The paper introduces a unified framework that simultaneously measures intrinsic (encoded) and extrinsic (expressed) gender bias in large language models using identical neutral prompts. It finds a consistent link between latent gender information and output bias, but shows that alignment via supervised fine‑tuning reduces expressed bias while leaving internal gender associations largely intact and reactivatable by adversarial prompts. The study also demonstrates that debiasing gains on structured benchmarks may not transfer to realistic tasks such as story generation.
By Nour Bouchouchi, Thibault Laugel, Xavier Renard, Christophe Marsala, Marie-Jeanne Lesot, Marcin Detyniecki
MERIT is a two‑stage framework for reviewer assignment that first trains a reviewer assessor using reinforcement learning to match paper‑specific expertise rubrics with reviewers’ prior work, guided by an LLM judge. The assessor’s predictions are then distilled into an embedding‑based retriever for efficient large‑scale assignment. Experiments show the 4B assessor outperforms larger general‑purpose LLMs on suitability classification, and the retriever achieves state‑of‑the‑art performance on LR‑Bench and the CMU Gold dataset.
By Zixuan Yang, Yibo Zhao, Weicong Liu, Xiang Li
MisEdu‑RAG is a dual‑hypergraph retrieval‑augmented generation framework designed to help novice math teachers diagnose and remediate student misconceptions. It structures pedagogical knowledge as a concept hypergraph and real student mistake cases as an instance hypergraph, performing two‑stage retrieval to ground responses in both layers. On the MisstepMath dataset, MisEdu‑RAG outperforms baseline models, improving token‑F1 by 10.95% and achieving up to 15.3% higher quality across five dimensions, especially in diversity and empowerment.
By Zhihan Guo, Yuting Lu, Jionghao Lin
The paper introduces GraphemeNet, a unified multi‑script handwritten character recognition architecture that explicitly encodes script‑geometric regularities. It uses two orthogonal binary axes: Persistent Scaffold Injection (PSI) to embed stroke‑level geometry into each encoder stage, and a choice between gated global pooling or a Stroke Topology Module for spatial relational reasoning. Across fourteen benchmarks in eight writing systems, GraphemeNet achieves state‑of‑the‑art performance with fewer parameters, demonstrating the effectiveness of structural‑prior efficiency for multi‑script HCR.
By Ranjit Raut, Aarav Subedi, Ashim Shrestha
The paper introduces a benchmark called Shedding Light to evaluate how well generative image models understand and reproduce lighting. The benchmark tests models by asking them to inpaint a simple object, called a light probe, into real photographs and then compares the generated probe to the ground truth to assess lighting direction, colour, and radiance. The authors provide a scalable protocol and open-source code and data for systematic assessment of photometric accuracy in future models.
By Justine Giroux, Jack Oliver Hilliard, Yannick Hold-Geoffroy, Javier Vazquez-Corral, Jean-Fran\c{c}ois Lalonde
CamPilot is a multi‑agent cinematic assistant that combines cinematographic planning with camera‑work control to generate more coherent and aesthetically pleasing movies from text prompts. It learns camera‑work planning from 14,000 professional films using a GRPO‑based learning paradigm, capturing motion patterns, composition principles, and cross‑shot relationships. The system is evaluated with a new benchmark, CamEval, and outperforms existing text‑to‑movie methods in cinematographic control and quality.
By Yang Wu, Stefano Petrangeli, Ishita Dasgupta, Yu Shen
The paper introduces LAION-Mobile, a new dataset of roughly one million smartphone images with EXIF metadata, designed to test deepfake detectors on modern computational photography content. Twelve state‑of‑the‑art detectors were evaluated on a 9,115‑image subset, revealing that none achieved an AUC above 0.624 on AI‑generated images and five performed worse than chance. The study also shows that threshold calibration on legacy data can hide high false‑alarm rates, with some detectors flagging 17–91 % of real photos when recalibrated for modern content.
By Achim von Stryk, Janis Keuper
UniH$^3$ is a new framework for all-in-one medical image restoration that unifies hierarchical homogeneity and heterogeneity. It introduces a Hierarchical Homogeneity Memory (H2M) module to distill and retrieve shared anatomical priors, and a Hierarchical Heterogeneity Balancer (H2B) to mitigate inter- and intra-task conflicts during training. Experiments on MedIR-2D-500K and MedIR-3D-3K show that UniH$^3$ achieves state‑of‑the‑art performance for both multi‑task and single‑task restoration.
By Zhiwen Yang, Jiayin Li, Chengyu Liu, Hui Zhang, Bingzheng Wei, Yan Xu
Tri‑DehazeGS is a Gaussian Splatting framework that decouples clean scene reconstruction from atmospheric haze by representing the scene with Gaussian primitives and the haze medium with an independent view‑shared tri‑plane field. It uses a physical scattering model to compose hazy observations and introduces Medium‑Decoupled Transmittance Gradient Compensation (MD‑TGC) to re‑balance gradients in low‑transmittance regions without altering forward rendering. Experiments on real and synthetic haze benchmarks demonstrate that this approach improves clean novel‑view reconstruction.
By Kui Jiang, Yang Gu, Jiacheng Liu, Shiyu Liu, Youyu Chen, Hui Liu
The paper introduces Sewer-Transformer-ML, a hierarchical vision Transformer that fuses multi‑level features for multi‑label sewer defect classification, and two lightweight variants, Sewer-MobileNet-ML and Sewer-Mobile-TransNet, tailored for resource‑constrained inspection scenarios. On the Sewer‑ML test set, Sewer‑Transformer‑ML‑Base achieved an $F2_{ ext{CIW}}$ of 65.68% and an $F1_{ ext{Normal}}$ of 92.68%, topping the public leaderboard and surpassing the next best method by 7.6 percentage points in $F2_{ ext{CIW}}$. The lightweight Sewer‑MobileNet‑ML reached a comparable $F2_{ ext{CIW}}$ of 65.73% with only 17 M parameters, a 95% reduction from the base model, while Sewer‑Mobile‑TransNet achieved 96.43% accuracy under the standard data split, and ablation studies highlighted the effectiveness of direct concatenation for Transformer features and attention‑based fusion for multiscale CNN features.
By Xu Fang, Zhuoran Wang, Qing Li, Shengyu Zhang, Guanzhi Deng, Jianbiao He, Qingquan Li
The paper introduces a new framework for unsupervised visible‑infrared person re‑identification that leverages modality‑unified prototypes. By contrasting with prototypes that unify both modalities, the method jointly optimizes similarity within and across modalities, improving modality invariance. A self‑distillation step refines instance‑prototype relationships using a steady teacher, resulting in a simple yet effective model validated on standard VI‑ReID benchmarks.
By Menglin Wang, Xiaojin Gong
LoopVAE introduces a recurrent depth architecture that reuses a scale‑ and loop‑conditioned core across different spatial scales while keeping resolution‑changing transitions separate. The four‑block core applies 28 block operations per encoder or decoder, enabling a 29M‑parameter convolutional model to achieve 0.28 rFID and 32.54 dB PSNR on ImageNet‑256 with roughly 65% fewer parameters than comparable VAEs. Experiments with both convolutional and Transformer operators, as well as ablations on parameter sharing, demonstrate competitive image quality metrics and reveal how targeted loop interventions and truncation affect reconstruction quality and computational trade‑offs.
By Zhiying Lu
The paper introduces the Consistency Memory Bank (COMB), a label‑free virtual staining framework designed to process gigapixel Whole Slide Images without the memory bottlenecks of patch‑based deep learning. COMB decouples context storage from computation, using a dynamic retrieval mechanism to fetch feature representations from adjacent tiles, local padding to resolve spatial discontinuities, and neighbor‑aware channel attention to stabilize statistical drift. The method achieves superior perceptual fidelity and tiling consistency compared to state‑of‑the‑art baselines, and its improved continuity suggests downstream benefits for tumor segmentation.
By Dou Hoon Kwark, Kianoush Falahkheirkhah, Ji-hun Oh, Shirui Luo, Volodymyr Kindratenko, Rohit Bhargava
InstantHDR is a feed-forward network that initializes high dynamic range (HDR) 3D scenes from uncalibrated multi-exposure low dynamic range (LDR) image collections in a single forward pass. It uses geometry-guided appearance modeling for multi-exposure fusion and a meta-network for scene-specific tone mapping. The authors also created a pre-training dataset, HDR-Pretrain, with 168 Blender-rendered scenes to support generalizable HDR models, achieving a speedup of about 700× over state‑of‑the‑art optimization methods while maintaining comparable quality after lightweight post‑optimization.
By Dingqiang Ye, Jiacong Xu, Jianglu Ping, Yuxiang Guo, Chao Fan, Vishal M. Patel
V‑Retrver is an evidence‑driven retrieval framework that treats universal multimodal retrieval as an agentic reasoning process grounded in visual inspection. It allows multimodal large language models to selectively acquire visual evidence through external tools, alternating between hypothesis generation and targeted visual verification. The approach is trained with a curriculum that blends supervised activation, rejection‑based refinement, and reinforcement learning, achieving an average 23.0% improvement in retrieval accuracy across multiple benchmarks.
By Dongyang Chen, Chaoyang Wang, Dezhao Su, Xi Xiao, Zeyu Zhang, Jing Xiong, Qing Li, Yuzhang Shang, Shichao Kan
The paper introduces DirectSwap, a mask‑free video head‑swapping method that leverages a newly created cross‑identity paired dataset, HeadSwapBench. By synthesizing expression‑synchronized video pairs from real footage, the authors provide frame‑aligned ground truth for full‑reference evaluation of identity, expression, pose, reconstruction fidelity, and temporal stability. DirectSwap outperforms traditional same‑identity masked reconstruction, especially when head silhouettes change, and can restore non‑head content without external segmentation.
By Yanan Wang, Shengcai Liao, Panwen Hu, Xin Li, Fan Yang, Guangxi Liu, Xiaodan Liang
ABACUS is a 3‑billion‑parameter vision‑language model that simultaneously tackles object counting, crowd counting, referring‑expression counting, and count‑faithful image generation. It introduces density‑aware adaptive zooming with an objectness map, a boundary‑aware count policy trained via GRPO to avoid over‑ or under‑counting at crop edges, and a cycle‑consistent GRPO strategy that scores generated images for count accuracy and aesthetic quality without external critics. The model sets new state‑of‑the‑art performance on seven benchmarks, outperforming both specialized and larger generalist models.
By Anindya Mondal, Sauradip Nag, Anjan Dutta
The paper evaluates the new YOLO26 architecture, which offers NMS-free end-to-end inference and is tailored for CPU-based edge devices, against three earlier Ultralytics models (YOLOv5u, YOLOv8, and YOLO11) in aquaculture fish mortality detection. Across nano, small, and medium scales, all models achieved similar detection accuracy on a full dataset, but differences emerged in data efficiency and deployment performance: YOLOv8 reached 90% mAP50 with only 400 images, while YOLO26 variants needed 1,000 images; YOLO26n was fastest on a Raspberry Pi 5 (7.51 FPS), whereas YOLOv5mu led on CPU-based hardware. The study concludes that architectural novelty alone does not dictate suitability for edge AI in aquaculture; training data size, target hardware, and inference needs must be jointly considered.
By Rakesh Ranjan, Gajanan S. Kothawade, Kata Sharrer, Scott Tsukuda, Christopher Good
HeteroPROMPT is a real‑time, privacy‑preserving framework for heterogeneous collaborative perception in autonomous systems. It aligns features from diverse sensors and models into a unified ego‑centric space using modular prompts and lightweight tuning, while keeping encoders and fusion stacks frozen. The system employs a metadata‑free autoencoder for modality classification and routing, achieving higher average precision on OPV2V‑H and V2XSet datasets with far fewer trainable parameters.
By Armin Maleki, Hayder Radha
FreeTransformSR is a lightweight image super‑resolution network that uses a channel‑wise free low‑rank learnable transform to adaptively modulate features with minimal parameters. It adds a local feature modulation branch with depthwise convolution and a soft complexity adaptive module that fuses local convolution and window self‑attention based on texture characteristics. The model also employs an adaptive intensity modulation strategy and achieves competitive PSNR/SSIM on five benchmark datasets while using only 595K parameters and running faster than competing methods.
By Hongji Li, Yunhui Li