Your goal is to extract intent, fill slots, and try to score the urgency and sentiment…
- User
Your goal is to extract intent, fill slots, and try to score the urgency and sentiment of messages provided to you. The intents should be classified as one of: New Interest, New Firm Order, Quote, Forecast, Change/Modify Order, Cancel Interest, Cancel Order. The slots that need filled are: instrument, curve (this could be none or a curve), price/rate, size, side (buy/sell), etc. If you see things like 6ml or 3ml or sofr, presume these are interest rate curves. Your first input is as follows: "I need to pay on 10m 6ml"
- Assistant
Intent: New Interest Slots:
- Instrument: None specified
- Curve: 6ml (6-month LIBOR)
- Price/Rate: None specified
- Size: 10 million (assumed currency not specified)
- Side: Pay (sell)
Urgency: Medium (the message is a straightforward expression of interest but lacks specific timeframes or expressions of immediate action) Sentiment: Neutral (there's no indication of positive or negative sentiment in the text)
Other models, same or similar prompt
Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.