Research map / Vision-language and multimodal / CLIP and zero-shot
VLN-PETL: Parameter-Efficient Transfer Learning for Vision-and-Language Navigation
Yanyuan Qiao, Zheng Yu, 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 |
|---|---|---|---|
| Add CIB (cross-modal interaction booster) to head-only tuning (ablation) | HAMT head tuning · REVERIE Val Unseen | SPL: 15.19 -> 21.62 | Improved |
| Add LEA (language encoder adapter) to head tuning (ablation) | HAMT head tuning · REVERIE Val Unseen | SPL: 15.19 -> 18.21 | Improved |
| Add LoRA to LEA+HIB+CIB (ablation) | VLN-PETL · REVERIE Val Unseen | SPL: 26.51 -> 27.67 | Improved |
| Use CIB instead of cross-modal adapter (CEA) (ablation) | VLN-PETL · REVERIE Val Unseen | SPL: 20.77 -> 21.62 | Improved |
| Use HIB instead of history adapter (HEA) (ablation) | VLN-PETL · REVERIE Val Unseen | SPL: SPL 18.98 -> 18.49; RGSPL 8.39 -> 9.32 | Mixed |
| Use learnable gate alpha instead of fixed 0.5 (ablation) | HIB and CIB gates · REVERIE Val Unseen | RGSPL: 14.50 -> 15.96 | Improved |
| Set gate temperature T to 0.1 versus 1 (ablation) | HIB and CIB gates · REVERIE Val Unseen | SPL: 27.67 vs 27.81 | No clear effect |
| Use BitFit or Prompt Tuning instead of Adapter/LoRA for VLN | HAMT · RxR Val Unseen | nDTW: BitFit 50.47, Prompt Tuning 44.94 vs Adapter 63.56 | Hurt |
| Use VLN-PETL (2.82% params) instead of full fine-tuning | HAMT · R2R Val Unseen | SR: 64.24 -> 65.47 | Improved |
| Use VLN-PETL instead of full fine-tuning on REVERIE | HAMT · REVERIE Val Unseen | SPL: 28.92 -> 27.67 | Hurt |
Later papers that compared with this method
2 later accepted papers report VLN-PETL as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- EvolveNav: Empowering LLM-Based Vision-Language Navigation via Self-Improving Embodied Reasoning (TPAMI 2026 · better on 1 of 1 shared result)
- Vision-and-Language Navigation via Causal Learning (CVPR 2024 · 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
- Parameter and Computation Efficient Transfer Learning for Vision-Language Pre-trained Models (NeurIPS 2023 · similarity 0.89)
- Towards Learning a Generic Agent for Vision-and-Language Navigation via Pre-Training (CVPR 2020 · similarity 0.89)
- Harnessing Input-Adaptive Inference for Efficient VLN (ICCV 2025 · similarity 0.89)
- Towards Long-Horizon Vision-Language Navigation: Platform, Benchmark and Method (CVPR 2025 · similarity 0.88)
- Vision-Language Model Selection and Reuse for Downstream Adaptation (ICML 2025 · similarity 0.87)
- PriorVLA: Prior-Preserving Adaptation for Vision-Language-Action Models (NeurIPS 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.
