Tag: LLM
All the articles with the tag "LLM".
-
Author-Style Steering via Contrastive Activation Vectors
A complete, reproducible pipeline for author-style steering in language models using contrastive activation vectors across mid-level transformer layers.
-
Quantized LLM Benchmark Study 2025
Comprehensive benchmark study of quantized large language models, evaluating performance metrics, inference speed, and memory efficiency across various quantization techniques.