vh/heretic
1
0
Fork 0
mirror of https://github.com/p-e-w/heretic.git synced 2026-09-27 14:31:26 -07:00
Code Issues Packages Projects Releases Wiki Activity

fix: various cleanups and improvements for the reproducibility system

Browse Source
This commit is contained in:
Philipp Emanuel Weidmann
2026-04-15 11:13:58 +05:30
parent e2c74bfb3c
commit b46396b785
5 changed files with 150 additions and 128 deletions
Download Patch File Download Diff File Expand all files Collapse all files
config.default.toml
-4
View File
@@ -144,7 +144,6 @@ split = "train[:400]"
column = "text"
residual_plot_label = '"Harmless" prompts'
residual_plot_color = "royalblue"
commit = ""
# Dataset of prompts that tend to result in refusals (used for calculating refusal directions).
[bad_prompts]
@@ -153,18 +152,15 @@ split = "train[:400]"
column = "text"
residual_plot_label = '"Harmful" prompts'
residual_plot_color = "darkorange"
commit = ""
# Dataset of prompts that tend to not result in refusals (used for evaluating model performance).
[good_evaluation_prompts]
dataset = "mlabonne/harmless_alpaca"
split = "test[:100]"
column = "text"
commit = ""
# Dataset of prompts that tend to result in refusals (used for evaluating model performance).
[bad_evaluation_prompts]
dataset = "mlabonne/harmful_behaviors"
split = "test[:100]"
column = "text"
commit = ""