wassname/eliciting_suppressed_knowledge
1
0
Fork 0
mirror of https://github.com/wassname/eliciting_suppressed_knowledge.git synced 2026-09-11 12:12:04 +08:00
Code Issues Packages Projects Releases Wiki Activity

wip

Browse Source
This commit is contained in:
wassname
2025-03-12 07:35:34 +08:00
parent fca49fbac9
commit 29d79b855c
1 changed files with 3 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
README.md
+3 -1
View File
@@ -1,8 +1,10 @@
# Eliciting Suppressed Knowledge (ESK)
# Eliciting Suppressed Knowledge (ESK) WIP
## Abstract
Transformer models possess knowledge they actively suppress during inference. By isolating and probing these suppressed activation, we demonstrate small improvements on TruthfulQA compared to standard activation probes. This confirms that supressed activations are a more useful source of knowledge than the model's direct outputs, or hidden states.
*This is a Work In Progress. While we have promising results, they should be improved by simple changes to the probing method. We are currently working on these improvements.*
## Background
Recent mechanistic interpretability research identifies competing neural dynamics in transformers: