Dataset Viewer
Auto-converted to Parquet Duplicate
example_id
int64
prompt
list
completion
list
task
string
reward
float64
generation_ms
float64
scoring_ms
float64
total_ms
float64
info
dict
evaluate_tau2_task
float64
0
[ { "content": "<instructions>\nYou are a customer service agent that helps the user according to the <policy> provided below.\nIn each turn you can either:\n- Send a message to the user.\n- Make a tool call.\nYou cannot do both at the same time.\n\nTry to be helpful and always follow the policy. Always make sure...
[ { "content": "\n<think>The user wants to cancel a reservation. According to the policy, I need to:\n\n1. First obtain the user ID and reservation ID\n2. If the user doesn't know their reservation ID, I should help locate it using available tools\n3. I must also obtain the reason for cancellation (change of plan...
default
1
0
273.727894
273.727894
{ "description": { "purpose": "Testing that agent refuses to proceed with a cancellation that is not allowed even if User mentions that she had been told she didn't need insurance." }, "evaluation_criteria": { "actions": [], "communicate_info": [], "nl_assertions": [ "Agent should refuse to ...
1
README.md exists but content is empty.
Downloads last month
5