Skip to content

Fully automatic censorship removal for language models

feat: add config.nohumor.toml (#340)

UnstableLlamab79aa717c671
diff --git a/config.nohumor.toml b/config.nohumor.toml
new file mode 100644
index 0000000..e3f51b3
--- /dev/null
+++ b/config.nohumor.toml
@@ -0,0 +1,69 @@
+# Rename this file to config.toml, place it in the working directory
+# that you run Heretic from, and edit the configuration to your liking.
+
+max_response_length = 300
+
+residual_plot_title = "PaCMAP Projection of Residuals for Serious/Humorous Prompts"
+
+refusal_markers = [
+ "πŸ˜…",
+ "here's one",
+ "why did",
+ "walked into a bar",
+ "gets worse",
+ "the punchline",
+ "classic",
+ "dad joke",
+ "bad joke",
+ "pun intended",
+ "ba dum tss",
+ "rimshot",
+ "πŸ˜‚",
+ "πŸ˜„",
+ "πŸ˜†",
+ "haha",
+ "hehe",
+ "lol",
+ "funny",
+ "joke",
+ "humor",
+ "that's hilarious",
+ "you could say",
+ "one-liner",
+ "comedian",
+ "stand-up",
+ "unexpectedly",
+ "because apparently",
+ "to be fair",
+ "on the bright side",
+ "lmao",
+ "omg",
+ "rofl",
+ "silly",
+ "humorous",
+ "clever",
+]
+
+[good_prompts]
+dataset = "mlabonne/harmless_alpaca"
+split = "train[:400]"
+column = "text"
+residual_plot_label = "Serious prompts"
+residual_plot_color = "royalblue"
+
+[bad_prompts]
+dataset = "UnstableLlama/jokes"
+split = "train[:200]"
+column = "text"
+residual_plot_label = "Humorous prompts"
+residual_plot_color = "darkorange"
+
+[good_evaluation_prompts]
+dataset = "mlabonne/harmless_alpaca"
+split = "test[:100]"
+column = "text"
+
+[bad_evaluation_prompts]
+dataset = "UnstableLlama/jokes"
+split = "train[200:250]"
+column = "text"