mlsanityData to 14 Oct 2026

When Semantics Saturate or Emerge: Adaptation-Conditional Semantic Utility in Source-Free Cross-Domain Few-Shot Learning

arXiv 2608.06673 ↗First posted 7 Aug 2026cs.CVcited by 0

Not on the map. Why?

Abstract (arXiv)

Language descriptions in source-free cross-domain few-shot learning (SF-CDFSL) are often selected according to zero-shot accuracy obtained with a frozen vision--language model. This paper asks whether that ranking remains valid after target-domain visual adaptation. Under a strictly paired protocol, we compare a generic class-name template with fixed detailed class descriptions before and after visual Low-Rank Adaptation (LoRA) on EuroSAT, CropDisease, ISIC, and ChestX. Let deltazero and deltalora denote the Detailed-minus-Base accuracy before and after adaptation, respectively. Two recurring regimes emerge. In semantic saturation, deltazero>0 but 0<deltaloralldeltazero: on EuroSAT and CropDisease, initial gains of 8.13--21.54 percentage points contract to 0.69--2.96 points after LoRA. In semantic emergence, deltazero≤0 but deltalora>0: on ISIC and ChestX, detailed descriptions become more useful only after the visual representation is updated. Training trajectories and sample-level decomposition show that saturation is driven mainly by Base-LoRA recovering errors already solved by detailed semantics, whereas emergence is associated with prediction turnover and newly formed Detailed-only correct decisions. Fixed-point-free shuffled-semantic controls, a second CLIP backbone, and multiple random seeds support the broad pattern while identifying ChestX 1-shot as a weak boundary case. These findings establish that zero-shot prompt quality is an incomplete proxy for adaptation-anchor quality and motivate evaluating language on both sides of the adaptation boundary.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 41 of its 46 references that are in the map. First posted on arXiv 7 Aug 2026.

Builds on

41 of 46 references.

References · 41

  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. 2017Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networkscited by 4,897
  4. 2017Prototypical Networks for Few-shot Learningcited by 3,015
  5. 2021Learning to Prompt for Vision-Language Modelscited by 2,281
  6. 2016Matching Networks for One Shot Learningcited by 2,255
  7. 2017EuroSAT: A Novel Dataset and Deep Learning Benchmark for Land Use and Land Cover Classificationcited by 1,496
  8. 2022Conditional Prompt Learning for Vision-Language Modelscited by 1,433
  9. 2017Learning to Compare: Relation Network for Few-Shot Learningcited by 1,143
  10. 2021CLIP-Adapter: Better Vision-Language Models with Feature Adapterscited by 1,026
  11. 2017ChestX-ray8: Hospital-scale Chest X-ray Database and Benchmarks on Weakly-Supervised Classification and Localization of Common Thorax Diseasescited by 972
  12. 2022MaPLe: Multi-modal Prompt Learningcited by 687
29 more
  1. 2019A Closer Look at Few-shot Classificationcited by 684
  2. 2019Skin Lesion Analysis Toward Melanoma Detection 2018: A Challenge Hosted by the International Skin Imaging Collaboration (ISIC)cited by 669
  3. 2022Tip-Adapter: Training-free Adaption of CLIP for Few-shot Classificationcited by 362
  4. 2022Prompt-aligned Gradient for Prompt Tuningcited by 341
  5. 2022Visual Classification via Description from Large Language Modelscited by 336
  6. 2019Meta-Dataset: A Dataset of Datasets for Learning to Learn from Few Examplescited by 315
  7. 2023Visual-Language Prompt Tuning with Knowledge-guided Context Optimizationcited by 308
  8. 2023Self-regulating Prompts: Foundational Model Adaptation without Forgettingcited by 272
  9. 2022What does a platypus look like? Generating customized prompts for zero-shot image classificationcited by 267
  10. 2016Using Deep Learning for Image-Based Plant Disease Detectioncited by 219
  11. 2020Cross-Domain Few-Shot Classification via Learned Feature-Wise Transformationcited by 185
  12. 2022Task Residual for Tuning Vision-Language Modelscited by 123
  13. 2024Low-Rank Few-Shot Adaptation of Vision-Language Modelscited by 104
  14. 2021Cross-domain Few-shot Learning with Task-specific Adapterscited by 80
  15. 2023StyleAdv: Meta Style Adversarial Training for Cross-Domain Few-Shot Learningcited by 53
  16. 2022Understanding Cross-Domain Few-Shot Learning Based on Domain Similarity and Few-Shot Difficultycited by 37
  17. 2021Ranking Distance Calibration for Cross-Domain Few-Shot Learningcited by 27
  18. 2024Flatten Long-Range Loss Landscapes for Cross-Domain Few-Shot Learningcited by 26
  19. 2022Wave-SAN: Wavelet based Style Augmentation Network for Cross-Domain Few-Shot Learningcited by 23
  20. 2024Prompt as Free Lunch: Enhancing Diversity in Source-Free Cross-domain Few-shot Learning through Semantic-Guided Promptingcited by 10
  21. 2024Reconstruction Target Matters in Masked Image Modeling for Cross-Domain Few-Shot Learningcited by 9
  22. 2024Step-wise Distribution Alignment Guided Style Prompt Tuning for Source-free Cross-domain Few-shot Learningcited by 5
  23. 2026Reclaiming Lost Text Layers for Source-Free Cross-Domain Few-Shot Learningcited by 5
  24. 2026Mind the Discriminability Trap in Source-Free Cross-domain Few-shot Learningcited by 4
  25. 2026Interpretable Cross-Domain Few-Shot Learning with Rectified Target-Domain Local Alignmentcited by 4
  26. 2025Revisiting Continuity of Image Tokens for Cross-domain Few-shot Learningcited by 3
  27. 2026Addressing Exacerbated Attention Sink for Source-Free Cross-Domain Few-Shot Learningcited by 2
  28. 2026Reviving In-domain Fine-tuning Methods for Source-Free Cross-domain Few-shot Learningcited by 1
  29. 2026Improving CLIP Adaptation by Breaking Tail Alignment for Source-Free Cross-Domain Few-Shot Learningcited by 1

Built on by

Not cited yet.