lukashelff commited on
Commit ·
fbad389
1
Parent(s): 1bc3855
fix readme
Browse files
README.md
CHANGED
|
@@ -1,3 +1,19 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
# Isomorphic Perturbation Testing (IPT)
|
| 2 |
|
| 3 |
|
|
|
|
| 1 |
+
---
|
| 2 |
+
title: Isomorphic Perturbation Testing
|
| 3 |
+
emoji: 🔍
|
| 4 |
+
colorFrom: blue
|
| 5 |
+
colorTo: purple
|
| 6 |
+
sdk: gradio
|
| 7 |
+
tags:
|
| 8 |
+
- evaluate
|
| 9 |
+
- metric
|
| 10 |
+
- reward-hacking
|
| 11 |
+
- RLVR
|
| 12 |
+
- logical-reasoning
|
| 13 |
+
- ILP
|
| 14 |
+
description: "Detects reward hacking in LLMs via Isomorphic Perturbation Testing (IPT)."
|
| 15 |
+
---
|
| 16 |
+
|
| 17 |
# Isomorphic Perturbation Testing (IPT)
|
| 18 |
|
| 19 |
|