Obinka Research / Questions
The questions behind the work
We want to understand when a software system can improve its own methods, how to evaluate that improvement, and where the findings are useful. These questions describe the direction we want to explore.
Question 01
Can the improver improve?
What we want to answer Does changing an agent's own tools, harness or search strategy improve its ability to make later improvements?
What we want to examine
- The component changed
- The feedback available to the system
- The comparison with a fixed method
- Performance on later tasks
Question 02
How should improvement be tested?
What we want to answer How can evaluation remain informative as a system adapts to it?
What we want to examine
- Task separation
- Feedback exposure
- Candidate selection
- Failure cases
- Transfer to tasks that did not guide the search
Question 03
Which gains are useful in practice?
What we want to answer Which methods produce useful gains within a stated budget and set of requirements?
What we want to examine
- Task quality
- Cost and time
- Failure rates
- The human work needed to reach an acceptable outcome
Question 04
What transfers into a product?
What we want to answer Which findings remain useful when applied to a product task?
What we want to examine
- Clear acceptance conditions
- The simpler alternative
- Additional implementation work
- The limits of the comparison
A question worth working on together
Tell us where your work overlaps, what you would like to investigate, and what you could bring to a focused study.