tnt-llm-classify
Categorize conversations between a human and an AI assistant using a predefined taxonomy, ensuring a thorough analysis of the dialogue to select the most relevant topic or intent. The output includes a detailed reasoning process and the chosen category, formatted in XML tags for clarity.
Prompt Text
Your task is to use the provided taxonomy to categorize the overall topic or intent of a conversation between a human and an AI assistant.
First, here is the taxonomy to use:
<taxonomy>
{taxonomy}
</taxonomy>
To complete the task:
1. Carefully read through the entire conversation, paying attention to the key topics discussed and the apparent intents behind the human's messages.
2. Consult the taxonomy and identify the single most relevant category that best captures the overall topic or intent of the conversation.
3. Write out a chain of reasoning for why you selected that category. Explain how the category fits the content of the conversation, referencing specific statements or passages as evidence. Output this reasoning inside <reasoning></reasoning> tags.
4. Output the name of the category you chose inside <category></category> tags.
That's it! Remember, choose the single most relevant category. Don't choose multiple categories. Think it through carefully and explain your reasoning before giving your final category choice.
Assign a single category to the following content:
<content>
{content}
</content>
Respond with. your reasoning and category within XML tags. Do not include the number, just the category text.Evaluation Results
1/22/2026
Overall Score
3.37/5
Average across all 3 models
Best Performing Model
Low Confidence
anthropic:claude-3-haiku
4.78/5
anthropic:claude-3-haiku
#1 Ranked
4.78
/5.00
adh
5.0
cla
4.7
com
4.7
GPT-4o Mini
#2 Ranked
4.39
/5.00
adh
4.3
cla
4.6
com
4.3
google:gemini-1.5-flash
#3 Ranked
0.94
/5.00
adh
1.0
cla
1.2
com
0.7
Test Case:
Tags
langsmith
wfh
ChatPromptTemplate
Tagging
