Spaces:
Sleeping
Sleeping
Add YAML metadata to Blog.md
Browse files
Blog.md
CHANGED
|
@@ -1,3 +1,22 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
# Who Audits the AI? Building an Adversarial Oversight Agent for Clinical Trials
|
| 2 |
|
| 3 |
**TL;DR**: Medical AI hallucinates fake protocol amendments, cites fabricated studies, and confidently clears patients who should never have been treated. We built SynthAudit.Env — a multi-agent environment where one AI generates these deceptive medical errors and another AI learns to catch them through reinforcement learning. 200 steps of GRPO training produced a 283% improvement in error detection, with the agent learning full ReAct reasoning chains from scratch.
|
|
|
|
| 1 |
+
---
|
| 2 |
+
title: "Who Audits the AI? Building an Adversarial Oversight Agent for Clinical Trials"
|
| 3 |
+
emoji: 🩺
|
| 4 |
+
colorFrom: indigo
|
| 5 |
+
colorTo: emerald
|
| 6 |
+
tags:
|
| 7 |
+
- openenv
|
| 8 |
+
- grpo
|
| 9 |
+
- clinical-trial
|
| 10 |
+
- reinforcement-learning
|
| 11 |
+
- multi-agent
|
| 12 |
+
- tool-calling
|
| 13 |
+
- pytorch
|
| 14 |
+
- medical-ai
|
| 15 |
+
- ai-safety
|
| 16 |
+
pinned: true
|
| 17 |
+
license: apache-2.0
|
| 18 |
+
---
|
| 19 |
+
|
| 20 |
# Who Audits the AI? Building an Adversarial Oversight Agent for Clinical Trials
|
| 21 |
|
| 22 |
**TL;DR**: Medical AI hallucinates fake protocol amendments, cites fabricated studies, and confidently clears patients who should never have been treated. We built SynthAudit.Env — a multi-agent environment where one AI generates these deceptive medical errors and another AI learns to catch them through reinforcement learning. 200 steps of GRPO training produced a 283% improvement in error detection, with the agent learning full ReAct reasoning chains from scratch.
|