Extending LRNNs' eigenvalues to [-1,1] enables robust state-tracking, significantly improving performance on parity and modular tasks.

Riccardo Grazzi, Julien Siems, Arber Zela et al.

Enhancing LLaMA-3.1-70B's reasoning via Additional Logic Training (ALT) with FLD×2 corpus, achieving 30-point boost in logic benchmarks.

Terufumi Morishita, Gaku Morio, Atsuki Yamaguchi et al.

This study evaluates pre-trained visual representations (PVR) in model-based reinforcement learning (MBRL), revealing limited improvements in sample efficiency and out-of-distribution generalization.

Moritz Schneider, Robert Krug, Narunas Vaskevicius et al.

Long-context RAG enhances LLM performance, but only a few models maintain accuracy above 64k tokens.

Quinn Leng, Jacob Portes, Sam Havens et al.

Proposes a feature learning theory for single- and multi-modal contrastive learning, highlighting signal-to-noise ratio (SNR) as key to generalization.

Wei Huang, Andi Han, Yongqiang Chen et al.

This paper introduces a nonlinear transformer framework with MLP layers that efficiently learns low-dimensional target functions in-context, requiring prompt length dependent only on the target function's subspace dimension r, not the ambient d.

Kazusato Oko, Yujin Song, Taiji Suzuki et al.

TAMA uses large multimodal models for few-shot time series anomaly detection, improving accuracy.

Jiaxin Zhuang, Leon Yan, Zhenwei Zhang et al.

ProD-RL leverages reinforcement learning to hierarchically decompose proofs, improving theorem proving success rate from 40.8% to 45.5%.

Kefan Dong, Arvind Mahankali, Tengyu Ma

Achieving domain-independent certified robustness via knowledge continuity, independent of input domains.

Alan Sun, Chiyu Ma, Kenneth Ge et al.

Outer-PPO introduces non-unit learning rates and momentum, improving performance by 5-10% in complex environments.

Charlie B. Tan, Edan Toledo, Benjamin Ellis et al.

TabM employs parameter-efficient ensembling of MLPs, achieving superior tabular data deep learning performance.

Yury Gorishniy, Akim Kotelnikov, Artem Babenko

Pre-trained vision-language flow model enables zero-shot multi-robot control with high-frequency continuous actions.

Kevin Black, Noah Brown, Danny Driess et al.

Fed-POE combines local and federated models with dynamic selection for adaptive prediction and fine-tuning, achieving significant performance gains.

Pouya M. Ghari, Yanning Shen

OGBench provides a comprehensive benchmark with 8 environments and 85 datasets to evaluate offline goal-conditioned RL algorithms across multiple capabilities.

Seohong Park, Kevin Frans, Benjamin Eysenbach et al.

Using sparse autoencoders (SAEs) for interpretable knowledge unlearning in language models; limited effectiveness with current techniques.

Eoin Farrell, Yeu-Tong Lau, Arthur Conmy

MazeNet is a deep learning method for solving OARSMT with 100% accuracy.

Gabriel Díaz Ramos, Toros Arikan, Richard G. Baraniuk

Introduced Concept-Guided Conditional Diffusion and Prototype Networks to enhance model interpretability.

Alba Carballo-Castro, Sonia Laguna, Moritz Vandenhirtz et al.

LAMPS predicts API call strategies to optimize scheduling, reducing latency by 27%-85% and TTFT by 4%-96%.

Rana Shahout, Cong Liang, Shiji Xin et al.

Incorporating known group invariances into Gaussian process kernels enhances Bayesian optimization sample efficiency, with theoretical bounds and practical validation.

Theodore Brown, Alexandru Cioba, Ilija Bogunovic

Proposed a biologically motivated evaluation framework; found scVI and PCA outperform emerging models in perturbation analysis.

Ihab Bendidi, Shawn Whitfield, Kian Kenyon-Dean et al.