Research map / Large language models / Agents, code and math
TEACh: Task-Driven Embodied Agents That Chat
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 the multimodal Episodic Transformer instead of language-only or random policies | EDH follower policy · EDH validation and test splits | SR and GC | Helped |
| Use the multimodal Episodic Transformer instead of the vision-only baseline | EDH follower policy · EDH unseen splits | SR and GC | Helped |
| Initialize with pretrained ALFRED weights (ablation) | Episodic Transformer · EDH | SR and GC | No clear effect |
| Add history loss over actions in both history and target sequence (ablation) | Episodic Transformer · EDH | SR: 5.76 -> 7.4; 4.96 -> 4.31; 4.8 -> 6.01; 5.02 -> 5.68 | Mixed |
| Initialize with ALFRED base-annotation weights (ablation) | Episodic Transformer · EDH | SR: 5.76 -> 10.2; 4.96 -> 5.56; 4.8 -> 7.06; 5.02 -> 5.24 | Mixed |
| Initialize with ALFRED weights trained using synthetic instructions (ablation) | Episodic Transformer · EDH | SR: 5.76 -> 8.55; 4.96 -> 7.83; 4.8 -> 4.2; 5.02 -> 3.88 | Mixed |
| Predict the full follower action trajectory from dialogue history | Episodic Transformer · TfD | SR: 0.00 -> 1.02; 0.00 -> 0.48; 0.00 -> 0.51; 0.00 -> 0.17 | Helped |
| Use hand-crafted rule-based Commander and Follower policies | Two-Agent Task Completion · Simple tasks, including Make Coffee | Success rate: 54.55 | Mixed |
| Use hand-crafted rule-based Commander and Follower policies for compositional tasks | Two-Agent Task Completion · Boil Potato, Make Plate of Toast, Make Sandwich, Prepare Breakfast | Success rate: 0.00; 0.00; 0.00; 0.00 | Hurt |
Closest papers by meaning
- PARTNR: A Benchmark for Planning and Reasoning in Embodied Multi-agent Tasks (ICLR 2025 · similarity 0.87)
- OpenEQA: Embodied Question Answering in the Era of Foundation Models (CVPR 2024 · similarity 0.87)
- Implicit Intelligence - Evaluating Agents on What Users Don’t Say (ICML 2026 · similarity 0.86)
- EmbodiedBench: Comprehensive Benchmarking Multi-modal Large Language Models for Vision-Driven Embodied Agents (ICML 2025 · similarity 0.86)
- Infer Human’s Intentions Before Following Natural Language Instructions (AAAI 2025 · similarity 0.86)
- Bootstrapping Cognitive Agents with a Large Language Model (AAAI 2024 · similarity 0.86)
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.