Research map / Adversarial robustness and security / Black-box and transfer attacks
Shedding Light on VLN Robustness: A Black-box Framework for Indoor Lighting-based Adversarial Attack
Chenyang Li, Wenbing Tang, Yihao Huang, Sinong Simon Zhan, Ming Hu, Xiaojun Jia, Yang Liu
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 dynamic lighting switches (DILA) on top of static attack (ablation) | ILA on SPOC · CHORES ObjectNav | ASR (%): 60.38 -> 96.23 | Improved |
| add DILA on top of SILA (ablation) | ILA on FLaRe · CHORES Fetch | ASR (%): 68.81 -> 93.58 | Improved |
| add DILA on top of SILA (ablation) | ILA on FLaRe · CHORES ObjectNav | ASR (%): 47.27 -> 52.73 | Improved |
| add DILA on top of SILA (ablation) | ILA on SPOC · CHORES Fetch | ASR (%): 100.00 -> 100.00 | No clear effect |
| use optimized global lighting (SILA) instead of random intensity | ILA on SPOC · CHORES ObjectNav | ASR (%): 23.23 -> 60.38 | Improved |
| use SILA lighting instead of texture-based genetic attack | ILA · CHORES ObjectNav (SPOC) | ASR (%): 54.90 -> 60.38 | Improved |
| use SILA lighting instead of texture attack on episode length | ILA on SPOC · CHORES ObjectNav | episode length: 137.04 (Texture-GA) vs 131.07 (SILA) | Mixed |
| use timestep-increasing weights in SILA loss vs unweighted (ablation) | SILA on FLaRe · CHORES ObjectNav | ASR (%): 44.44 -> 47.27 | Improved |
| use increasing weights instead of decreasing weights (ablation) | SILA on SPOC · CHORES ObjectNav | ASR (%): 57.55 -> 60.38 | Improved |
| use one-step lookahead trigger instead of random switch timing (ablation) | DILA on SPOC · CHORES ObjectNav | ASR (%): 85.85 -> 96.23 | Improved |
| use moderate step size 0.05 vs 0.01 (ablation) | SILA on SPOC · CHORES ObjectNav | ASR (%): 50.94 -> 60.38 | Improved |
| use bounds [0.0, 1.5] vs [0.0, 2.0] (ablation) | SILA on SPOC · CHORES ObjectNav | ASR (%): 56.73 -> 60.38 | Improved |
Closest papers by meaning
- BadVLA: Towards Backdoor Attacks on Vision-Language-Action Models via Objective-Decoupled Optimization (NeurIPS 2025 · similarity 0.84)
- Everyday Object Meets Vision-and-Language Navigation Agent via Backdoor (NeurIPS 2024 · similarity 0.83)
- AdvEDM: Fine-grained Adversarial Attack against VLM-based Embodied Agents (NeurIPS 2025 · similarity 0.83)
- SceneTAP: Scene-Coherent Typographic Adversarial Planner against Vision-Language Models in Real-World Environments (CVPR 2025 · similarity 0.83)
- Steering Away from Harm: An Adaptive Approach to Defending Vision Language Model Against Jailbreaks (CVPR 2025 · similarity 0.82)
- ScribbleLight: Single Image Indoor Relighting with Scribbles (CVPR 2025 · similarity 0.82)
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.
