Cross-Prompt Inference Mitigates Rare-Class Suppression in Imbalanced VLM Finetuning
cs.CV
Submitted: 2026-10-15
Updated: 2026-10-15
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Sources
- Qwen3-VL Technical Report
- Qwen2.5-VL Technical Report
- LPT: Long-tailed Prompt Tuning for Image Classification
- Follow-Up Differential Descriptions: Language Models Resolve Ambiguities for Image Classification
- Gemma 3 Technical Report
- Understanding Catastrophic Forgetting in Language Models via Implicit Inference
- Fine-Tuning can Distort Pretrained Features and Underperform Out-of-Distribution
- Visual Instruction Tuning
- Does VLM Classification Benefit from LLM Description Semantics?
- Fine-Grained Visual Classification of Aircraft
- Visual Classification via Description from Large Language Models
- What does a platypus look like? Generating customized prompts for zero-shot image classification
- Waffling around for Performance: Visual Classification with Random Words and Broad Concepts
- Long-Tail Learning with Foundation Model: Heavy Fine-Tuning Hurts
- Test-Time Prompt Tuning for Zero-Shot Generalization in Vision-Language Models
- From Head to Tail: Towards Balanced Representation in Large Vision-Language Models through Adaptive Data Calibration
- VL-LTR: Learning Class-wise Visual-Linguistic Representation for Long-Tailed Visual Recognition
- SigLIP 2: Multilingual Vision-Language Encoders with Improved Semantic Understanding, Localization, and Dense Features
- InternVL3.5: Advancing Open-Source Multimodal Models in Versatility, Reasoning, and Efficiency
- Robust fine-tuning of zero-shot models
Related papers
- Loss Knows Best: Detecting Annotation Errors in Videos via Loss Trajectories
- AnchorWeave: World-Consistent Video Generation with Retrieved Local Spatial Memories
- Benchmarking the Robustness of Foundation Models for Mammography under Domain Shift
- MambaX-Net: Dual-Input Mamba-Enhanced Cross-Attention Network for Longitudinal MRI Segmentation
- TeleOCR: Navigating Document Parsing Across Digital and Camera-Captured Documents
- A Survey on Efficient Vision-Language-Action Models