A comprehensive index of artificial intelligence and machine-learning research with AI-generated summaries, citation metrics, and direct links to papers and code.
Unknown
Introduces RAS, a training-free sampling strategy for Diffusion Transformers that dynamically adjusts token updates based on region-adaptive criteria.
Jakub Konecný, H. B. McMahan, Felix X. Yu, et al.
This paper introduces strategies to improve communication efficiency in federated learning, including structured updates and compressed updates, reducing communication costs while maintaining model accuracy.
Unknown
This paper proposes a method to evaluate LLMs with 200x less data while maintaining accuracy, reducing evaluation costs.
Unknown
Rocketeval introduces an automated LLM evaluation method using grading checklists to reduce cost and improve efficiency.
Unknown
This paper surveys the emerging field of machine unlearning, which aims to remove the influence of specific training data from trained models to comply with data deletion requests.
Unknown
This paper derives neural scaling laws from the data distribution, predicting scaling regimes and suggesting empirical tests on toy datasets.
Dan Zhang, Tao Feng, Lilong Xue, et al.
This survey provides a comprehensive overview of Parameter-Efficient Fine-Tuning (PEFT) methods for foundation models, categorizing and analyzing approaches to reduce computational costs while maintaining performance.
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
This paper proposes a method to factorize spatio-temporal foundation models to reduce computational costs while maintaining zero/few-shot generalization performance.
Unknown
This paper proposes methods to reduce overthinking in large reasoning models, cutting computational costs while maintaining performance.
Unknown
This paper proposes an economical communication pipeline for LLM-based multi-agent systems, using a graph-based description of reasoning to reduce token costs while maintaining performance.
B. Sheth
Proposes a learning-based framework for personalized information filtering agents, implemented as 'Newt', using relevance feedback and genetic algorithms to adapt to user interests.