# arXiv cs.AI: A two-path network for low-light image enhancement [Canonical reading](https://escapemeridian.com/readings/arxiv-cs-ai-a-two-path-network-for-low-light-image-enhancement-208b0f9a) The authors of “Cue-guided dual-path reconstruction with adaptive multi-scale response refinement for low-light image enhancement” present CueDR-Net, a network designed to brighten low-light images while reconstructing local content. They argue that existing methods can introduce colour deviations or blur details. CueDR-Net derives luminance variation, colour, and texture cue maps from an input image, then uses bidirectional attention to exchange features between a cue stream and an image stream. One path predicts pixel wise adjustment maps for iterative intensity changes; a second predicts an additive correction for local content. Adaptive Multi-Scale Response Refinement Convolution (AMRRConv) refines features for both paths, combining learnable local transition responses with multi-scale features. The authors report competitive enhancement performance on LOLv1, LSRW-HUAWEI, and LSRW-NIKON, and say AMRRConv achieves competitive low-light detection performance with a small neck parameter count when integrated into YOLOv11. The approach combines brightness adjustment with local content reconstruction. The abstract reports comparative results and ablation studies but gives no numerical results or specific examples, so the extent of the reported performance is not clear from this evidence. Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims. ## Original source arXiv cs.AI · Article Original source date: Mar 17, 2026 (2026-03-17). Date basis: Date basis not provided. Added to Escape Meridian: 2026-11-08T19:54:36.688Z. First observed: 2026-11-06T05:03:32.444Z. Retrieved: 2026-11-08T19:53:25.127Z. - [Cue-guided dual-path reconstruction with adaptive multi-scale response refinement for low-light image enhancement](https://arxiv.org/abs/2603.16482) ## Limits The abstract reports competitive performance and ablation studies but provides no numerical results or specific examples for judging the size of the gains. ## Open question What evaluation results support the authors’ description of CueDR-Net’s enhancement performance on LOLv1, LSRW-HUAWEI, and LSRW-NIKON? ## Source-stated The authors propose CueDR-Net, combining brightness adjustment and local content reconstruction through bidirectional cue interaction and adaptive multi-scale refinement. > we propose CueDR-Net, a dual-path network that combines brightness adjustment with local content reconstruction through bidirectional cue interaction and adaptive multi-scale refinement. - [Supporting source](https://arxiv.org/abs/2603.16482) ## Source-stated CueDR-Net uses one path to predict pixel wise adjustment maps and another to predict an additive correction for local content reconstruction. > The brightness adjustment path predicts pixel wise adjustment maps that control iterative intensity changes in the input image. In parallel, the residual path uses the exchanged features to predict an additive correction for local content reconstruction. - [Supporting source](https://arxiv.org/abs/2603.16482) ## Source-stated The authors report competitive enhancement performance on three named datasets and competitive low-light detection performance for AMRRConv integrated into YOLOv11. > Evaluations on LOLv1, LSRW-HUAWEI, and LSRW-NIKON demonstrate competitive enhancement performance, supported by ablation studies. AMRRConv also achieves competitive low-light detection performance with a small neck parameter count when integrated into YOLOv11. - [Supporting source](https://arxiv.org/abs/2603.16482) ## Source and contributor context Shi, Yicui, Chen, Yuhan, Huang, Xiangfei, Wang, Zhenguo, Yu, Wenxuan, Fang, Ying, Li, Jie, Li, Guofa · Article author These are credits stated by the source; they do not establish a profile match or authorship of every passage. ## Source media No suitable media was found by the supported page extraction. ## Terms and references Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement. ## Model parameters The internal numbers a model adjusts during training; their count is a rough measure of a model's size. - [Definition reference](https://developers.google.com/machine-learning/glossary#parameter) ## arXiv arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed. - [Explore this context](/entities/arxiv) - [Reference](https://arxiv.org/) - [Reference](https://info.arxiv.org/help/api/index.html) ## Editorial provenance Version 1. Publication ID: 9fbfc4a6-2765-4d52-8c56-ea6241d21762. Prepared by gpt-6-luna; grounding review by gpt-6-luna. Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-08T19:54:36.688Z. This checks fidelity to the source, not whether its claims are independently true. Editorial selection: accept; reasons: material\_development. Observation: 3afbe3130c6e17dd55e585d838df74631c0870691debb73d0225d5dec5677033. Source text hash: 51a75beeaa152e1a7ab6afb493560eed26eb765723c28716b147e26724e194dd. ## Revision and source-check history ### Version 1 — 2026-11-08T19:54:36.688Z The authors of “Cue-guided dual-path reconstruction with adaptive multi-scale response refinement for low-light image enhancement” present CueDR-Net, a network designed to brighten low-light images while reconstructing local content. They argue that existing methods can introduce colour deviations or blur details. CueDR-Net derives luminance variation, colour, and texture cue maps from an input image, then uses bidirectional attention to exchange features between a cue stream and an image stream. One path predicts pixel wise adjustment maps for iterative intensity changes; a second predicts an additive correction for local content. Adaptive Multi-Scale Response Refinement Convolution (AMRRConv) refines features for both paths, combining learnable local transition responses with multi-scale features. The authors report competitive enhancement performance on LOLv1, LSRW-HUAWEI, and LSRW-NIKON, and say AMRRConv achieves competitive low-light detection performance with a small neck parameter count when integrated into YOLOv11. The approach combines brightness adjustment with local content reconstruction. The abstract reports comparative results and ablation studies but gives no numerical results or specific examples, so the extent of the reported performance is not clear from this evidence. ## Original source arXiv cs.AI · Article Original source date: Mar 17, 2026 (2026-03-17). Date basis: Date basis not provided. Added to Escape Meridian: 2026-11-08T19:54:36.688Z. First observed: 2026-11-06T05:03:32.444Z. Retrieved: 2026-11-08T19:53:25.127Z. - [Cue-guided dual-path reconstruction with adaptive multi-scale response refinement for low-light image enhancement](https://arxiv.org/abs/2603.16482) ## Limits The abstract reports competitive performance and ablation studies but provides no numerical results or specific examples for judging the size of the gains. ## Open question What evaluation results support the authors’ description of CueDR-Net’s enhancement performance on LOLv1, LSRW-HUAWEI, and LSRW-NIKON? ## Source-stated The authors propose CueDR-Net, combining brightness adjustment and local content reconstruction through bidirectional cue interaction and adaptive multi-scale refinement. > we propose CueDR-Net, a dual-path network that combines brightness adjustment with local content reconstruction through bidirectional cue interaction and adaptive multi-scale refinement. - [Supporting source](https://arxiv.org/abs/2603.16482) ## Source-stated CueDR-Net uses one path to predict pixel wise adjustment maps and another to predict an additive correction for local content reconstruction. > The brightness adjustment path predicts pixel wise adjustment maps that control iterative intensity changes in the input image. In parallel, the residual path uses the exchanged features to predict an additive correction for local content reconstruction. - [Supporting source](https://arxiv.org/abs/2603.16482) ## Source-stated The authors report competitive enhancement performance on three named datasets and competitive low-light detection performance for AMRRConv integrated into YOLOv11. > Evaluations on LOLv1, LSRW-HUAWEI, and LSRW-NIKON demonstrate competitive enhancement performance, supported by ablation studies. AMRRConv also achieves competitive low-light detection performance with a small neck parameter count when integrated into YOLOv11. - [Supporting source](https://arxiv.org/abs/2603.16482) ## Source and contributor context Shi, Yicui, Chen, Yuhan, Huang, Xiangfei, Wang, Zhenguo, Yu, Wenxuan, Fang, Ying, Li, Jie, Li, Guofa · Article author These are credits stated by the source; they do not establish a profile match or authorship of every passage. ## Source media No suitable media was found by the supported page extraction. ## Terms and references Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement. ## Model parameters The internal numbers a model adjusts during training; their count is a rough measure of a model's size. - [Definition reference](https://developers.google.com/machine-learning/glossary#parameter) ## arXiv arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed. - [Explore this context](/entities/arxiv) - [Reference](https://arxiv.org/) - [Reference](https://info.arxiv.org/help/api/index.html) ## Editorial provenance Version 1. Publication ID: 9fbfc4a6-2765-4d52-8c56-ea6241d21762. Prepared by gpt-6-luna; grounding review by gpt-6-luna. Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-08T19:54:36.688Z. This checks fidelity to the source, not whether its claims are independently true. Editorial selection: accept; reasons: material\_development. Observation: 3afbe3130c6e17dd55e585d838df74631c0870691debb73d0225d5dec5677033. Source text hash: 51a75beeaa152e1a7ab6afb493560eed26eb765723c28716b147e26724e194dd. ### uncertain — 2026-11-08T19:54:36.688Z — anthropic/claude-opus-5.5 Publication: 9fbfc4a6-2765-4d52-8c56-ea6241d21762 attribution — pass: The reading attributes CueDR-Net and its performance claims to the paper's authors. Its 'arXiv cs.AI' label matches the feed; the paper is primarily cs.CV and cross-listed in cs.AI. Results are framed as author reports, not established facts. > we propose CueDR-Net, a dual-path network that combines brightness adjustment with local content reconstruction summary — uncertain: The method description is faithful but hard for a general reader to follow. It omits the source's stated purpose: low-light image enhancement supports optical sensing and scene observation when limited photon capture and sensor noise degrade images. Central terms such as YOLOv11, neck parameter count, ablation studies and bidirectional attention are unexplained, and term-library tooltips are not shown. Smallest repair: add the source's stated motivation in one plain sentence and attach tooltips or suggested term additions. The sentence 'The approach combines brightness adjustment...' repeats an earlier point. > Low-light image enhancement (LLIE) supports optical sensing and scene observation when limited photon capture and sensor noise degrade image quality. date — pass: originalPublished 2026-03-17 matches the v1 submission date. It is not the 4 Nov 2026 revision date or the November feed date. > \[Submitted on 17 Mar 2026 (v1), last revised 4 Nov 2026 (this version, v2)\] support — pass: All three claims quote the abstract accurately and keep the authors' hedge of 'competitive' performance. The limitation, that the abstract gives no numbers, is accurate. The question follows from that gap without leading the reader, and the reading draws no unsupported benefits or costs. > Evaluations on LOLv1, LSRW-HUAWEI, and LSRW-NIKON demonstrate competitive enhancement performance, supported by ablation studies. media — pass: The selected media is the verified direct link to the original arXiv abstract page, matching the source URL. No better source media was found. ## Recorded references Explicit citations in public readings and public file attachments to this reading. Shared topics are not citations. Other artifact types need an explicit reference adapter. No public references recorded here yet. [Reference data (JSON)](/readings/208b0f9a4448ee27170cd6006f7dfc695fe430af5ebb81cef1550ff38d76c6de/references.json) ## Corrections and challenges [Send a correction or an opposing source](https://escapemeridian.com/corrections/) [JSON record](https://escapemeridian.com/readings/arxiv-cs-ai-a-two-path-network-for-low-light-image-enhancement-208b0f9a.json)