Run1 pareto frontier optimization
[COLM'26] SkillLearnBench is the first benchmark for evaluating continual learning methods that automatically generate agent skills.
npx -y skills add cxcscmu/SkillLearnBench --skill run1_pareto-frontier-optimizationAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
What its author says it does
Copied from the file, not written here
Identifying Pareto-optimal points in a multi-objective optimization space.
SKILL.md
0.7 KB, 168 tokens by cl100k_base, as published. Nobody here has run it
Pareto Logic
A solution $(F1_a, \Delta_a)$ dominates $(F1_b, \Delta_b)$ if:
- $(F1_a \ge F1_b)$ and $(\Delta_a \le \Delta_b)$
- At least one inequality is strict.
To extract the Pareto frontier from a results list:
- Initialize an empty list
pareto_points. - For each candidate $A$:
- Check if any other candidate $B$ exists such that $B$ dominates $A$.
- If no $B$ dominates $A$, $A$ is on the frontier.
Efficiency Tip: For large search spaces, sort by $F1$ descending, then iterate through to keep only those that show a decrease in $\Delta$.