Preprint2025
Improving reasoning performance in large language models via representation engineering
Unknown
This paper proposes a representation engineering approach that reads LLM activations from the residual stream during reasoning tasks to improve performance.
0Jan 1, 2025Large Language ModelsReasoning