mirror of
https://github.com/wassname/llm-moral-foundations2.git
synced 2026-08-11 11:20:53 +08:00
19 lines
413 B
Markdown
19 lines
413 B
Markdown
Unbiased Assessment of LLM Moral Foundations: Controlling for Positional Effects and Response Steering
|
|
|
|
Difference from previous work
|
|
- control for positional bias
|
|
- use mechinterp representation steering
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Links:
|
|
- https://github.com/peterkirgis/llm-moral-foundations
|
|
|
|
|
|
|
|
TODO
|
|
- [ ] add amoral reasoning steering https://huggingface.co/soob3123/amoral-qwen3-14B
|