AI agents came up with the entire solution and wrote all of the solution code. What I (the human) did was set the direction of the work, hand work off between agents, and make submissions and record their scores.
| Item | Details |
|---|---|
| Models | Claude Opus 5.5 (claude-opus-5-5, effort max), OpenAI gpt-6.1-sol / gpt-6-astra (reasoning max) |
| Harnesses | Claude Code and Codex CLI |
| /// Computes single-source unorthodox shortest paths with respect to the given shortest-path tree `sp`. | |
| /// | |
| /// `is_consistent[u][i]` is true iff the `i`-th edge of `u` is consistent. | |
| /// O(m log n) time. | |
| pub fn sssup(g: &Vec<Vec<(usize, i64)>>, sp: SSSP, is_consistent: &Vec<Vec<bool>>) -> Vec<i64> { | |
| let mut dist = vec![i64::max_value(); g.len()]; | |
| let mut uf = UnionFind::new(g.len()); | |
| let mut que = std::collections::BinaryHeap::new(); | |
| for u in 0..g.len() { | |
| for i in 0..g[u].len() { |
| use io; | |
| function input() { | |
| local infile = io.openRead(infile); | |
| R = infile.readInt(); | |
| C = infile.readInt(); | |
| F = infile.readInt(); | |
| N = infile.readInt(); | |
| B = infile.readInt(); | |
| T = infile.readInt(); |
| package lib.graph; | |
| import static java.lang.Math.*; | |
| import java.util.*; | |
| public class MinCostCirculation { | |
| // 最小費用循環流を求める | |
| // O(n^2 m log(nC))だが超高速 | |
| // 一つの頂点に接続する辺の容量の和がオーバーフローしないようにすること |