A comprehensive index of artificial intelligence and machine-learning research with AI-generated summaries, citation metrics, and direct links to papers and code.
Felix Sattler, Simon Wiedemann, Klaus‐Robert Müller, et al.
Sparse ternary compression (STC) enables communication-efficient federated learning with non-i.i.d. data by compressing both upstream and downstream updates via sparsification, ternarization, and Golomb encoding.
Md. Mostafizer Rahman, Yutaka Watanobe
Explores ChatGPT's opportunities and threats in education and research, with coding experiments and surveys on programming learning.
Chao Dong, Chen Change Loy, Kaiming He, et al.
This paper introduces SRCNN, a deep CNN that learns an end-to-end mapping for single image super-resolution, outperforming sparse-coding methods.
Aditya Bharat Soni, Boxuan Li, Xingyao Wang, et al.
OpenHands-Versa is a generalist agent using a minimal set of general tools—code editing/execution, web search, multimodal browsing, and file access—that outperforms specialized agents across SWE-Bench Multimodal, GAIA, and The Agent Company.
Richard S. Sutton
This paper presents positive results for all control tasks previously reported as failures by Boyan and Moore, using sparse-coarse-coded function approximators (CMACs) and online learning.
Bogdan Georgiev, J. Gómez-Serrano, Terence Tao, et al.
This paper demonstrates AlphaEvolve, an LLM-guided evolutionary coding agent, as a tool for autonomously discovering novel mathematical constructions, rediscovering best-known solutions and improving several across 67 problems.
Shangao Li, Yao Zhang, Volker Tresp, et al.
QuoteBench reveals that matched execution scores hide command-path failures in LLM coding agents, showing that boundary adaptation, not raw generation, separates models.
Yue Wang, Qiuzhi Liu, Jiahao Xu, et al.
Identifies 'underthinking' in o1-like LLMs, where frequent thought switching harms reasoning, and proposes a decoding penalty (TIP) to improve accuracy on math problems.
Xinyu Zhao, Zhen Tan, Vaishnav Tadiparthi, et al.
SkillComposer formalizes structured skill composition as task-conditioned skill sequence prediction, jointly predicting skill subset, count, and order via a constrained autoregressive decoder, improving coding agent pass rates by up to 23.1pp.
Sarel Weinberger, Amir Hozez
Prompt wording in coding agents changes where effort is spent—reasoning vs. tool use—without improving success, with cost multipliers up to 18x.
Kangsan Kim, Min-Bong Kang, Taeil Kim, et al.
This paper introduces Memory Transfer Learning (MTL), showing that cross-domain memory pools improve coding agent performance by 3.7% through transfer of meta-knowledge, with abstraction level dictating transferability.
Unknown
This review comprehensively evaluates protein language models and their encoding strategies for protein function prediction, aiming to help researchers understand and apply these models.