A comprehensive index of artificial intelligence and machine-learning research with AI-generated summaries, citation metrics, and direct links to papers and code.
Yan Zhou, Yue Ouyang, Kaiyang Zheng, et al.
CoBa introduces compute-balanced routing for cost-effective test-time scaling, dynamically allocating reasoning compute under wall-clock and serving constraints.
Canbin Huang, Tianyuan Shi, Yuhua Zhu, et al.
Lookahead routing predicts latent representations of potential model outputs to guide LLM selection, improving routing decisions without full inference.
Adam Fisch, Shubhendu Trivedi, Fantine Huot, et al.
This paper formalizes AI model routing with costly value estimation as Pandora's Box, introducing closed-form policies that decide when refining value estimates is worth the cost.
Unknown
This paper provides a closer look into mixture-of-experts (MoE) in large language models, analyzing their architecture, routing mechanisms, and performance trade-offs.
Danyang Zhang, Junhao Song, Ziqian Bi, et al.
A comprehensive review of Mixture-of-Experts (MoE) in large language models, covering architecture, routing, meta-learning, and applications.
Unknown
This paper presents a methodology for scaling diffusion transformers to 16 billion parameters, including design choices for expert routing algorithms.
Huiyuan Tian, Bonan Xu, Shijian Li
This paper introduces the Expert Subspace Separation Index (ESSI) and controlled experiments to show that sparse MoE routing selects token-relevant experts from overlapping subspaces, a pattern termed 'coherent overlap' that preserves functional valu
Unknown
This paper proposes expert choice routing for mixture-of-experts, where each expert selects top tokens, improving load balance and training efficiency.
Unknown
This paper proposes an adaptive routing framework that dynamically selects between multiple vision-language models for computer use agents to optimize cost and performance.
Unknown
Masrouter introduces a learned routing model that selects the most cost-effective LLM for each query in multi-agent systems, reducing costs while maintaining performance.
Zihan Li, Feiyang Liu, Dandan Shan, et al.
OPERA is a multi-agent ensemble framework that treats expert weight assignment as offline policy learning to enable deployable biomedical AI without retraining.
Ahmed Heakl, Martin Gubri, Salman Khan, et al.
Dr. LLM retrofits frozen LLMs with lightweight routers trained via MCTS to skip, execute, or repeat layers, improving accuracy and efficiency without altering base weights.