# Collin Burns

Collin Burns is an American speedcuber and AI alignment researcher who set the world record for solving the Rubik's Cube in 5.25 seconds in 2015, and later co-developed the MMLU benchmark for evaluating large language models.

Collin Burns is an American speedcuber and artificial intelligence researcher. He first gained international recognition as a teenager for his exceptional skill in solving the Rubik's Cube, setting a world record in 2015. Burns later transitioned to a career in AI alignment, contributing to the development of evaluation benchmarks for large language models.

## Speedcubing career

Burns's competitive speedcubing career peaked in the mid-2010s. In 2014, he set a North American continental record of 5.93 seconds for the 3×3×3 cube. That same year, at the US Nationals 2014 speedcubing competition, he defeated the then-world champion, Feliks Zemdegs, marking a significant upset in the sport.

On April 25, 2015, Burns achieved his most notable feat: a world record single solve of 5.25 seconds. The solve occurred at a competition held at Central Bucks West High School in Doylestown, Pennsylvania. He was 15 years old at the time and used a YuXin brand 3×3×3 cube. His time was 0.3 seconds faster than the previous world record of 5.55 seconds, set by Mats Valk. Burns's record stood for approximately seven months, until Lucas Etter solved the cube in 4.90 seconds on November 21, 2015.

## Transition to AI research

Following his speedcubing career, Burns shifted his focus to academic and research pursuits in artificial intelligence. He became involved in the field of AI alignment, which studies how to ensure that advanced AI systems behave in accordance with human intentions and values.

In 2020, Burns collaborated with Dan Hendrycks and other researchers to develop the Massive Multitask Language Understanding (MMLU) benchmark. MMLU is a dataset designed to evaluate the knowledge and reasoning capabilities of large language models across a wide range of subjects, from humanities to STEM fields. The benchmark has since become a standard tool for assessing the performance of models developed by organizations such as [openai](https://www.wikiprompt.org/wiki/openai), [anthropic](https://www.wikiprompt.org/wiki/anthropic), and [google-deepmind](https://www.wikiprompt.org/wiki/google-deepmind).

## Contributions to AI evaluation

The MMLU benchmark, co-created by Burns, addresses the need for robust evaluation methods in the rapidly advancing field of [large language models](https://www.wikiprompt.org/wiki/large-language-model). It tests models on tasks that require both factual knowledge and multi-step reasoning, providing a more comprehensive measure of capability than earlier benchmarks. The development of MMLU has influenced subsequent research on model evaluation and safety, including efforts in [RLHF](https://www.wikiprompt.org/wiki/rlaif) and other alignment techniques.

Burns's work sits at the intersection of [machine learning](https://www.wikiprompt.org/wiki/machine-learning) and AI safety, a field that has grown in prominence as models become more powerful. His contributions are part of a broader movement within the AI community to ensure that systems are not only capable but also aligned with human goals.

## Personal life and legacy

Burns's dual career highlights a rare combination of skills: world-class dexterity in speedcubing and analytical rigor in AI research. His speedcubing record remains a notable milestone in the sport's history, and his later work on MMLU has had a lasting impact on how AI models are evaluated.

As of the mid-2020s, Burns continues to be active in AI alignment research, contributing to ongoing efforts to make advanced AI systems safe and beneficial. His trajectory from competitive cubing to AI research serves as an example of how diverse talents can converge in the service of addressing complex technological challenges.

## External links

- [Wikipedia: Collin Burns](https://en.wikipedia.org/wiki/Collin_Burns)

---
Source: https://www.wikiprompt.org/wiki/collin-burns
License: CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/)
Last updated: 2026-09-14T04:25:52.215645+00:00