Research map / Large language models / AI and society
Fix the Mind, Not the Move: Interpretable AI Assistance via Knowledge-Gap Localization
Ayano Hiranaka, Ya-Chuan Hsu, Stefanos Nikolaidis, Erdem Biyik, Daniel Seita
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 auxiliary mixup generalization loss (ablation) | SENSEI · Breakfast | System recall: 0.221 -> 0.762 | Improved |
| add auxiliary mixup generalization loss (ablation) | SENSEI · Overcooked | System F1: 0.698 -> 0.726 | Improved |
| add auxiliary mixup generalization loss (ablation) | SENSEI · Rover | System F1: 0.566 -> 0.561 | No clear effect |
| anchor training on expert instead of a random student (ablation) | SENSEI · Breakfast | System F1: 0.752 -> 0.709 | Hurt |
| anchor training on expert instead of a random student (ablation) | SENSEI · Overcooked | System F1: 0.708 -> 0.726 | No clear effect |
| anchor training on expert instead of a random student (ablation) | SENSEI · Rover | System F1: 0.584 -> 0.561 | Hurt |
| add expert trajectory input (ablation) | SENSEI · Breakfast | System F1: 0.728 -> 0.709 | Mixed |
| separate localization from editing instead of end-to-end decoder | SENSEI · Overcooked | System F1: 0.028 -> 0.726 | Improved |
| use trained localizer instead of random historical localization | SENSEI · Overcooked | Localization F1: 0.500 -> 0.726 | Improved |
| replace trained SENSEI with prompted LLMs | GPT-5.2 · Breakfast | System F1: 0.709 -> 0.126 | Hurt |
| give users SENSEI guidance after first attempt | Overcooked players · User study | plan-step IoU with expert: 0.736 -> 0.844 | Improved |
Closest papers by meaning
- When Models Know More Than They Can Explain: Quantifying Knowledge Transfer in Human-AI Collaboration (NeurIPS 2025 · similarity 0.86)
- Structure Enables Effective Self-Localization of Errors in LLMs (ICML 2026 · similarity 0.85)
- ThinkBot: Embodied Instruction Following with Thought Chain Reasoning (ICLR 2025 · similarity 0.85)
- HumanSense: From Multimodal Perception to Empathetic Context-Aware Responses Through Reasoning MLLMs (AAAI 2026 · similarity 0.84)
- RealWebAssist: A Benchmark for Long-Horizon Web Assistance with Real-World Users (AAAI 2026 · similarity 0.84)
- MetaMind: Modeling Human Social Thoughts with Metacognitive Multi-Agent Systems (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.
