togethercomputer

togethercomputer / erdos-minimum-overlap

Public

New State-of-the-Art on Erdős' Minimum Overlap Problem

79
9
100% credibility
Found Mar 08, 2026 at 79 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Jupyter Notebook
AI Summary

This repository shares numerical data for step function constructions that achieve a new record upper bound on Erdős' minimum overlap problem using AI optimization.

How It Works

1
🔍 Stumble upon a math breakthrough

You hear about a new record in solving a famous puzzle called Erdős' minimum overlap problem while browsing math news or GitHub.

2
🎉 See the exciting new best result

You read how AI agents created a step function that gives the tightest upper bound yet, beating previous records from experts and other AIs.

3
📖 Learn the puzzle simply

The page explains the problem with pictures and a table comparing all the top solutions over the years, making it easy to grasp why this matters.

4
📊 Compare the achievements

You check the results table showing steps, dates, and bounds, spotting how this one edges out the competition with more precise steps.

5
🔬 Explore the solution data

You peek at the ready-to-use number lists from older methods and the new AI-improved one, ready for checking or plotting.

🏆 Appreciate the math progress

You feel thrilled knowing math puzzles are getting solved better thanks to clever AI, with everything here to verify and share.

Sign up to see the full architecture

4 more

Sign Up Free

Star Growth

See how this repo grew from 79 to 79 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is erdos-minimum-overlap?

This Jupyter Notebook repo tackles the Erdős minimum overlap problem—a 1955 combinatorics challenge on partitioning sets with minimal shifts—by delivering explicit step function constructions for the tightest known upper bound on constant C5 (0.380871). AI agents optimize these via sequential linear programming, starting from prior state-of-the-art like Haugland and TTT-Discover. Users get numpy arrays of h-values for direct import, plus visuals and verification to compute overlaps in their own notebooks.

Why is it gaining traction?

It nudges past state-of-the-art GitHub records from AlphaEvolve (0.380924) and TTT-Discover (0.380876), proving AI's punch in discovering reinforcement learning algorithms for math puzzles. Devs dig the ready comparisons and finer 600-step granularity versus older 51-step baselines. The hook: plug-and-play data for benchmarking your own Erdős solvers.

Who should use this?

Combinatorics researchers tightening bounds on Erdős problems, AI engineers testing sequential optimization on open math challenges, or grad students in analysis verifying minimum overlap via Jupyter experiments.

Verdict

Worth forking for state-of-the-art Erdős minimum overlap benchmarks, despite 79 stars and 1.0% credibility score signaling early maturity—README docs shine, but add your tests before production math pipelines.

(178 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.