Showing cs.LGShow all
2 papers · 1 filter
cs.LG2026
ROAST: Rollout-based On-distribution Activation Steering Technique
Xuanbo Su, Hao Luo, Yingfang Zhang +1
Activation steering provides parameter-efficient control over large language models (LLMs) at inference time, but many methods rely on off-distribution supervision and discrete mas…
cs.LG2025
Learning to Rank Chain-of-Thought: Using a Small Model
Eric Hanchen Jiang, Haozheng Luo, Shengyuan Pang +9
Large Language Models (LLMs) struggle with reliable mathematical reasoning, and current verification methods are often computationally expensive. This paper introduces the Energy O…