TS-Arena employs a live forecasting protocol with pre-registration to prevent data leakage, enabling real-time evaluation of TSFMs with dynamic rankings.

Marcel Meyer, Sascha Kaltenpoth, Henrik Albers et al.

YOTO framework combines differentiable subset selection and multi-task learning for efficient gene selection and prediction in single-cell transcriptomics.

Daphné Chopard, Jorge da Silva Gonçalves, Irene Cannistraci et al.

Diffusion Maps Kernel Ridge Regression (DM-KRR) enables long-term prediction of complex dynamical systems, outperforming state-of-the-art methods with higher accuracy and data efficiency.

Jiwoo Song, Daning Huang, John Harlim

DTDR leverages dynamic tool dependency modeling, boosting function calling success by 23%-104% through context-aware retrieval.

Bhrij Patel, Davide Belli, Amir Jalalirad et al.

Posterior Behavioral Cloning (PostBC) models demonstrator behavior's posterior, improving coverage and RL finetuning efficiency.

Andrew Wagenmaker, Perry Dong, Raymond Tsao et al.

SoliReward enhances video generation reward models' robustness using single-item binary annotations and Hierarchical Progressive Query Attention.

Jiesong Lian, Ruizhe Zhong, Zixiang Zhou et al.

Proposes On-policy Expert Corrections (OEC) to mitigate covariate shift in multi-turn LM training, improving software engineering task success rates by 13-14%.

Niklas Lauffer, Xiang Deng, Srivatsa Kundurthy et al.

ParaFormer integrates PageRank-enhanced attention to mitigate over-smoothing in deep graph transformers, achieving superior results on large-scale datasets.

Chaohao Yuan, Zhenjie Song, Ercan Engin Kuruoglu et al.

FLAME combines Legendre Memory and normalization flow for lightweight, probabilistic time series forecasting with state-of-the-art zero-shot performance.

Xingjian Wu, Hanyin Cheng, Xiangfei Qiu et al.

Proposes KHRONOS, a kernel-based neural surrogate for multi-fidelity aerodynamic prediction, reducing parameters by 94% and accelerating training/inference.

Apurba Sarker, Reza T. Batley, Darshan Sarojini et al.

SEMDICE is an off-policy state entropy maximization method using stationary distribution correction, enabling learning from arbitrary datasets with theoretical guarantees.

Jongmin Lee, Meiqi Sun, Pieter Abbeel

Integrates conformal prediction into multi-armed bandits, ensuring finite-sample coverage and reward efficiency under weak arm separability.

Simone Cuonzo, Nina Deliu

Proposed Gaussian Quant (GQ), a training-free vector quantization method, outperforms VQGAN on ImageNet.

Tongda Xu, Wendi Zheng, Jiajun He et al.

Proposed a randomized greedy compression algorithm, proving the existence of pruned/quantized subnetworks in wide MLPs.

Houssam El Cheairi, David Gamarnik, Rahul Mazumder

GraphDeT introduces edge denoising as an auxiliary task, boosting GNN domain adaptation performance by up to 21.83%.

Haiyang Yu, Meng-Chieh Lee, Xiang song et al.

Proposes Soft Q-based Diffusion Fine-Tuning (SQDF), integrating KL regularization to enhance target reward and diversity in text-to-image generation.

Hyeongyu Kang, Jaewoo Lee, Woocheol Shin et al.

Proposes a training-free activation-space whitening method for policy violation detection, achieving 86.0% F1 score, outperforming fine-tuned baselines by up to 9.1 points.

Oren Rachmil, Avishag Shapira, Roy Betser et al.

ASCIIBench evaluates multimodal models' spatial reasoning via ASCII art, revealing significant representation bottlenecks.

Kerry Luo, Michael Fu, Joshua Peguero et al.

Models Transformer attention as an interacting particle system; analyzes mean-field limits revealing multi-cluster formation and phase transitions.

Philippe Rigollet

Flow matching (FM) with paths (OT/VP) outperforms diffusion models in tabular data synthesis, achieving high utility with low function evaluations.

Bahrul Ilmi Nasution, Floor Eijkelboom, Mark Elliot et al.