Decomposition-based causal discovery (DCD) separates trend, seasonal, residual components, improving causal inference in non-stationary time series by 20-30% over baselines.

Muhammad Hasan Ferdous, Md Osman Gani

ContextEvolve uses three agents to compress optimization context, improving ADRS performance by up to 33.3% while cutting tokens by 29.0%.

Hongyuan Su, Yu Zheng, Yong Li

Proposes Stackelberg game-based reward shaping to optimize LLM alignment, boosting user utility by over 66% win-tie rate.

Haichuan Wang, Tao Lin, Lingkai Kong et al.

Proposes DC-CoT with RL training to identify and execute parallel subtasks, reducing longest path length by 35-40%.

Arvind Mahankali, Kaiyue Wen, Tengyu Ma

Understanding generalization via embedding dimension and distributional convergence using Wasserstein distance and Lipschitz constants.

Junjie Yu, Zhuoli Ouyang, Haotian Deng et al.

The paper proves static black-box evaluation cannot guarantee post-update alignment; larger models hide more latent biases, activated by single gradient steps.

Yavuz Bakman, Duygu Nur Yaldiz, Eleni Triantafillou et al.

Mechanistic framework models catastrophic forgetting as feature encoding transformations, validated via geometric analysis and Vision Transformer experiments.

Sergi Masip, Gido M. van de Ven, Javier Ferrando et al.

Proposes STARS, a Riemannian optimization-based inference-time method to enhance diversity by maximizing activation space volume with orthogonal steering directions.

Dongxuan Zhu, Ly Tran Ho Khanh, Andy Yat-Ming Cheung et al.

CORDS maps discrete objects to continuous fields, enabling exact decoding of variable-sized sets with high accuracy.

Tin Hadži Veljković, Erik Bekkers, Michael Tiemann et al.

Introduces Distributional Active Inference (DAIF), integrating active inference into distributional RL for model-free, efficient control in complex environments.

Abdullah Akgül, Gulcin Baykal, Manuel Haußmann et al.

Introduced a contextual runtime monitor framework, significantly enhancing safety in autonomous driving.

Alejandro Luque-Cerpa, Mengyuan Wang, Emil Carlsson et al.

Study shows effective dimension of representation geometry strongly predicts deep neural network generalization.

Sumit Yadav

Introduces Self-Distillation Fine-Tuning (SDFT), enabling continual learning with reduced catastrophic forgetting and improved new-task accuracy.

Idan Shenfeld, Mehul Damani, Jonas Hübotter et al.

Keel integrates Highway connections into Post-LayerNorm, enabling stable training of Transformer models over 1000 layers, outperforming Pre-LN in stability and expressivity.

Chen Chen, Lai Wei

ULEE combines unsupervised pretraining, self-generated goals, and meta-learning to enhance exploration and adaptation, outperforming baselines.

Octavio Pappalardo

Proposes OPSD, a self-distillation method where a single large model acts as both teacher and student, using known correct answers to improve reasoning efficiency.

Siyan Zhao, Zhihui Xie, Mengchen Liu et al.

K-Myriad maximizes collective state entropy across multiple policies, enhancing exploration in high-dimensional continuous RL tasks.

Vincenzo De Paola, Mirco Mutti, Riccardo Zamboni et al.

TriPlay-RL uses tri-role self-play reinforcement learning for LLM safety alignment, improving adversarial effectiveness by 20%-50%.

Zhewen Tan, Wenhan Yu, Jianfeng Si et al.

Using polytope decomposition, enhance ReLU network approximation, especially near singular points.

ZeYu Li, ShiJun Zhang, TieYong Zeng et al.

Proposes ATLAS, a latent-space transformer framework combining diffusion, stochastic interpolants, and CRPS for state-of-the-art probabilistic medium-range weather forecasting, outperforming ECMWF IFS.

Jean Kossaifi, Nikola Kovachki, Morteza Mardani et al.