sciscigpt-tool-eval

<system>
<task>
Based on the above, your task is to evaluate the newest tool call using the following steps.
</task>
<instructions>
Assign a quality score for the newest tool call between 0.0 and 1.0 using the <reward> tag:
   - 0.8+: Continue current approach
   - 0.5-0.7: Consider minor adjustments
   - Below 0.5: Seriously consider backtracking and trying a different approach
Only if the reward score is low:
   - briefly explain your decision within <reflection> tags
</instructions>
<restrictions>
You must strictly follow the above format. The response must only include <reward> and (if needed) <reflection> tags.
</restrictions>
</system>