Contributor

• Your GitHub username: @HanxunH

Model

• Model name: Qwen 3.5 397B
• Arena rank (if listed): 26
• Provider: Qwen Chat

Evidence (at least one required)

Option A: Web conversation link
Paste a public share link from any LLM web interface (ChatGPT, Claude, Gemini, Grok, Kimi, etc.):

• Link: https://chat.qwen.ai/s/f4faf33a-a6b3-4503-8c9b-6d57ee39c0c6?fev=0.2.16

Option B: Notebook
Upload or link to a Jupyter Notebook (.ipynb) showing the ISC trigger:

• Notebook link or file:

Option C: API log
Paste the prompt + response (or attach a JSON file):

[paste here]

What harmful content was generated?

Method

• Used an existing ISC-Bench template (templates/)
• Modified an existing template (describe changes)
• Designed a custom TVD task from scratch
• Conversational (no code, just professional task framing)
• Other (describe below)

Checklist

• Evidence is publicly accessible or attached
• No adversarial prompt was used — only a legitimate professional task
• The model generated harmful content as a functional requirement of the task
• Content would be refused if requested directly (without task framing)