ModelLens learns a performance-aware latent space from leaderboard interactions to recommend models without running them, outperforming metadata-based baselines on 162K records.

Rui Cai, Weijie Jacky Mo, Xiaofei Wen et al.

STABLE uses RLPR to achieve rollback-free stable brick generation, improving inference speed by 94%.

Chenhui Xu, Ziyue Bai, Fuxun Yu et al.

MIND uses sliced Wasserstein for efficient, robust evaluation of generative models, requiring fewer samples and being faster than FID.

Quentin Berthet, Yu-Han Wu, Clement Crepy et al.

AOPD replaces negative reinforcement with local divergence minimization, reducing variance and exploration issues in on-policy distillation, boosting math reasoning by 4.09/8.34 points.

Nan Jia, Haojin Yang, Xing Ma et al.

Proposes SelectivelyRM, an optimal transport-based framework for robust reward modeling from noisy preferences, outperforming SOTA by 5-10% in error metrics.

Licheng Pan, Haochen Yang, Haoxuan Li et al.

DiBA (Diagonal and Binary Matrix Approximation) compresses neural network weights by decomposing matrices into three diagonal and two binary factors, improving storage efficiency.

Nobutaka Ono

Proposes a linear-constrained diffusion sampling method combining Langevin mixing and information-theoretic bounds, improving image inpainting and super-resolution.

Ahmad Aghapour, Erhan Bayraktar, Asaf Cohen

Introduces Knowledge-Free Correlated Agreement (KFCA) to incentivize federated learning without ground truth.

Leon Witt, Togrul Abbasli, Kentaroh Toyoda et al.

Proposes online natural language feedback to improve language model alignment via ICL and fine-tuning, achieving up to 35% performance recovery with 50x fewer samples.

Christine Ye, Joe Benton

Jordan-RoPE uses complex Jordan blocks for non-semisimple relative positional encoding, outperforming RoPE on WikiText-103.

Yaobo Zhang

Uni-OPD unifies on-policy distillation with dual strategies, boosting multi-task and multi-modal model performance.

Wenjin Hou, Shangpin Peng, Weinong Wang et al.

This study compares additive and multiplicative RNN architectures incorporating action embeddings for RL, showing multiplicative models improve efficiency and accuracy.

Matthew Schlegel, Volodymyr Tkachuk, Adam White et al.

Proposes Policy Gradient Penalty (PGP) for constrained exploration, achieving ε-approximate optimality and feasibility via quadratic regularization.

Florian Wolf, Ilyas Fatkhullin, Niao He

Proposes a distributional alignment game framework for answer-level fine-tuning, utilizing GRPO for efficient training, achieving significant performance gains.

Mehryar Mohri, Jon Schneider, Yifan Wu

PAINT integrates adaptive partial solution masking and sparse energy interpolation, boosting self-distilled reasoning performance.

Zhiquan Tan, Yinrong Hong

Proposes HDET method to improve optimization quality and generalization of large models via automatic learning rate exploration.

Hailing Cheng, Tao Huang, Chen Zhu et al.

Efficient learning by implicit exploration in bandit problems with side observations, achieving near-optimal regret guarantees.

Tomas Kocak, Gergely Neu, Michal Valko et al.

TCOD employs a progressive trajectory length schedule to stabilize on-policy distillation in multi-turn autonomous agents, boosting success rate by up to 18 points.

Jiaqi Wang, Wenhao Zhang, Weijie Shi et al.

Kolmogorov-Arnold Networks achieve universality with a single non-affine function.

Vugar Ismailov

Proposed a Collocation-based Robust Physics-Informed Neural Network (CRVPINN) for simulating pollution propagation under thermal inversion conditions on Spitsbergen.

Leszek Siwik, Maciej Sikora, Natalia Leszczyńska et al.