LoRA fine-tuned OCR models convert handwritten pottery records to structured data, reducing error rates to below 1.5%.

Gissu Valentina Naghavi, Dominik Hagmann, Martin Kampel et al.

S2GE enhances decoder usability in graph-augmented LLMs via sampling and alignment, significantly boosting accuracy.

Xiaoyu Guo, Pengcheng Chen, Jiong Yu et al.

Introduced Answer Probing-Guided Tree Search (APTS) to enhance solution diversity across three reasoning tasks.

Yi Fang, Que Shen, Chengpeng Li et al.

Cen-Prune enhances visual token pruning diversity via centering geometry correction, boosting LVLM performance.

Shunjie Wen, Jaeyeon Lee, Dong-Wan Choi

Introduced CBUS, a cognitive memory-bottleneck framework, to simulate children’s reading behavior, reducing the gap with real students by over 50%.

Krisztian Balog, Arild Michel Bakken

PersonaGen-1M uses MinHash LSH and semantic deduplication to build 1.03M intent-annotated buyer personas.

Dmitrij Żatuchin, Daniil Dzemesjuk

Proposes Discrete Diffusion Bridges (DDB) to resolve spatiotemporal misalignment in image generation using hybrid absorption and information-guided noise scheduling.

Xing Xie, Jiawei Liu, Shijun Zhou et al.

RIDGE selects frames in long videos using temporal signals, achieving top performance across four benchmarks.

Shanqing Xu, Meng Luo, Mengchen Qian et al.

EMERGE-Policy employs a graph-structured multi-agent framework, coordinating specialized sub-agents and skills to achieve system-level robot policies without fine-tuning.

Zhirui Fang, Qingchi Yu, Ziyang Chen et al.

Sequence-level distillation from Qwen3-14B raises small-model legal argument coverage, reaching ARCscore 0.781 on CANLII.

Mohamed Elaraby, Ahmed Elhady, Diane Litman

EVAR framework enhances evidence faithfulness and task performance in long-form narrative reasoning while controlling inference cost.

Peilin Liu, Zhiquan Ji, Jinglong Ping

VeloBins learns velocity and uncertainty via bin classification, reducing errors by 3-53% across four datasets.

Maulana Bisyir Azhari, Seungwook Lee, Donghun Han et al.

ACTD employs vocabulary and sequence alignment with residual regularization, achieving state-of-the-art results on five reasoning benchmarks, with +4.5 points improvement.

Huiyi Zhang, Zijian Li, Xiaocheng Feng et al.

Proposes an evidence-gated multi-agent cloud MLOps framework enabling end-to-end automation from natural language tasks to verified deployment.

Sagar Srinivas Sakhinana, Venkataramana Runkana

RePair method improves cross-modal retrieval accuracy on Flickr30K and COCO30K by repairing retrieval errors.

Siyi Liu, Xiaorong Zhu, Enjun Du et al.

3D-MRL enhances zero-shot 3D shape recognition accuracy to 50.9% using nested representation learning.

Márcus Lobo, Vitor Matias, Jeová Farias et al.

RAGDiffusion++ generates high-fidelity garment images using macro-retrieval and micro-alignment, reducing FID by 14.3%.

Yuhan Li, Xianfeng Tan, Fangao Zeng et al.

Using noise-to-signal ratio (NSR) to evaluate the reliability of LLMs' cultural positioning, with 42% of pairs exceeding NSR>1, indicating high response noise and bias towards Western cultures.

An Duy Nguyen, Muhammad Aurangzeb Ahmad

RACER combines reinforcement learning and four-agent collaboration, improving KG-enhanced reasoning by about 5% on average.

Yuwei Lou, Hao Hu, Yuzhou Jiang et al.

GoM framework enhances generalization in single-image multi-view synthesis using scene-disjoint validation and targeted diffusion adaptation, ranking first.

Jie Li, Xingchen Zou, Yuxuan Liang