Research map / Vision-language and multimodal / Robot action and navigation
NavGPT-2: Unleashing Navigational Reasoning Capability for Large Vision-Language Models
Gengze Zhou, Yicong Hong, Zun Wang, Xin Eric Wang, Qi Wu
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 LLM latents (frozen VLM + Q-former) as visual-linguistic representation instead of LXMERT-style transformer | DUET global-action policy · R2R Test Unseen | SR: 69 -> 71 (FlanT5-XL w/ PREVALENT) | Improved |
| use FlanT5-XXL w/ PREVALENT NavGPT-2 over NaviLLM | NavGPT-2 · R2R Test Unseen | SR: 68 -> 72 | Improved |
| remove the navigation-specific policy network (ablation) | NavGPT-2 FlanT5-XL · R2R Val Unseen | SR: 67.52 -> 21.46 | Hurt |
| pretrain Q-former on navigational reasoning data (ablation) | NavGPT-2 FlanT5-XL · R2R Val Unseen | SR: 66.75 -> 67.52 | Improved |
| scale LLM from FlanT5-XL to FlanT5-XXL (ablation) | NavGPT-2 · R2R Val Unseen | SR: 67.52 -> 71.31 | Improved |
| use decoder-only Vicuna instead of encoder-decoder FlanT5 (ablation) | NavGPT-2 · R2R Val Unseen | SR: 71.31 (FlanT5-XXL) -> 53.77 (Vicuna-7B), 48.28 (Vicuna-13B) | Hurt |
| scale Vicuna from 7B to 13B (ablation) | NavGPT-2 · R2R Val Unseen | SR: 53.77 -> 48.28 | Hurt |
| use NavGPT-2 instead of DUET with limited training data (10%) | NavGPT-2 FlanT5-XL · R2R Val Unseen (10% train) | SR: 48.19 -> 52.02 | Improved |
| zero-shot transfer to unseen environments (HM3D) using LLM-space visual features | NavGPT-2 FlanT5-XXL · HM3D zero-shot | SR: 25.60 -> 47.20 | Improved |
| zero-shot transfer to RxR English | NavGPT-2 FlanT5-XXL · RxR-EN zero-shot | SR: 25.07 -> 28.75 | Improved |
| swap in a stronger vision encoder for DUET (ablation) | DUET global branch · R2R Val Unseen | SR | No clear effect |
| remove local branch from DUET (ablation) | DUET · R2R Test Unseen | SR: 69 -> 67 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- DUET — DUET: Cross-Modal Semantic Grounding for Contrastive Zero-Shot Learning (AAAI 2023) · better on 7 of 7 shared results
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
- NavGPT: Explicit Reasoning in Vision-and-Language Navigation with Large Language Models (AAAI 2024 · similarity 0.96)
- Run, Ruminate, and Regulate: A Dual-process Thinking System for Vision-and-Language Navigation (AAAI 2026 · similarity 0.91)
- Aux-Think: Exploring Reasoning Strategies for Data-Efficient Vision-Language Navigation (NeurIPS 2025 · similarity 0.90)
- FantasyVLN: Unified Multimodal Chain-of-Thought Reasoning for Vision-and-Language Navigation (CVPR 2026 · similarity 0.90)
- NavBench: Probing Multimodal Large Language Models for Embodied Navigation (NeurIPS 2025 · similarity 0.90)
- Insight-V: Exploring Long-Chain Visual Reasoning with Multimodal Large Language Models (CVPR 2025 · similarity 0.89)
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.
