Proposes Video-in-the-Loop (ViTL), a two-stage long-video QA framework combining localization and span-aware answering, achieving up to 8.6% improvement with 50% less input.

Chendong Wang, Donglin Bai, Yifan Yang et al.

OptAgent combines multi-agent shopping simulation with genetic search, improving Etsy query fitness 21.98% over original queries and 3.36% over BoN.

Divij Handa, David Blincoe, Orson Adams et al.

Introduced HVGC framework and BridgeDiT model for text-to-sounding video generation, outperforming existing methods.

Kaisi Guan, Xihua Wang, Zhengfeng Lai et al.

Proximal Sampler with Restricted Gaussian Oracle achieves O(d² log(1/ε)) complexity for uniform sampling in convex bodies.

Thanh Dang, Jiaming Liang

TokenFlow employs preemptive scheduling and proactive KV cache management to boost streaming throughput by 82.5%, reducing P99 TTFT by 80.2%.

Junyi Chen, Chuheng Du, Renyuan Liu et al.

FSFSplatter achieves high-precision surface reconstruction with sparse views in 2 minutes, reducing error by 28.39%.

Yibin Zhao, Yihan Pan, Jun Nan et al.

Proposes a bi-level optimization framework combining hydraulic modeling, state estimation, and stealthy FDI attacks to analyze cyber-physical threats in gas pipelines.

Tejaswini Sanjay Katale, Lu Gao, Yunpeng Zhang et al.

EqM learns a time-invariant energy gradient landscape, enabling optimization-based sampling with FID 1.90, outperforming diffusion/flow models.

Runqian Wang, Yilun Du

Self-Forcing++ extends high-quality video generation to 4 minutes by sampling and distribution matching, outperforming baseline models by over 50x in length with improved fidelity.

Justin Cui, Jie Wu, Ming Li et al.

Diffusion^2 uses 3D point clouds to generate RF heatmaps with just 1.9 dB error, 27x faster.

Kyoungjun Park, Yifan Yang, Changhan Ge et al.

This study demonstrates that a standard Transformer, trained directly on Cartesian coordinates without graph priors, can achieve energy and force prediction accuracy comparable to state-of-the-art equivariant GNNs on OMol25, with faster inference.

Tobias Kreiman, Yutong Bai, Fadi Atieh et al.

Introduces Ground-Truth Alignment framework to diagnose reasoning-execution gaps in VLM mobile agents.

Lingzhong Dong, Ziqi Zhou, Shuaibo Yang et al.

Analyzed diagonal linear networks using Dynamical Mean-Field Theory to reveal initialization's impact on solutions.

Sota Nishiyama, Masaaki Imaizumi

R-AIRL uses adversarial IRL to learn reasoning rewards from expert demonstrations, improving training, inference reranking, and error localization.

Claudio Fanconi, Nicolás Astorga, Mihaela van der Schaar

Unified protocol shows synthetic data can accelerate LLM pre-training; optimal mix around 30% synthetic improves efficiency 5-10x.

Feiyang Kang, Newsha Ardalani, Michael Kuchnik et al.

Proposes innovations-representation-based TS-GPT for engineering time series, improving causal modeling and probabilistic forecasting.

Lang Tong, Xinyi Wang

The study explores the generalization of preference optimization under noisy feedback, proposing the GPO method and validating its effectiveness.

Shawn Im, Sharon Li

MAHALO framework achieves multi-objective alignment via PRM-guided decoding, enhancing performance in math reasoning and human values.

Yiran Shen, Yu Xia, Jonathan Chang et al.

ModernVBERT is a 250M-parameter vision-language encoder that outperforms larger models in document retrieval by optimizing attention masks, image resolution, and training strategies.

Paul Teiletche, Quentin Macé, Max Conti et al.

This study introduces a systematic multi-turn RL framework for large language models, emphasizing environment, reward, and policy pillars, validated across TextWorld, ALFWorld, and SWE-Gym with key improvements of up to 88%.

Ruiyi Wang, Prithviraj Ammanabrolu