importance sampling
A statistical technique used to estimate properties of a distribution while minimizing variance. In AI, it is often applied in reinforcement learning and probabilistic models to improve estimation accuracy.
- Active Measurement: Efficient Estimation at Scale
- Amortized Sampling with Transferable Normalizing Flows
- Breaking the Frozen Subspace: Importance Sampling for Low-Rank Optimization in LLM Pretraining
- GVPO: Group Variance Policy Optimization for Large Language Model Post-Training
- Leveraging Importance Sampling to Detach Alignment Modules from Large Language Models
- MRO: Enhancing Reasoning in Diffusion Language Models via Multi-Reward Optimization
- Model Selection for Off-policy Evaluation: New Algorithms and Experimental Protocol
- Tapered Off-Policy REINFORCE - Stable and efficient reinforcement learning for large language models