LiveCodeBench Pro — Übersicht

LiveCodeBench Pro ist ein kontaminationsfreier Benchmark zur Bewertung von LLMs im Bereich Competitive Programming und quasi-Nachfolger von Live Code Bench v6. Der LCB-Pro Benchmark enthält 584 Aufgaben aus Coding-Competitions wie Codeforces, ICPC und IOI. Die Aufgaben werden in Echtzeit gesammelt, bevor Lösungen oder Editorials online erscheinen und die Trainingsdaten kontaminieren könnten. Ein Team internationaler Olympiade-Medaillengewinner annotiert jedes Problem hinsichtlich algorithmischer Kategorien und kognitiver Anforderungsprofile (Wissen, Logik, Beobachtung) und führt zeilenweise Fehleranalysen fehlgeschlagener Modell-Einreichungen durch.

Paper Leaderboard GitHub Dataset NeurIPS 2025

Beispielaufgaben aus dem LiveCodeBench Pro Benchmark

Die folgenden Beispielaufgaben zeigen typische Fragestellungen, die im LiveCodeBench Pro Benchmark vorkommen.

Codeforces 1983D - Swap Dilemma: You are given two arrays a and b, both of length n, consisting of distinct positive integers. In one move, you can choose indices l and r (l ≤ r) and swap a_l and a_r, then choose indices p and q (p ≤ q) in b such that r−l = q−p and swap b_p and b_q. Determine if it is possible to make both arrays the same using any number of such operations.

Check that a and b are permutations of each other (same multiset). Then observe the operation is equivalent to swapping adjacent elements in both arrays simultaneously. Count inversions in both permutations: arrays can be made equal if and only if the parity of the number of inversions needed is the same. Output 'YES' or 'NO'.

Codeforces 626F - Group Projects: There are n students in a class working on group projects. The i-th student takes a_i minutes to finish. Students divide into groups; the imbalance of a group is max(a_i) - min(a_i) in that group (0 for single-student groups). Count the number of ways to divide students into groups so that the total imbalance is at most k. Output the answer modulo 10^9 + 7.

Sort the array a. Use DP with state dp[i][j][s] where i is the number of students placed, j is the number of 'open' groups (groups with only one element so far), and s is the current total imbalance. Transition: student i+1 can start a new group (j increases by 1, imbalance decreases by a[i+1]) or join an existing open group (j decreases by 1, imbalance increases by a[i+1]). Optimize space with rolling arrays.

Codeforces 1704F - Colouring Game: Alice and Bob play a game on a 1×n strip of cells, each initially white, red, or blue. On each turn, a player must color exactly two adjacent cells — one red and one blue (in either order). A player who cannot move loses. Given the initial coloring, determine the winner assuming both play optimally.

Model the strip as a combinatorial game. Segment the strip into independent intervals separated by already-colored cells. Compute the Sprague-Grundy value for each interval based on the pattern of available moves, then XOR all Grundy values together. If the XOR is non-zero, the first player wins; otherwise the second player wins.