Research map / Large language models / AI and society
Proact-VL: A Proactive VideoLLM for Real-Time AI Companions
Weicai Yan, Yuhong Dai, Qi Ran, Haodong Li, Wang Lin, Tao Jin, Xing Xie, Hao Liao, Jianxun Lian
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add regression loss L_reg (ablation) | Proact-VL response head · Live Gaming Benchmark loss ablation | F1: 11.03 -> 60.08 | Improved |
| add regression loss L_reg (ablation) | Proact-VL response head · Live Gaming Benchmark loss ablation | TimeDiff: 18.50 -> 3.41 | Improved |
| add classification loss L_cls (ablation) | Proact-VL response head · Live Gaming Benchmark loss ablation | F1: 47.39 -> 60.08 | Improved |
| add classification loss L_cls (ablation) | Proact-VL response head · Live Gaming Benchmark loss ablation | TimeDiff: 8.28 -> 3.41 | Improved |
| add both losses vs L_cls only (ablation) | Proact-VL · Live Gaming Benchmark loss ablation | CC: 45.54 -> 50.91 | Improved |
| add classification loss L_cls (ablation) | Proact-VL · Live Gaming Benchmark loss ablation | recall: 67.02 -> 62.41 | Hurt |
| use Proact-VL instead of LiveCC-7B-Base | real-time commentary · Live Gaming Benchmark overall | CC: 38.88 -> 49.23 | Improved |
| use Proact-VL instead of LiveCC-7B-Instruct | real-time commentary · Live Gaming Benchmark overall | F1: 56.05 -> 64.87 | Improved |
| use Proact-VL instead of GPT-4o | guidance · Live Gaming Benchmark guidance | CC: 50.00 -> 42.60 | Hurt |
| use Proact-VL instead of Gemini 2.5 Pro | proactive timing · Live Gaming Benchmark overall | TimeDiff: 2.59 -> 1.71 | Improved |
| use Proact-VL fine-tuned from Qwen2-VL instead of base Qwen2-VL | Qwen2-VL · multiple benchmarks overall | CC: 31.31 -> 55.19 | Improved |
| use Proact-VL from Qwen3-VL instead of base Qwen3-VL | Qwen3-VL · multiple benchmarks overall | F1: 35.79 -> 60.52 | Improved |
Closest papers by meaning
- IPIBench: Evaluating Interactive Proactive Intelligence of MLLMs under Continuous Streams (NeurIPS 2026 · similarity 0.87)
- Proactive Agent: Shifting LLM Agents from Reactive Responses to Active Assistance (ICLR 2025 · similarity 0.86)
- VirtualEnv: A Platform for Embodied AI Research (AAAI 2026 · similarity 0.86)
- ProAct: A Benchmark and Multimodal Framework for Structure-Aware Proactive Response (ICML 2026 · similarity 0.86)
- Learning Interactive Real-World Simulators (ICLR 2024 · similarity 0.86)
- Steve-Eye: Equipping LLM-based Embodied Agents with Visual Perception in Open Worlds (ICLR 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
