← Search

Ly Tran Ho Khanh

2 accepted papers

2026

Exploring Diverse Generation Paths via Inference-time Stiefel Activation Steering

ICLR 2026poster

Language models often default to a narrow set of high-probability outputs, leaving their generation paths homogeneous and prone to mode collapse. Sampling-based strategies inject randomness but still struggle to guarantee diversity across multiple concurrent generation runs. We address this limitati…

Cited by 0SourceScholar
2026

Test-time Diverse Reasoning by Riemannian Activation Steering

AAAI 2026technical

Best-of-N reasoning improves the accuracy of language models in solving mathematical tasks by sampling multiple candidate solutions and then selecting the best one based on some criteria. A critical bottleneck for this strategy is the output diversity limit, which occurs when the model generates sim

Cited by 0SourcePDFScholar