synthetic benchmarks
Artificial datasets created for evaluating the performance of AI algorithms under controlled conditions, offering insights into generalization capabilities without the noise of real-world data.
- Constrained Sampling for Language Models Should Be Easy: An MCMC Perspective
- Deciphering the Extremes: A Novel Approach for Pathological Long-tailed Recognition in Scientific Discovery
- FAME: Adaptive Functional Attention with Expert Routing for Function-on-Function Regression
- Flatten Graphs as Sequences: Transformers are Scalable Graph Generators
- Length Generalization via Auxiliary Tasks
- Masked Diffusion Models as Energy Minimization
- Neural Mutual Information Estimation with Vector Copulas
- Optimizing the Unknown: Black Box Bayesian Optimization with Energy-Based Model and Reinforcement Learning
- PaTH Attention: Position Encoding via Accumulating Householder Transformations
- SONAR: Long-Range Graph Propagation Through Information Waves
- Sample-efficient Learning of Concepts with Theoretical Guarantees: from Data to Concepts without Interventions
- Split Gibbs Discrete Diffusion Posterior Sampling
- Thought Communication in Multiagent Collaboration
- Toward Interpretable Evaluation Measures for Time Series Segmentation
- Train on Pins and Test on Obstacles for Rectilinear Steiner Minimum Tree