Research map / Label-efficient and robust learning / Recognition and long-tail
First Contact: Unsupervised Human-Machine Co-Adaptation via Mutual Information Maximization
Siddharth Reddy, Sergey Levine, Anca Dragan
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 |
|---|---|---|---|
| replace random interfaces with MIMI-optimized interface | cursor control interface · user study, 'The system improved over time' | Likert rating: 3.33 -> 5.83 | Improved |
| replace random interfaces with MIMI-optimized interface | cursor control interface · user study, 'I felt in control' | Likert rating: 4.08 -> 4.92 | No clear effect |
| replace random interfaces with MIMI-optimized interface | cursor control interface · user study, 'System responded as I expected' | Likert rating: 4.17 -> 4.75 | No clear effect |
| replace random interfaces with MIMI-optimized interface | cursor control interface · user study, 'I improved at using the system' | Likert rating: 5.58 -> 6.08 | No clear effect |
| set time offset Delta to 1 in generalized MI objective (ablation) | MIMI · SAvDRL Lunar Lander | Spearman rho with ground-truth reward | Hurt |
| raise time offset Delta to max episode length 10^3 (ablation) | MIMI · SAvDRL Lunar Lander | Spearman rho with ground-truth reward | Improved |
| use MI score to rank interfaces without ground-truth reward | MIMI · offline evaluation, X2T | Spearman rho: 0.45 | Improved |
| use MI score to rank interfaces across 5 domains | MIMI · offline evaluation, 4 of 5 domains | Spearman rho: 0.43 | Improved |
| optimize MI reward with human-in-the-loop RL instead of random interface | MIMI · 2D cursor control | ground-truth reward (negative distance): 0.87 | Improved |
| learn interface online for 200 episodes | MIMI · Lunar Lander with hand gestures | landing success | Improved |
Closest papers by meaning
- Behavioral Exploration: Learning to Explore via In-Context Adaptation (ICML 2025 · similarity 0.85)
- Towards Better Robot Learners: Leveraging Implicit and Explicit Human Feedback Together in Human Robot Interactions (AAAI 2025 · similarity 0.85)
- TokenHSI: Unified Synthesis of Physical Human-Scene Interactions through Task Tokenization (CVPR 2025 · similarity 0.84)
- ManipTrans: Efficient Dexterous Bimanual Manipulation Transfer via Residual Learning (CVPR 2025 · similarity 0.84)
- PEAC: Unsupervised Pre-training for Cross-Embodiment Reinforcement Learning (NeurIPS 2024 · similarity 0.84)
- COOPERA: Continual Open-Ended Human-Robot Assistance (NeurIPS 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.
