VideoMLA introduces low-rank latent KV cache, reducing memory by 92.7% for minute-scale video diffusion while maintaining high quality.

Hidir Yesiltepe, Jiazhen Hu, Tuna Han Salih Meral et al.

NeuROK employs a transformer-based encoder-decoder to learn a low-dimensional latent space for 4D object dynamics, trained on large-scale geometric trajectories, bypassing predefined physical models.

Chen Geng, Guangzhao He, Yue Gao et al.

LLMSurgeon formulates data mixture diagnosis as a label-shift inverse problem, achieving 94.46% accuracy on the LLMSurgeon benchmark.

Yaxin Luo, Jiacheng Cui, Xiaohan Zhao et al.

YoCausal employs a two-level causality benchmark using real-world videos and natural reversal, evaluating 13 SOTA video diffusion models' causal understanding via RSI and CCI metrics.

You-Zhe Xie, Yu-Hsuan Li, Jie-Ying Lee et al.

Proposed VisAnomReasoner fine-tuned on VisAnomBench achieves 74.30% precision and 72.17% F1 in time-series anomaly detection, surpassing baselines by over 21 and 23 points.

Xiaona Zhou, Muntasir Wahed, Tianjiao Yu et al.

SchGen introduces a semantic-grounded code model for PCB schematic generation, achieving 82% valid circuits with 60.5% functional correctness from natural language prompts.

Qinpei Luo, Ruichun Ma, Xinyu Zhang et al.

Introduces GPIC, a 28 trillion-pixel large-scale image corpus with permissive licensing, to advance visual generative modeling.

Keshigeyan Chandrasegaran, Kyle Sargent, Suchir Agarwal et al.

HullFT employs convex reconstruction and gradient caching for efficient test-time fine-tuning, improving speed and quality tradeoff in large language models.

Alaa Khamis, Alaa Maalouf

Proposes COMPOSE, a dual-graph framework combining citation and formal theorem graphs, generating plausible future theorems with 108K training pairs and 47K future papers tested.

David Busbib, Michael Werman

Proposes MedCase-Structured, a pipeline combining LLMs and terminology validation to generate HL7 FHIR R4 clinical datasets for diagnostic reasoning, with an 82.5% success rate.

Valentina Bui Muti, Eugénie Dulout, Ziquan Fu

Combining multi-objective genetic programming with survival tree optimization, this study enhances predictive accuracy and interpretability in survival analysis, validated on two real-world datasets.

Thalea Schlender, Peter A. N. Bosman, Tanja Alderliesten

SCOPE integrates a frozen LLM with an open-set plugin classifier, achieving 91.05% open-set detection accuracy and 96.63% anomaly correction in ATC readback monitoring.

Qihan Deng, Minghua Zhang, Yang Yang et al.

NEO-ov, a fully native end-to-end vision-language model, supports multi-image and video understanding with superior fine-grained perception and spatial reasoning.

Haiwen Diao, Jiahao Wang, Penghao Wu et al.

Proposes Bidirectional Evolutionary Search (BES), combining forward candidate evolution with backward goal decomposition to enhance exploration and verification in language models.

Guowei Xu, Zhenting Qi, Huangyuan Su et al.

Proposes Calibrated Collective Oversight (CCO), integrating multiple auxiliary signals with Conformal Decision Theory for online calibration, ensuring AI behavior aligns with safety targets.

William Overman, Mohsen Bayati

VisualMem introduces a structured visual memory module integrated with text memory, achieving 95% accuracy in personal entity recall, surpassing caption-based methods by over 40%.

Viet Nguyen, Thao Nguyen, Vishal M. Patel et al.

OmniVerifier-M1 employs symbolic bounding boxes and decoupled reinforcement learning to enhance visual verification accuracy, achieving 0.68 on ViVerBench.

Xinchen Zhang, Bowei Liu, Jiale Liu et al.

Proposes CAPO, a cross-annotator preference optimization method, enabling LLMs to learn and reproduce stable individual explanation behaviors, outperforming prompting and SFT.

Beiduo Chen, Pingjun Hong, Ziyun Zhang et al.

LearnWeak framework uses a stronger reference agent to identify model weaknesses, synthesizes targeted tasks, and improves small CUAs by 11.6% on average across 8 domains.

Suji Kim, Kangsan Kim, Sung Ju Hwang

FluxMem models memory as a dynamically evolving heterogeneous graph with three stages, achieving state-of-the-art results in complex reasoning and web navigation tasks.

Jizhan Fang, Buqiang Xu, Zhixian Wang et al.