Research map / Detection and segmentation / Open-set and few-shot detection
Multiple Thinking Achieving Meta-Ability Decoupling for Object Navigation
Ronghao Dang, Lu Chen, Liuyi Wang, Zongtao He, Chengju Liu, Qijun Chen
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 search thinking (object association via DOA-style attention) (ablation) | MT · AI2-Thor ALL | SR: 43.48 -> 75.94 | Improved |
| add navigation thinking (ablation) | MT · AI2-Thor ALL | SR: 75.94 -> 79.62 | Improved |
| add exploration thinking (ablation) | MT · AI2-Thor ALL | SR: 79.62 -> 81.97 | Improved |
| add obstacle thinking (ablation) | MT · AI2-Thor ALL | SR: 81.97 -> 83.14 | Improved |
| add multiple thinking collaboration (MTC) module (ablation) | MT · AI2-Thor ALL | SR: 82.26 -> 83.14 | Improved |
| add meta-ability reward R_MA (during first 0.2M episodes) (ablation) | MT · AI2-Thor ALL | SR and SPL: SR 81.96 -> 83.14; SPL 49.54 -> 50.23 | Improved |
| keep meta-ability reward throughout all training episodes (ablation) | MT · AI2-Thor ALL | SR: 83.14 -> 81.31 | Hurt |
| MT decoupled meta-ability model vs DOA | MT · AI2-Thor ALL | SR and SPL: SR 74.32 -> 83.14; SPL 40.27 -> 50.23 | Improved |
| MT decoupled meta-ability model vs DOA | MT · RoboTHOR ALL | SR and SPL: SR 36.22 -> 42.17; SPL 22.12 -> 29.14 | Improved |
| zero-shot MT-ZS with GloVe target similarity instead of one-hot | MT-ZS · Zero-shot AI2-Thor unseen 18/4 | SR: 57.32 (ZSON) -> 68.61 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- PONI — PONI: Potential Functions for ObjectGoal Navigation With Interaction-Free Learning (CVPR 2022) · better on 2 of 2 shared results
- VTNet — VTNet: Visual Transformer Network for Object Goal Navigation (ICLR 2021) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Search for or Navigate to? Dual Adaptive Thinking for Object Navigation (ICCV 2023 · similarity 0.88)
- Run, Ruminate, and Regulate: A Dual-process Thinking System for Vision-and-Language Navigation (AAAI 2026 · similarity 0.87)
- Aux-Think: Exploring Reasoning Strategies for Data-Efficient Vision-Language Navigation (NeurIPS 2025 · similarity 0.85)
- Enhancing Multi-Robot Semantic Navigation Through Multimodal Chain-of-Thought Score Collaboration (AAAI 2025 · similarity 0.84)
- Scene Map-based Prompt Tuning for Navigation Instruction Generation (CVPR 2025 · similarity 0.84)
- FantasyVLN: Unified Multimodal Chain-of-Thought Reasoning for Vision-and-Language Navigation (CVPR 2026 · 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.
