mlsanityData to 14 Oct 2026

Lost in Translation? Vocabulary Alignment for Source-Free Adaptation in Open-Vocabulary Semantic Segmentation

arXiv 2509.15225 ↗First posted 18 Sep 2025cs.CVcited by 1

Not on the map. Why?

Abstract (arXiv)

We introduce VocAlign, a novel source-free domain adaptation framework specifically designed for VLMs in open-vocabulary semantic segmentation. Our method adopts a student-teacher paradigm enhanced with a vocabulary alignment strategy, which improves pseudo-label generation by incorporating additional class concepts. To ensure efficiency, we use Low-Rank Adaptation (LoRA) to fine-tune the model, preserving its original capabilities while minimizing computational overhead. In addition, we propose a Top-K class selection mechanism for the student model, which significantly reduces memory requirements while further improving adaptation performance. Our approach achieves a notable 6.11 mIoU improvement on the CityScapes dataset and demonstrates superior performance on zero-shot segmentation benchmarks, setting a new standard for source-free adaptation in the open-vocabulary setting.

Key facts. Cited by 1 paper in the map (data through 14 Oct 2026). Builds on 50 of its 61 references that are in the map. First posted on arXiv 18 Sep 2025.

Builds on

50 of 61 references.

References · 50

  1. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,744
  2. 2021LoRA: Low-Rank Adaptation of Large Language Modelscited by 14,380
  3. 2019Parameter-Efficient Transfer Learning for NLPcited by 4,236
  4. 2021Prefix-Tuning: Optimizing Continuous Prompts for Generationcited by 3,915
  5. 2021The Power of Scale for Parameter-Efficient Prompt Tuningcited by 3,795
  6. 2015Domain-Adversarial Training of Neural Networkscited by 3,044
  7. 2021Learning to Prompt for Vision-Language Modelscited by 2,281
  8. 2022Conditional Prompt Learning for Vision-Language Modelscited by 1,433
  9. 2015Learning Transferable Features with Deep Adaptation Networkscited by 1,329
  10. 2016Deep CORAL: Correlation Alignment for Deep Domain Adaptationcited by 1,200
  11. 2016Semantic Understanding of Scenes through the ADE20K Datasetcited by 1,069
  12. 2021CLIP-Adapter: Better Vision-Language Models with Feature Adapterscited by 1,026
38 more
  1. 2021Open-vocabulary Object Detection via Vision and Language Knowledge Distillationcited by 906
  2. 2016COCO-Stuff: Thing and Stuff Classes in Contextcited by 864
  3. 2020AdapterFusion: Non-Destructive Task Composition for Transfer Learningcited by 788
  4. 2022Language-driven Semantic Segmentationcited by 717
  5. 2018Deep Visual Domain Adaptation: A Surveycited by 635
  6. 2021Extract Free Dense Labels from CLIPcited by 583
  7. 2022Open-Vocabulary Semantic Segmentation with Mask-adapted CLIPcited by 565
  8. 2018Domain Adaptive Faster R-CNN for Object Detection in the Wildcited by 443
  9. 2021CLIP-NeRF: Text-and-Image Driven Manipulation of Neural Radiance Fieldscited by 370
  10. 2023LoraHub: Efficient Cross-Task Generalization via Dynamic LoRA Compositioncited by 364
  11. 2021DAFormer: Improving Network Architectures and Training Strategies for Domain-Adaptive Semantic Segmentationcited by 300
  12. 2023Side Adapter Network for Open-Vocabulary Semantic Segmentationcited by 300
  13. 2018Strong-Weak Distribution Alignment for Adaptive Object Detectioncited by 270
  14. 2016FCNs in the Wild: Pixel-level Adversarial and Constraint-based Adaptationcited by 243
  15. 2023CAT-Seg: Cost Aggregation for Open-Vocabulary Semantic Segmentationcited by 219
  16. 2022MIC: Masked Image Consistency for Context-Enhanced Domain Adaptationcited by 178
  17. 2022ZegCLIP: Towards Adapting CLIP for Zero-shot Semantic Segmentationcited by 178
  18. 2021Source-Free Domain Adaptation for Semantic Segmentationcited by 168
  19. 2022HRDA: Context-Aware High-Resolution Domain-Adaptive Semantic Segmentationcited by 164
  20. 2021Self-supervised Augmentation Consistency for Adapting Semantic Segmentationcited by 149
  21. 2017Convolutional neural network architecture for geometric matchingcited by 138
  22. 2019Domain Adaptation for Semantic Segmentation with Maximum Squares Losscited by 138
  23. 2021Cross-Domain Adaptive Teacher for Object Detectioncited by 126
  24. 2023CLIPSelf: Vision Transformer Distills Itself for Open-Vocabulary Dense Predictioncited by 118
  25. 2023Align Your Prompts: Test-Time Prompting with Distribution Alignment for Zero-Shot Generalizationcited by 101
  26. 2022Decorate the Newcomers: Visual Domain Prompt for Continual Test Time Adaptationcited by 98
  27. 2022Visual Prompt Tuning for Test-time Domain Adaptationcited by 97
  28. 2023SED: A Simple Encoder-Decoder for Open-Vocabulary Semantic Segmentationcited by 97
  29. 2022SePiCo: Semantic-Guided Pixel Contrast for Domain Adaptive Semantic Segmentationcited by 87
  30. 2021Generalize then Adapt: Source-Free Domain Adaptive Semantic Segmentationcited by 81
  31. 2021Domain Adaptive Semantic Segmentation with Self-Supervised Depth Estimationcited by 80
  32. 2023SILC: Improving Vision Language Pretraining with Self-Distillationcited by 66
  33. 2023Introducing Language Guidance in Prompt-based Continual Learningcited by 49
  34. 2022PiPa: Pixel- and Patch-wise Self-supervised Learning for Domain Adaptative Semantic Segmentationcited by 41
  35. 2022Non-Contrastive Learning Meets Language-Image Pre-Trainingcited by 25
  36. 2023APoLLo: Unified Adapter and Prompt Learning for Vision Language Modelscited by 19
  37. 2024Open-Set Domain Adaptation for Semantic Segmentationcited by 13
  38. 2024MICDrop: Masking Image and Depth Features via Complementary Dropout for Domain-Adaptive Semantic Segmentationcited by 13

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 1.

1 paper.

  1. 2026A Training-Free Guess What Vision Language Model from Snippets to Open-Vocabulary Object Detectioncited by 1

Similar, not cited

Close in wording, no citation either way

  1. 2024OpenDAS: Open-Vocabulary Domain Adaptation for 2D and 3D Segmentation
  2. 2024VLMs meet UDA: Boosting Transferability of Open Vocabulary Segmentation with Unsupervised Domain Adaptation
  3. 2024Training-Free Open-Vocabulary Segmentation with Offline Diffusion-Augmented Prototype Generation
  4. 2024Renovating Names in Open-Vocabulary Segmentation Benchmarks
  5. 2024Transferable and Principled Efficiency for Open-Vocabulary Segmentation
  6. 2026Towards Continual Test-Time Adaptation of Vision-Language Models in Open-Vocabulary Semantic Segmentation
  7. 2024Adapting Vision-Language Model with Fine-grained Semantics for Open-Vocabulary Segmentation
  8. 2025Test-Time Adaptation of Vision-Language Models for Open-Vocabulary Semantic Segmentation