Research map / Multi-agent systems and games / Human-AI and goal-conditioned
CoMind: Understanding Collaborative Human Activity from Multiple Minds and Views
Alexey Gavryushin, Dingxi Zhang, Zhao Huang, Alexandros Delitzas, Jiaqi Chen, Ben Ellis, Cedric Zöllner, Manthan Patel, Manuel Kaufmann, Marc Pollefeys, Xi Wang
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 |
|---|---|---|---|
| fine-tune on CoMind (LoRA) | Qwen3-VL 8B Instruct · Object Interaction Anticipation | Action Verb accuracy: 0.0873 -> 0.6429 | Improved |
| fine-tune on CoMind (LoRA) | Qwen3-VL 32B Instruct · Object Interaction Anticipation | Action Verb accuracy: 0.1380 -> 0.6508 | Improved |
| fine-tune on CoMind | Qwen3-VL 32B Instruct · Joint Attention Estimation | L-IoU@0.5: 0.0001 -> 0.1562 | Improved |
| fine-tune on CoMind | Qwen3-VL 8B Instruct · Joint Attention Estimation | Cue Type accuracy: 0.0074 -> 0.4775 | Improved |
| fine-tune on CoMind | Qwen3-VL 32B Instruct · Handover Prediction | Initiator accuracy: 0.6143 -> 0.7286 | Improved |
| fine-tune on CoMind | Qwen3-VL 8B/32B · Handover Prediction | TTH accuracy: TTH 0.0524 -> 0.1143 (8B), 0.0381 -> 0.0905 (32B) | Mixed |
| provide transcripts of the last 10 s as input (ablation) | Gemini 3 Flash · Object Interaction Anticipation | Cue Type accuracy: 0.2954 -> 0.5404 | Improved |
| provide transcripts as input (ablation) | Qwen3-VL 8B Instruct · Object Interaction Anticipation | IoU@0.5: 0.0909 -> 0.0024 | Hurt |
| add context frames to input (ablation) | Gemini 3 Flash · Object Interaction Anticipation | Action Verb accuracy: 0.0623 -> 0.0781 | Improved |
| zero-shot proprietary VLMs on spatial grounding | Claude/GPT models · Joint Attention Estimation | L-IoU@0.5: 0.0044 vs 0.1772 random | Hurt |
Closest papers by meaning
- SoMi-ToM: Evaluating Multi-Perspective Theory of Mind in Embodied Social Interactions (NeurIPS 2025 · similarity 0.88)
- IndEgo: A Dataset of Industrial Scenarios and Collaborative Work for Egocentric Assistants (NeurIPS 2025 · similarity 0.86)
- Learning from Observer Gaze: Zero-Shot Attention Prediction Oriented by Human-Object Interaction Recognition (CVPR 2024 · similarity 0.86)
- Beyond Single Stationary Policies: Meta-Task Players as Naturally Superior Collaborators (NeurIPS 2024 · similarity 0.85)
- LifeEval: A Multimodal Benchmark for Assistive AI in Egocentric Daily Life Tasks (CVPR 2026 · similarity 0.85)
- HumanSense: From Multimodal Perception to Empathetic Context-Aware Responses Through Reasoning MLLMs (AAAI 2026 · similarity 0.85)
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.
