Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Watch-And-Help: A Challenge for Social Perception and Human-AI Collaboration
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use only the graph representation of the last observation instead of the full demonstration (ablation) | Goal inference model · test-1 | recall: 0.96 -> 0.75 | Hurt |
| Use Alice's actions as input instead of the symbolic state demonstration (ablation) | Goal inference model · test-1 | precision: 0.85 -> 0.99 | Helped |
| Use Alice's actions as input instead of the symbolic state demonstration (ablation) | Goal inference model · test-1 | recall: 0.96 -> 0.99 | Helped |
| Generalize from single-activity goals to multi-activity goals | Goal inference model · test-2 | precision: 0.85 -> 0.68 | Hurt |
| Generalize from single-activity goals to multi-activity goals | Goal inference model · test-2 | recall: 0.96 -> 0.64 | Hurt |
| Use a random inferred goal instead of a goal inferred from Alice's demonstration (ablation) | HP · Help stage, overall test set | cumulative reward | Hurt |
| Use a hierarchical RL low-level policy rather than a regression planner for subgoal execution (ablation) | Hierarchical helping agent · Help stage | helping performance | Hurt |
| Pair the human-like agent with a random-goal HP agent rather than having it work alone | Human-like agent · Help stage, overall test set | success rate | No clear effect |
| Pair real humans with a random-goal HP agent instead of the human-like agent | HP with random goal inference · Human-AI collaboration | success rate | Helped |
| Use the human-like agent rather than real humans to execute the same tasks alone | Single-agent task execution · Human experiment | number of steps | No clear effect |
Closest papers by meaning
- Constrained Human-AI Cooperation: An Inclusive Embodied Social Intelligence Challenge (NeurIPS 2024 · similarity 0.86)
- PARTNR: A Benchmark for Planning and Reasoning in Embodied Multi-agent Tasks (ICLR 2025 · similarity 0.85)
- RealWebAssist: A Benchmark for Long-Horizon Web Assistance with Real-World Users (AAAI 2026 · similarity 0.85)
- Designs for Enabling Collaboration in Human-Machine Teaming via Interactive and Explainable Systems (NeurIPS 2024 · similarity 0.85)
- MetaMind: Modeling Human Social Thoughts with Metacognitive Multi-Agent Systems (NeurIPS 2025 · similarity 0.84)
- SOTOPIA: Interactive Evaluation for Social Intelligence in Language Agents (ICLR 2024 · similarity 0.84)
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.