rejection sampling
PulseAugur coverage of rejection sampling — every cluster mentioning rejection sampling across labs, papers, and developer communities, ranked by signal.
-
New method learns Random Geometric Graphs in probabilistic metric spaces
Researchers have developed a novel method for learning Random Geometric Graphs (RGGs) within probabilistic metric spaces. This technique is applicable to diverse datasets regardless of their observable types, distributi…
-
New theory explains how verifier imperfection impacts LLM test-time scaling
A new paper titled "ROC-n-reroll: How verifier imperfection affects test-time scaling" explores the theoretical underpinnings of improving language model performance through additional compute during inference. The rese…
-
AI alignment techniques detailed: from SFT to ensemble methods
The author outlines various techniques for training AI systems to be aligned and behave ethically. These methods involve leveraging internal model states or external outputs as reward signals, adjusting training data di…
-
New VEGAS metric aligns video captions with viewer attention
Researchers have introduced VEGAS (Video caption Evaluation via GAze Score), a novel metric designed to improve video captioning by aligning generated text with individual viewer attention. Unlike traditional methods, V…
-
New Selective Importance Sampling method improves LLM alignment
Researchers have introduced Selective Importance Sampling (SIS), a novel plug-in method designed to enhance the alignment of large language models (LLMs) during reinforcement learning post-training. This approach addres…
-
New research disentangles synthetic data scaling methods
A new research paper from arXiv explores two methods for scaling synthetic data generation: Source Expansion (SE) and Fixed-Source Synthesis (FSS). The study isolates FSS by keeping the source material and teacher model…
-
New IVRS Method Enhances Bayesian Machine Learning Posterior Approximation
Researchers have introduced Implicit Variational Rejection Sampling (IVRS), a novel method designed to enhance posterior approximation in Bayesian machine learning. This technique combines implicit distributions modeled…