User Satisfaction Classifier

A ModernBERT-based classifier that detects user satisfaction and dissatisfaction reasons from follow-up messages in conversational AI systems.

Key Insight

Follow-up messages alone contain sufficient signal for classification. No conversation context needed—just pass the user's response directly.

Labels

Label Description Example
SAT User is satisfied "Thanks!", "Perfect", "Great!"
NEED_CLARIFICATION User needs more explanation "What do you mean?", "Can you explain?"
WRONG_ANSWER System provided incorrect information "No, that's wrong", "That's not right"
WANT_DIFFERENT User wants alternative options "Show me others", "What else?"

License

Apache 2.0

Downloads last month
15
Safetensors
Model size
0.1B params
Tensor type
F32
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for llm-semantic-router/feedback-detector

Finetuned
(984)
this model