CARE: Class-Adaptive Expert Consensus for Reliable Learning with Long-Tailed Noisy Labels

Learning from real-world data is frequently hindered by the compound challenge of long-tailed class distributions and noisy annotations. Existing methods partially address these issues but typically ignore the non-uniform impact of label noise across classes, resulting in ineffective correction for tail classes and over-regularization for head classes. To address this issue, we propose Class-Adaptive Rectification with Experts (CARE), a parameter-efficient framework that leverages three complementary supervision sources from vision-language models (VLM): observed noisy labels, VLM text embeddings, and visual features. CARE introduces a class-adaptive expert consensus mechanism that enforces stricter agreement for tail classes and more permissive agreement for head classes based on class frequency. By aggregating high-confidence predictions across these sources, CARE filters unreliable signals and recalibrates class distributions, yielding more reliable rectification under long-tailed distributions. Extensive experiments on both synthetic and real-world benchmarks demonstrate that CARE consistently outperforms state-of-the-art methods, achieving up to 3.0\% performance gains. The source code is available at https://github.com/qwq123-study/CARE.

Modern AI systems often learn from large collections of real-world data, but these datasets are rarely perfect. Some categories contain many examples for taining while others have very few, and labels may contain mistakes introduced by human annotation or automatic data collection systems. This makes it difficult for AI models to learn reliably, especially for rare categories. To address this problem, we developed a new training approach called CARE. Rather than depending on only one type of information, CARE learns from several clues at the same time. It considers the original labels in the dataset, the visual patterns inside the images, and the broader knowledge captured by modern AI systems that connect images with language. CARE further adjusts how carefully it makes decisions depending on whether a category appears often or only rarely in the data. For rare categories, CARE applies stricter checks before trusting a prediction, while for common categories it allows more flexibility. By combining only the most reliable signals from different sources, CARE filters out incorrect information and helps rebalance how different categories are learned, leading to more accurate learning overall.