evidence lower bound
The evidence lower bound (ELBO) is a concept from variational inference that provides a lower bound on the log likelihood of the observed data. In the context of probabilistic models, optimizing the ELBO helps approximate posterior distributions of latent variables, thereby facilitating more efficient inference.
- Bayesian Ego-graph inference for Networked Multi-Agent Reinforcement Learning
- Brain-like Variational Inference
- Disentangled Cross-Modal Representation Learning with Enhanced Mutual Supervision
- EVOREFUSE: Evolutionary Prompt Optimization for Evaluation and Mitigation of LLM Over-Refusal to Pseudo-Malicious Instructions
- Flow based approach for Dynamic Temporal Causal models with non-Gaussian or Heteroscedastic Noises
- Functional Complexity-adaptive Temporal Tensor Decomposition
- Latent Space Factorization in LoRA
- Next Semantic Scale Prediction via Hierarchical Diffusion Language Models