Classifier Retraining Boosts Long-Tailed Recognition Accuracy.

Juan Terven, Diana Margarita C\'ordova Esparza, Julio Alejandro Romero Gonzalez, Edgar Arturo Ch\'avez Urbiola, Francisco Javier Willars Rodriguez, Juan Bautista Hurtado Ramos, Alfonso Ramirez Pedraza· July 14, 2026 View original

Key takeaways

  • Long-tailed recognition is a common challenge in real-world datasets.
  • Retraining only the classifier after Balanced Softmax training significantly boosts few-shot accuracy.
  • The BS-cRT method is a simple yet effective two-stage approach.
  • It offers substantial performance gains on underrepresented classes across various benchmarks.

Who benefits

HealthcareE-commerceSecurityAutonomous Vehicles

Summary

This paper introduces BS-cRT, a two-stage method for long-tailed recognition that significantly improves few-shot accuracy by retraining only the classifier after initial Balanced Softmax training. The approach consistently outperforms existing baselines across various image recognition datasets.

In machine learning, long-tailed recognition problems arise when some classes have many more training examples than others, leading models to perform poorly on rare "tail" classes. Existing methods often modify loss functions or representations to address this imbalance. This research investigates a simpler, two-stage approach called BS-cRT. The first stage involves training a backbone network and a cosine classifier using Balanced Softmax, a technique designed to mitigate class imbalance. After this initial training, the backbone network is frozen, and only the classifier is retrained. This retraining occurs on balanced episodic batches, maintaining the empirical-prior Balanced Softmax objective. Evaluations across multiple long-tailed datasets, including CIFAR-100-LT, ImageNet-LT, and Places-LT, show that this classifier-only retraining step consistently improves few-shot accuracy. For instance, at an imbalance factor of 100, few-shot gains were substantial, demonstrating that a focused retraining of the classifier can effectively reduce errors in tail classes without complex architectural changes. The study also analyzes limitations of other boundary-probe extensions, reinforcing BS-cRT as a practical and strong baseline.

Why it matters

Professionals working with real-world datasets often encounter class imbalance. This method offers a straightforward and effective way to improve model performance on underrepresented classes, which is crucial for applications like fraud detection, medical diagnosis, or product recommendation where rare events are critical.

How to implement this in your domain

  1. 1Apply Balanced Softmax as the initial training strategy for image classification models on imbalanced datasets.
  2. 2Implement a second training stage where the backbone is frozen, and only the classifier weights are updated using balanced episodic batches.
  3. 3Evaluate the few-shot accuracy improvements on tail classes using relevant metrics for your specific application.
  4. 4Consider integrating this two-stage approach into existing model training pipelines for long-tailed recognition tasks.

Original post by Juan Terven, Diana Margarita C\'ordova Esparza, Julio Alejandro Romero Gonzalez, Edgar Arturo Ch\'avez Urbiola, Francisco Javier Willars Rodriguez, Juan Bautista Hurtado Ramos, Alfonso Ramirez Pedraza

"arXiv:2607.09832v1 Announce Type: new Abstract: Long-tailed recognition methods often modify losses, margins, or representations to reduce the dominance of frequent classes. We ask whether, after Balanced Softmax training, the remaining tail error can be reduced by retraining onl…"

View on X

Originally posted by Juan Terven, Diana Margarita C\'ordova Esparza, Julio Alejandro Romero Gonzalez, Edgar Arturo Ch\'avez Urbiola, Francisco Javier Willars Rodriguez, Juan Bautista Hurtado Ramos, Alfonso Ramirez Pedraza on X · view source

Want to go deeper?

Turn these trends into skills with Learnijoy's hands-on AI & tech courses.

Explore courses

More in AI Research

AI ResearchAI Engineering & DevTools

Emotional Preferences Regulate Goal Priorities in Reinforcement Learning Agents

This paper proposes a computational framework where higher-level goals autonomously generate state-dependent emotional preferences to regulate the priorities of competing lower-level objectives in reinforcement learning agents. It demonstrates how this emergent preference function exhibits contextual priority switching and improves performance over fixed-preference strategies in multi-objective exploration environments.

Shiqi Liu, Yihua Tan, Hu Fu, Guanyu QiAug 28, 2026
AI Engineering & DevToolsAI Research

New Framework Unifies Task Detection and Adaptation for Continual Learning

This paper proposes FiUni, a Fisher-guided unified framework for task-free continual learning in LLMs that combines batch-level task detection with parameter-efficient adaptation. FiUni uses Fisher information matrix (FIM) properties to dynamically determine whether to reuse, expand, or create new low-rank adaptation (LoRA) subspaces, effectively mitigating catastrophic forgetting without explicit task boundaries.

Dezheng Han, Anbang Zhang, Zhihao Zhu, Shuaishuai GuoAug 28, 2026
AI Engineering & DevToolsAI Research

Soft EMG Interface Enables Machine Learning-Powered Silent Speech Recognition

This paper introduces a soft, active electromyography (EMG) interface worn on the hand that enables word-level silent speech recognition (SSR) using machine learning. The device acquires stable EMG signals from a fingertip electrode near the lips, achieving 97.2% accuracy on a 30-word vocabulary and demonstrating real-time drone control in noisy environments.

Yuta Kurotaki, Shusuke Yamakoshi, Reitaro Yoshida, Yutaka Isoda, Tamami Takano, Yuji Isano, Yusuke Miyake, Kentaro Kuribayashi, Hiroki OtaAug 28, 2026