agentsclimarketplace

Pareto optimization

Skill cxcscmu/SkillLearnBench/skills/b1-one-shot-gemini-3.1-pro-preview/dbscan-parameter-tuning/pareto-optimization

[COLM'26] SkillLearnBench is the first benchmark for evaluating continual learning methods that automatically generate agent skills.

Install
npx -y skills add cxcscmu/SkillLearnBench --skill pareto-optimization

Assembled 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

Algorithm to identify Pareto-optimal points from a set of multi-objective evaluations.

SKILL.md

1.4 KB, 319 tokens by cl100k_base, as published. Nobody here has run it

Pareto Optimization

When evaluating multiple conflicting objectives (e.g., maximize F1 score, minimize distance), a point is Pareto optimal if no other point is better in all objectives.

Usage

import numpy as np

def is_pareto_efficient(costs, maximize=[True, False]):
    ""\"
    Find the pareto-efficient points
    costs: An (n_points, n_costs) array
    maximize: Boolean array indicating if the corresponding objective should be maximized
    ""\"
    is_efficient = np.ones(costs.shape[0], dtype=bool)
    
    # Adjust costs so we can simply look for points that are strictly "less than or equal"
    # For objectives we want to maximize, we negate them.
    adjusted_costs = np.copy(costs)
    for i, max_obj in enumerate(maximize):
        if max_obj:
            adjusted_costs[:, i] = -adjusted_costs[:, i]
            
    for i, c in enumerate(adjusted_costs):
        if is_efficient[i]:
            # Keep any point with a lower cost or not strictly dominated
            # A point is strictly dominated if it's >= in all dimensions and > in at least one
            is_efficient[is_efficient] = np.any(adjusted_costs[is_efficient] < c, axis=1) | np.all(adjusted_costs[is_efficient] == c, axis=1)
            is_efficient[i] = True  # And keep self
    return is_efficient

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.