The BasePeople

Vladimir Solodkin

Titles as the sources gave them. In 1 entry, with 0 checked quotations.

In the Base

Every entry they appear in.

  1. ResearcharXivWrote it

    WRDO: Weighted Refusal Direction Optimization in Aligned Language Models

    According to the abstract excerpt, the authors propose WRDO, which combines per-prompt refusal directions with learned weights, and compare it with single-direction methods on three open-weight models using three guard judges and HarmBench. The excerpt is cut off before results.