CodeGoat24/UnifiedReward-Edit-qwen35-27b

VISIONConcurrent Unit Cost:2Model Size:27BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Mar 8, 2026License:mitArchitecture:Transformer Open Weights Featherless Exclusive Cold

CodeGoat24/UnifiedReward-Edit-qwen35-27b is a 27 billion parameter unified reward model developed by CodeGoat24, specifically designed for evaluating both Text-to-Image and Image-to-Image generation tasks. This model excels at assessing image editing quality, supporting pairwise ranking, pairwise scoring, and pointwise scoring based on instruction-following and overall image quality. It is optimized for providing feedback on generated and edited images, making it suitable for applications requiring automated visual content evaluation.

Loading preview...

UnifiedReward-Edit-qwen35-27b Overview

CodeGoat24/UnifiedReward-Edit-qwen35-27b is a 27 billion parameter reward model developed by CodeGoat24, uniquely designed to evaluate both Text-to-Image (T2I) and Image-to-Image (I2I) generation. This model's primary differentiator is its specialized capability in assessing image editing quality, providing a unified framework for multimodal understanding and generation.

Key Capabilities

  • Unified Reward System: Evaluates both T2I and I2I outputs within a single model architecture.
  • Image Editing Assessment: Supports detailed evaluation of edited images through multiple methods:
    • Pairwise Rank: Directly compares two edited images to determine which is superior.
    • Pairwise Score: Assigns individual scores to each image in a pair.
    • Pointwise Score: Rates a single image on two distinct axes: adherence to instructions and overall image quality.
  • Training Data: Utilizes preprocessed data from EditScore and EditReward datasets, enhancing its ability to judge image editing tasks.

Good For

  • Automated Evaluation: Ideal for developers needing to automatically assess the quality and instruction-following of generated or edited images.
  • Image Generation Research: Useful for researchers working on improving T2I and I2I models by providing a robust reward signal.
  • Quality Control: Can be integrated into pipelines for filtering or ranking image outputs based on predefined criteria.