Research map / Large language models / AI and society
Effective Human-AI Teams via Learned Natural Language Rules and Onboarding
Hussein Mozannar, Jimin J Lee, Dennis Wei, Prasanna Sattigeri, Subhro Das, David Sontag
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 |
|---|---|---|---|
| onboarding with IntegrAI regions vs Human+AI without onboarding | Human-AI team · BDD user study | participant accuracy (%): 77.2 -> 82.6 | Improved |
| onboarding with DOMINO-based baseline regions vs no onboarding | Human-AI team · BDD user study | participant accuracy (%): 77.2 -> 80.4 | No clear effect |
| add AI-integration recommendations to onboarding | Onboard(ours) · BDD user study | participant accuracy (%): 82.6 -> 79.9 | No clear effect |
| add AI-integration recommendations to onboarding | Onboard(ours) · BDD user study | time per example (s): 5.936 -> 7.622 | Hurt |
| onboarding vs no onboarding | Human+AI · MMLU user study | participant accuracy (%): 75.0 -> 74.4 | No clear effect |
| show only AI-integration recommendations | Human+AI · MMLU user study | participant accuracy (%): 75.0 -> 69.8 | Hurt |
| onboarding vs no onboarding | Human+AI · MMLU user study | time per example (s): 23.623 -> 20.977 | Improved |
| IntegrAI regions vs DOMINO | region discovery · BDD | test error (%): 18.9 -> 17.8 | Improved |
| IntegrAI regions vs DOMINO | region discovery · MMLU | test error (%): 48.1 -> 45.3 | Improved |
| IntegrAI regions vs DOMINO | region discovery · DynaSent | test error (%): 20.0 -> 20.2 | No clear effect |
| include examples outside the region in description (S- not empty, m=5) (ablation) | IntegrAI-Describe · MS-COCO | METEOR: 24.3 -> 26.1 | Improved |
| more refinement iterations m=0 to m=5 (ablation) | IntegrAI-Describe · MS-COCO | SPICE: 43.8 -> 45.2 | Improved |
Closest papers by meaning
- When Models Know More Than They Can Explain: Quantifying Knowledge Transfer in Human-AI Collaboration (NeurIPS 2025 · similarity 0.85)
- Improving Knowledge Extraction from LLMs for Task Learning through Agent Analysis (AAAI 2024 · similarity 0.84)
- Generalizing Verifiable Instruction Following (NeurIPS 2025 · similarity 0.84)
- Mars: Situated Inductive Reasoning in an Open-World Environment (NeurIPS 2024 · similarity 0.84)
- MLRC-Bench: Can Language Agents Solve Machine Learning Research Challenges? (NeurIPS 2025 · similarity 0.84)
- Synergistic Multi-Agent Framework with Trajectory Learning for Knowledge-Intensive Tasks (AAAI 2025 · 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.
