CRISP enhances VSSD for remote sensing segmentation via DCO and OMP, achieving ~30M parameters with significant accuracy gains.

Kangning Wang, Haopeng Zhang, Zhiguo Jiang

This study critically evaluates automatic creativity metrics and LLM-based judgments, revealing weak correlation with human assessments and systemic biases, highlighting evaluation limitations.

Alessandro Tutone, Giorgio Franceschelli, Mirco Musolesi

ReWorld uses split training and pose-indexed landmark bank to balance long-term memory and real-time control, achieving state-of-the-art fidelity.

Zhifei Chen, Luozhou Wang, Guibao Shen et al.

Introduces SWE Refactor Bench with a three-stage evaluation for long-horizon repository migration; only 5.4% success rate across 520 runs, highlighting current model limitations.

Deyao Hong, Yizhe Chi, Wenyi Li et al.

Expert-grounded distillation creates an 8B-parameter vision-language model for scalable road safety audits, outperforming larger teachers and proprietary models with 81% accuracy.

Md Thamed Bin Zaman Chowdhury, Moazzem Hossain

Introduces an adaptive sampling method using leave-one-out denoiser, reducing sampling complexity.

Daniil Dmitriev, Zhihan Huang, Yuting Wei

Prime Agent employs recursive language models and persistent tools to achieve 95.5% on ARC-AGI-3, enabling long-horizon autonomous evaluation and multi-agent coordination.

Seth Karten, Alex L. Zhang, Kevin Thomas et al.

ConvergeFlow uses flow matching in continuous embedding space to provably converge to valid token embeddings, enabling direct token prediction without a cross-entropy decoder.

Na Li, Yuchen Jiao, Changxiao Cai et al.

Using Bayesian latent ability models, the study finds independent reasoning boosts skill development, while frequent AI help can cause dependency and performance decline.

Shang Wu, Catarina G Belem, Shuyuan Fu et al.

Proposed adapter-based few-shot continual learning framework with SSL pretraining and LoRA, boosting malware packet recognition accuracy.

Kyle Stein, Guillermo Francia, III Eman El-Sheikh et al.

RVM fine-tunes diffusion velocity fields without trajectories, matching or exceeding policy-gradient baselines at lower cost; no numerical scores are provided.

Jaemoo Choi, Wei Guo, Yuchen Zhu et al.

ActPair framework combines action-aligned retrieval and pairwise multimodal reranking, significantly improving text-based person anomaly search accuracy.

Thanh-Khoi Nguyen, Thanh-Nhan Vo, Trong-Thuan Nguyen et al.

Proposes Safety-Direction Penalty (SDP) based on representation geometry to mitigate reasoning-induced safety misalignment, restoring safety in Qwen2.5-3B and 7B models.

Yipeng Zhao, Qishun Yang, Shenzhe Zhu et al.

GeoWAM predicts future scene geometry with point clouds, significantly improving autonomous driving trajectory planning.

Yiren Lu, Xin Ye, Jiaming Liu et al.

Proposes Primal–Dual neural framework for timely classification, balancing sensitivity, specificity, and monitoring costs with guarantees.

Jiaming Qiu, Yingye Zheng, Ying-Qi Zhao

Proposes INDI, a method that distills behavior-level intent into vision-language-action models, boosting success rates by over 20% on benchmarks.

Sangoh Lee, Sangwoo Mo, Wook-Shin Han

ProxyFormer uses a proxy architecture for ultra-long context and high-resolution generation, extending sequence length and retrieval accuracy.

Zhongpan Tang

Reward-free continual learning with DreamerV3-based latent world models enables space robots to adapt after hardware failures without external rewards.

Andrej Orsula, Miguel Olivares-Mendez, Carol Martinez

MomADv2 employs selective state-space memory and flow-matching residual correction, significantly improving long-horizon autonomous driving stability and safety.

Ziying Song, Shengkai Zhang, Lin Liu et al.

Proposes a claim-based cross-lingual biography enrichment framework leveraging non-English Wikipedia, achieving 75.17 F1 in claim extraction and 86.2% claim coverage.

Yifei Song, Ziyang Chen, Emil Sayilov et al.