mailitics

Tag: alignment

  • Alignment of Diffusion Model and Flow Matching for Text-to-Image Generation

    Alignment of Diffusion Model and Flow Matching for Text-to-Image Generation arXiv:2602.00413v1 Announce Type: new Abstract: Diffusion models and flow matching have demonstrated remarkable success in text-to-image generation. While many existing alignment methods primarily focus on fine-tuning pre-trained generative models to maximize a given reward function, these approaches require extensive computational resources and may not generalize…

  • Why AI Alignment Starts With Better Evaluation

    Why AI Alignment Starts With Better Evaluation You can’t align what you don’t evaluate The post Why AI Alignment Starts With Better Evaluation appeared first on Towards Data Science. Hailey Quach Go to original source

  • Risk Phase Transitions in Spiked Regression: Alignment Driven Benign and Catastrophic Overfitting

    Risk Phase Transitions in Spiked Regression: Alignment Driven Benign and Catastrophic Overfitting arXiv:2510.01414v1 Announce Type: new Abstract: This paper analyzes the generalization error of minimum-norm interpolating solutions in linear regression using spiked covariance data models. The paper characterizes how varying spike strengths and target-spike alignments can affect risk, especially in overparameterized settings. The study presents…

  • Any-Step Density Ratio Estimation via Interval-Annealed Secant Alignment

    Any-Step Density Ratio Estimation via Interval-Annealed Secant Alignment arXiv:2509.04852v1 Announce Type: new Abstract: Estimating density ratios is a fundamental problem in machine learning, but existing methods often trade off accuracy for efficiency. We propose textit{Interval-annealed Secant Alignment Density Ratio Estimation (ISA-DRE)}, a framework that enables accurate, any-step estimation without numerical integration. Instead of modeling infinitesimal…

  • Many of Your DPOs are Secretly One: Attempting Unification Through Mutual Information

    Many of Your DPOs are Secretly One: Attempting Unification Through Mutual Information arXiv:2501.01544v1 Announce Type: cross Abstract: Post-alignment of large language models (LLMs) is critical in improving their utility, safety, and alignment with human intentions. Direct preference optimisation (DPO) has become one of the most widely used algorithms for achieving this alignment, given its ability…

  • On Robust Cross Domain Alignment

    On Robust Cross Domain Alignment arXiv:2412.15861v1 Announce Type: new Abstract: The Gromov-Wasserstein (GW) distance is an effective measure of alignment between distributions supported on distinct ambient spaces. Calculating essentially the mutual departure from isometry, it has found vast usage in domain translation and network analysis. It has long been shown to be vulnerable to contamination…