ML-Embed integrates 3D-ML framework with MEL, MLL, MRL for efficient, inclusive multilingual embeddings; achieves SOTA on 9/17 MTEB benchmarks.

Ziyin Zhang, Zihan Liao, Hang Yu et al.

TAB-VLM tests 600 questions on 1,600 Indian artifacts; GPT-5.2 leads, yet reaches only 58.7% overall accuracy.

Mukul Ranjan, Prince Jha, Khushboo Kumari et al.

Proposed Nonlinear Bipolar Compensation method improves low-bit quantized model accuracy by 4.6% on average.

Peilin Sun, Jianxin Wu

Proposes a self-adaptive interleaved reasoning framework with hierarchical data pipeline, boosting multimodal generation fidelity and efficiency.

Qingyang Liu, Bingjie Gao, Canmiao Fu et al.

EponaV2 employs future depth and semantic prediction with flow matching to enhance autonomous driving planning, outperforming perception-free models with +1.3PDMS and +5.5EPDMS improvements.

Jiawei Xu, Zhizhou Zhong, Zhijian Shu et al.

Introduces SciPaths, a dataset with 262 expert-annotated pathways, using ML models like GPT-4 to predict scientific dependencies with F1=0.189.

Eric Chamoun, Yizhou Chi, Yulong Chen et al.

TOPOS framework uses TOPOS-VAE and TOPOS-DiT for high-fidelity 3D head generation, outperforming existing methods.

Bojun Xiong, Zoubin Bi, Xinghui Peng et al.

DiffPhD accelerates elastodynamics of heterogeneous materials via GPU, achieving 10x speedup.

Shih-Yu Lai, Sung-Han Tien, Jui-I Huang et al.

EvoLib constructs an evolving knowledge library via self-supervised abstraction extraction, boosting performance in mathematical reasoning and code generation without parameter updates.

Weijia Xu, Alessandro Sordoni, Chandan Singh et al.

CDD method enhances RAG model accuracy under knowledge conflict, especially in entity swap and logical contradiction scenarios.

Yihang Chen, Pin Qian, Su Wang et al.

Proposes a neighborhood-based Neural ODE training framework combining Jacobian accuracy and statistical properties, improving chaotic system modeling.

Joon-Hyuk Ko, Andrus Giraldo, Deok-Sun Lee

MARS employs a three-tier memory architecture with adaptive lifecycle management, achieving 26.4% HR@1 improvement in recommendation tasks.

Xiang Shen, Yuhang Zhou, Yifan Wu et al.

BBCritic achieves GUI semantic alignment via contrastive learning, surpassing binary classification.

Yuchen Sun, Pei Fu, Shaojie Zhang et al.

CreFlow employs a compositional LTL reward model with localized correction and success-guided reflow loss, boosting sparse-reward embodied video RL success by 23.8%.

Zhenyang Ni, Yijiang Li, Ruochen Jiao et al.

Introduces Riemannian Hodge Message Passing (RHMP), achieving top performance across seven physical benchmarks.

Dongzhe Zheng, Christine Allen-Blanchette

Full Jacobian eigendecomposition reveals spectral evolution from non-normal to symmetric layers, with training inducing low-rank bottlenecks and topology-spectral coupling.

Jesseba Fernando, Grigori Guitchounts

Flow-matching loss enables multi-strategy trajectory sampling, driving autonomous skill evolution.

Mingda Zhang, Tiesunlong Shen, Haoran Luo et al.

Receding Horizon Multi-Agent Deceptive Path Planner uses Boltzmann distribution for dynamic deception.

Xubin Fang, Brian M. Sadler, Rick S. Blum

Proposes accuracy drop as a layer relevance metric, outperforming cosine similarity in large language models.

Cristian Hinostroza, Rodrigo Toro Icarte, Christ Devia et al.

PBT-Bench evaluates AI's property-based testing ability using 100 curated problems with 365 semantic bugs across 40 Python libraries, revealing significant model performance gaps.

Lucas Jing, Xinqi Wang, Liao Zhang et al.