LLM Inference

Scalable oversight via confidence-guided reasoning; enables safe self-correction matching parallel decoding accuracy at orders-of-magnitude lower compute.

A test-time LLM ensembling strategy that dynamically adapts to balance performance, diversity, and consistency, achieving state-of-the-art results.

Training-free decoding that mitigates LLM hallucinations by contrasting a base model with a masked-retrieval variant, boosting summarisation by up to 18.6%.