A comprehensive index of artificial intelligence and machine-learning research with AI-generated summaries, citation metrics, and direct links to papers and code.
Atindra Jha, Margaret Li, Jure Leskovec, et al.
Mixture-of-Experts models overfit more to repeated training data than dense models, with degradation increasing with sparsity; masking-based regularization can mitigate but not eliminate the gap.
Arian Hosseini, Alessandro Sordoni, Daniel Toyama, et al.
This paper reveals a significant reasoning gap in LLMs when solving compositional math problems, showing that performance on standard benchmarks masks systematic differences in reasoning abilities.
Wenrui Bao, Zhiben Chen, Dan Xu, et al.
Learn2PD trains a lightweight filter to adaptively unmask tokens in diffusion LLMs, achieving up to 22.58x speedup without performance loss.
Giang Nguyen, Štefan Dlugolinský, Martin Bobák, et al.
This survey provides a comprehensive overview and comparison of Machine Learning and Deep Learning frameworks and libraries for large-scale data mining, highlighting trends and massive parallelism support in the era of Big Data.
David Silver, Thomas Hubert, Julian Schrittwieser, et al.
AlphaZero is a single reinforcement learning algorithm that, starting from random play and given only game rules, achieves superhuman performance in chess, shogi, and Go through self-play.
Saman Marandi, Yu-Shu Hu, Mohammad Modarres
This paper presents a framework for automated construction of Dynamic Master Logic models as knowledge graphs from technical documentation using retrieval-augmented large language models, validated on a nuclear reactor system.
Theresia Veronika Rampisela, Henrik Palmer Olsen, Giovanni Colavizza
LegalPincite is a large-scale legal IR dataset from CJEU judgments with masked queries, full corpora, and multi-level ground-truth citations to address data leakage and oversimplified retrieval settings.
Unknown
This paper introduces a unified framework for model editing, including the Equality-constrained Mass Model Editing (EMMET) algorithm for Transformers, which improves the memorization objective of editing methods like ROME.
Ilya Mikhelson
This paper introduces the Socratic Test, an automated conversational assessment that uses dynamic scaffolding and multimodal workspaces to map a student's Zone of Proximal Development, replacing deficit-based grading with an additive, mastery-oriente
Unknown
CoMAS is a novel framework for co-evolving multi-agent systems where agents autonomously improve by learning from interaction rewards.
Siyan Zhao, Devaansh Gupta, Qinqing Zheng, et al.
Proposes d1, a framework adapting pre-trained masked diffusion LLMs into reasoning models via masked SFT and a novel critic-free RL algorithm diffu-GRPO.
Jiayuan Liu, Tianqi Li, Shiyi Du, et al.
Expanding context windows in LLM agents systematically degrades cooperation in multi-agent social dilemmas, a phenomenon termed the memory curse.