• Livestreams
• Assistance
• Interactive Environments
• Viewer Activity
• Host Behavior

Innovation

• mixed-initiative
• role-conditioned assistance
• autoregressive policy
• trajectory engine
• Marker-Aware Multiturn Supervised Fine-Tuning