Research map / Label-efficient and robust learning / Test-time adaptation
Test-Time Mixture of World Models for Embodied Agents in Dynamic Environments
Jinwoo Jang, Minjong Yoo, Sihyung Yoon, Honguk Woo
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 |
|---|---|---|---|
| use multi-granular (object plus scene) routing instead of object-only (ablation) | TMoW · VirtualHome (unseen, zero-shot) | success rate (%): 65.25 -> 80.74 | Improved |
| use multi-granular routing instead of scene-only (ablation) | TMoW · VirtualHome (unseen, zero-shot) | success rate (%): 8.74 -> 80.74 | Improved |
| add test-time prototype refinement (ablation) | TMoW · VirtualHome (unseen, zero-shot) | success rate (%): 73.30 -> 80.74 | Improved |
| add test-time prototype refinement (ablation) | TMoW · VirtualHome (unseen, zero-shot) | pending steps: 14.85 -> 13.12 | Improved |
| distilled mixture augmentation instead of training new model from scratch (ablation) | TMoW · VirtualHome (unseen, 1-shot) | success rate (%): 59.84 -> 81.56 | Improved |
| use top-3 routing instead of top-1 (ablation) | TMoW · VirtualHome (unseen, zero-shot) | success rate (%): 65.43 -> 80.16 | Improved |
| increase top-K from 3 to 5 (ablation) | TMoW · VirtualHome (unseen, zero-shot) | success rate (%): 80.16 -> 77.52 | Hurt |
| increase top-K from 3 to 7 (ablation) | TMoW · VirtualHome (unseen, zero-shot) | success rate (%): 80.16 -> 66.01 | Hurt |
| TMoW vs SayCanPay in zero-shot unseen domains | TMoW · VirtualHome (unseen) | success rate (%): 49.53 -> 80.16 | Improved |
| TMoW vs SayCanPay on RLBench unseen domains | TMoW · RLBench (unseen) | success rate (%): 38.54 -> 62.75 | Improved |
| TMoW vs SayCanPay in 5-shot expansion | TMoW · VirtualHome (unseen, 5-shot) | success rate (%): 58.88 -> 83.61 | Improved |
Closest papers by meaning
- World Model Implanting for Test-time Adaptation of Embodied Agents (ICML 2025 · similarity 0.91)
- ModularAgent: A Task-Aware Modular Framework for Joint Optimization of Multimodal Large Language Models and World Models (CVPR 2026 · similarity 0.89)
- COMBO: Compositional World Models for Embodied Multi-Agent Cooperation (ICLR 2025 · similarity 0.88)
- EmbodiedBench: Comprehensive Benchmarking Multi-modal Large Language Models for Vision-Driven Embodied Agents (ICML 2025 · similarity 0.88)
- Embodied Multi-Modal Agent trained by an LLM from a Parallel TextWorld (CVPR 2024 · similarity 0.88)
- From Assumptions to Actions: Turning LLM Reasoning into Uncertainty-Aware Planning for Embodied Agents (ICLR 2026 · similarity 0.87)
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.
