Tag
This paper presents BayesPO, a Bayesian prompt optimization framework using gradient-guided discrete MCMC with parallel tempering, achieving improved accuracy on instruction-induction tasks.
Introduces Depth-Entropy Guided Sampling (DEGS), a training-free test-time method that exploits layer-wise entropy collapse in LLMs to improve reasoning without RL training, achieving competitive results with RL-posttrained models.
This paper introduces Entropy-Guided Power Sampling (EGPS), a training-free and verifier-free sampler that improves the efficiency of power sampling for enhancing base language model reasoning. EGPS achieves up to 12.6x speedup over standard Metropolis-Hastings sampling while reaching best or tied-best accuracy on benchmarks like MATH500, HumanEval, and GPQA.
This paper proposes a four-phase method for constructing causal graphs that model LLM inference processes, using counterfactual augmentation to enable stable causal discovery and provide transparent, concept-level explainability.
bde is a Python package that brings sampling-based Bayesian Deep Learning to practitioners via the MILE method, combining JAX's speed with scikit-learn's API for tabular supervised learning tasks.