# COLLAR: Cascaded Object-Level Latent Refinement for High-Fidelity Conditional Generation Source: [COLLAR: Cascaded Object-Level Latent Refinement for High-Fidelity Conditional Generation]() Collection index: [llms.txt]() Data updated: 2026-06-02T05:00:51.558Z Authors: Xinlong Zhang, Jia Wei, Xiaoyu Zhang, Teng Zhou, Chengyu Lin, Yongchuan Tang Published: 2026-05-31T02:10:34.000Z ## Abstract Achieving high-fidelity object-level control in Diffusion Transformers remains a significant challenge despite the introduction of structural priors like depth and Canny maps. Current object-level conditional generation methods frequently suffer from visual artifacts and struggle to maintain precise control over objects within small localized regions. To address these limitations, we propose Cascaded Object-Level Latent Refinement (COLLAR), a training-free framework that progressively optimizes object-level features via the Field-of-View (FoV) expansion. First, we propose the Cross-Scale Semantic Alignment (CSSA) module to address spatial-semantic gaps by injecting object-level features into extended-FoV branches via attention mechanisms. To further optimize these features, the Cyclic Feature Injection (CFI) module introduces a reciprocal background feedback mechanism. It leverages a frequency-based adaptive strategy to selectively update the global backbone with context-aligned local information. Finally, the extended-FoV branch serves as a hub for feature optimization, ensuring that object-level features are integrated into the global generation process without compromising final image quality. Extensive experiments on the COCO-MIG and COCO-POS benchmarks demonstrate that our approach consistently outperforms state-of-the-art methods across semantic alignment, image quality, and spatial fidelity. ## Recorded benchmark results A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage. - [Canny-conditioned Multi-instance Generation on COCO-MIG Canny (test)](): Canny-conditioned Multi-instance Generation; COCO-MIG Canny (test); ISR (n=2): 34.9. - [Depth-conditioned Multi-instance Generation on COCO-MIG Depth (test)](): Depth-conditioned Multi-instance Generation; COCO-MIG Depth (test); Image Success Ratio (n=2): 65.3. - [Multi-instance Generation on COCO-POS Canny (test)](): Multi-Instance Generation; COCO-POS Canny (test); G-CLIP: 27.1. - [Depth-conditioned Image Generation on Conditional Generation Benchmark SD3-based](): Depth-conditioned Image Generation; Conditional Generation Benchmark SD3-based; G-CLIP: 29.11. - [Multi-instance Generation on COCO-POS Depth (test)](): Multi-Instance Generation; COCO-POS Depth (test); G-CLIP: 27.19. ## Original sources and implementation [Original paper]() [PDF]()