TriggerBench
A benchmark dataset for evaluating AI system performance on trigger condition recognition and prospective memory tasks, providing a structured dialogue and trigger-task evaluation framework.
Timeline (last 90 days)
The researchers evaluated on a synthetic task set constructed by referencing the published structure of TriggerBench, which contains 48 blind-written dialogues and 175 tasks.
The researchers commit to conducting a formal evaluation on TriggerBench once its official data is released, as future work.
TriggerBench is a standardized benchmark dataset specifically designed for evaluating trigger-based memory tasks, published by an independent party.
All Facts (3)
The researchers evaluated on a synthetic task set constructed by referencing the published structure of TriggerBench, which contains 48 blind-written dialogues and 175 tasks.
50%UnverifiedTriggerBench is a standardized benchmark dataset specifically designed for evaluating trigger-based memory tasks, published by an independent party.
50%UnverifiedThe researchers commit to conducting a formal evaluation on TriggerBench once its official data is released, as future work.
50%