Propose an architectural change to address the high variance in importance sampling-based partition function estimation. Discuss the tradeoffs involved.

Implement bridge sampling to estimate the partition function, which can reduce variance by using auxiliary distributions to bridge between different regions of the parameter space. This approach trades off increased computational complexity for improved accuracy and reduced variance.