Analyzed representation flow in SSMs and TBMs using centered kernel alignment and variance metrics, revealing layer-wise information flow differences.

Nhat M. Hoang, Do Xuan Long, Cong-Duy Nguyen et al.

Introduced Control-Augmented Autoregressive Diffusion (CADA) for data assimilation, significantly improving stability and accuracy.

Prakhar Srivastava, Farrin Marouf Sofian, Francesco Immorlano et al.

Introduced UserLMs to simulate human behavior in multi-turn dialogues, reducing GPT-4o performance from 74.6% to 57.4%.

Tarek Naous, Philippe Laban, Wei Xu et al.

LMCACHE employs batch data movement and modular APIs to enable efficient cross-engine and hierarchical KV cache management for enterprise-scale LLM inference.

Yuhan Liu, Yihua Cheng, Jiayi Yao et al.

SIGMA-GEN enables single-pass multi-subject identity-preserving image generation, achieving state-of-the-art performance using the SIGMA-SET27K dataset.

Oindrila Saha, Vojtech Krs, Radomir Mech et al.

This paper introduces AI-Driven Research for Systems (ADRS), using OpenEvolve to automate algorithm optimization, achieving up to 5.0x speedup and 50% cost reduction.

Audrey Cheng, Shu Liu, Melissa Pan et al.

SimulatorArena uses user profiles to create high-fidelity simulators, achieving ρ=0.7 correlation with human judgments, enabling scalable multi-turn evaluation.

Yao Dou, Michel Galley, Baolin Peng et al.

Combining large language models with multi-layer scene graphs enables active semantic perception, improving indoor scene understanding speed and accuracy.

Huayi Tang, Pratik Chaudhari

BlockRank method enhances ICR efficiency through structured attention; experiments show superior performance over existing methods.

Nilesh Gupta, Chong You, Srinadh Bhojanapalli et al.

Even with perfect retrieval, long context length degrades LLM performance; propose shortening context strategy to improve performance.

Yufeng Du, Minyang Tian, Srikanth Ronanki et al.

Introduced LUR and RLUR methods to enhance uncertainty detection efficiency in driver action recognition.

Koen Vellenga, H. Joe Steinhauer, Jonas Andersson et al.

Proposes DeSA, a two-stage training framework that decouples search and answer generation, significantly improving recall and accuracy metrics.

Yiding Wang, Zhepei Wei, Xinyu Zhu et al.

Derives closed-form Busemann functions in Wasserstein space, enabling efficient distribution slicing and distance computation.

Clément Bonet, Elsa Cazelles, Lucas Drumetz et al.

High-dimensional analysis of linear attention models reveals prediction decomposition, implicit regularization, and phase transitions in low-rank tasks.

Kaito Takanami, Takashi Takahashi, Yoshiyuki Kabashima

TAG amplifies tangential score components to steer diffusion sampling toward high-density regions, reducing hallucinations and improving semantic fidelity.

Hyunmin Cho, Donghoon Ahn, Susung Hong et al.

A.I.R. introduces training-free adaptive iterative reasoning for frame selection, boosting VideoQA accuracy and efficiency.

Yuanhao Zou, Shengji Jin, Andong Deng et al.

JEF-Hinter distills offline trajectories into compact, context-aware hints, boosting web agent performance by 15-20% with minimal inference overhead.

Hadi Nekoei, Aman Jaiswal, Patrice Bechard et al.

A martingale-based sequential audit framework detects token misreporting in LLMs with fewer than 70 samples, controlling false positives below 5%.

Ander Artola Velasco, Stratis Tsirtsis, Manuel Gomez-Rodriguez

ChronoEdit leverages pretrained video generative models with explicit temporal reasoning to enhance physical consistency in image editing, achieving a 4.42/5 score on PBench-Edit.

Jay Zhangjie Wu, Xuanchi Ren, Tianchang Shen et al.

PolyKAN framework uses polyhedral analysis for KAN compression, offering theoretical guarantees on model reduction and error control.

Di Zhang