Tag
Proposes a 'lift' method for training input-convex neural networks (ICNNs) that uses an unconstrained hypernetwork to emit non-negative inter-layer weights, softening the loss landscape and escaping gradient attenuation, achieving lower test loss than projected gradient descent and softplus reparametrization.
This paper analyzes the confidence calibration of 11 popular LLMs, finding that they are generally overconfident, especially on hard tasks, and underconfident on easy tasks. It introduces LifeEval, a test for evaluating calibration across difficulty levels.
SurvivalPFN is a prior-data fitted network that amortizes Bayesian inference for survival analysis via in-context learning, achieving strong predictive performance across 61 datasets without task-specific training or hyperparameter tuning.
This paper introduces DiMS, a dynamical system sampler that guarantees exact sampling from the submanifold of minimum loss solutions in neural networks, enabling better uncertainty quantification in Bayesian inference.
BaLoRA introduces a Bayesian extension to Low-Rank Adaptation (LoRA) that provides calibrated uncertainty estimates and improves prediction accuracy by narrowing the gap with full fine-tuning.
DyStruct is a training-free Bayesian decoding framework for discrete Diffusion Language Models that enables flexible-length generation by dynamically determining expansion size and decoding order, improving accuracy on math and code tasks.
This paper presents a Bayesian inverse problem framework for rain field reconstruction using Commercial Microwave Links and Diffusion Model priors, demonstrating improved accuracy over existing baselines.