Research map / Vision-language and multimodal / Robot action and navigation
Path-level Hindsight Instructions for Semantic Exploration in Vision-Language Navigation
Sungjune Kim, Sangpil Kim, Honglak Lee
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 Phi-Nav hindsight instructions (DAgger) | CMA-D-P-A · R2R-CE val unseen | SR: 32.10 -> 34.08 | Improved |
| add Phi-Nav hindsight instructions (DAgger) | CMA-D · R2R-CE val unseen | SR: 26.64 -> 27.62 | Improved |
| add Phi-Nav (scheduled sampling) | ETPNav · R2R-CE val unseen | SR: 57.21 -> 62.58 | Improved |
| add Phi-Nav (scheduled sampling) | ETPNav · R2R-CE val unseen | SPL: 49.15 -> 52.74 | Improved |
| add Phi-Nav | ETPNav · RxR-CE val unseen | SR: 54.79 -> 55.83 | Improved |
| add Phi-Nav | ETPNav · RxR-CE val unseen | NE: 5.64 -> 5.67 | Hurt |
| add alignment-weighted hindsight loss (TIAW) instead of fixed lambda=1.0 (ablation) | Phi-Nav on ETPNav · R2R-CE val unseen | SPL: 46.32 -> 51.32 | Improved |
| use fixed lambda=0.1 instead of lambda=1.0 (ablation) | Phi-Nav on ETPNav · R2R-CE val unseen | SPL: 46.32 -> 48.66 | Improved |
| add landmark-aware scoring to TIAW (ablation) | Phi-Nav on ETPNav · R2R-CE val unseen | SR: 61.18 -> 62.58 | Improved |
| use Qwen2.5-VL-7B instead of 3B speaker (ablation) | Phi-Nav on ETPNav · R2R-CE val unseen | SPL: 47.32 -> 52.74 | Improved |
| use Qwen2.5-VL-7B instead of Qwen3-VL-8B speaker (ablation) | Phi-Nav on ETPNav · R2R-CE val unseen | SR: 60.05 -> 62.58 | Improved |
| add expert-in-context example to zero-shot speaker (ablation) | Phi-Nav on ETPNav · R2R-CE val unseen | SPL: 50.55 -> 52.74 | 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.
- ETPNav — ETPNav: Evolving Topological Planning for Vision-Language Navigation in Continuous Environments (TPAMI 2024) · better on 6 of 6 shared results
- CMA-D — CMAD: Correlation-Aware and Modalities-Aware Distillation for Multimodal Sentiment Analysis with Missing Modalities (ICCV 2025) · better on 3 of 3 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
- AgentVLN: Towards Agentic Vision-and-Language Navigation (ECCV 2026 · similarity 0.91)
- Progress-Think: Semantic Progress Reasoning for Vision-Language Navigation (CVPR 2026 · similarity 0.90)
- SeqWalker: Sequential-Horizon Vision-and-Language Navigation with Hierarchical Planning (AAAI 2026 · similarity 0.90)
- Instruction Decomposition and Action Alignment for Vision-Language Navigation (ICML 2026 · similarity 0.90)
- NavGPT: Explicit Reasoning in Vision-and-Language Navigation with Large Language Models (AAAI 2024 · similarity 0.90)
- Vision-Language Navigation with Energy-Based Policy (NeurIPS 2024 · 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.
