Two of HackMIT’s top eight finalists were disqualified last month after it was discovered that they had “misrepresented” what they accomplished, according to a HackMIT blog post. One of the teams, a group of three MIT sophomores calling themselves Seamless, originally came in second and were awarded $3,000. The other team, AgileAssault, did not make the top three, but received a $1,000 prize for placing in the top eight.
Seamless presented code that could allegedly animate still images and smooth the transition between looping videos, basing their work on a Microsoft paper published last year. The team consisted of Kevin Kwok ’17, Nikhil Buduma ’17, and Guillermo Webster ’17. The contestants claimed to have implemented an algorithm that was originally the result of the collaboration of multiple people at Microsoft Research and the University of Illinois at Urbana-Champaign.
HackMIT, which with 900 participants was among the largest hackathons in the country this year, was prompted to investigate the possibility that Seamless had cheated when other participants pointed out that the demo videos they presented as their results were identical to those published by Microsoft Research.
The competition operates under an “honor code” according to its organizers — participants are not allowed to work on projects before the hackathon starts and must state any borrowed code. “When an issue comes up like this, HackMIT’s policy is to give the team the benefit of the doubt,” HackMIT Director Katie Siegel ’16 said in an interview with The Tech. “We sit down and talk it over with them, and try to get a better sense of what is going on. That being said, we also don’t want to reward people who misrepresent their accomplishments.”
“They did implement everything they said, but going over it there was some confusion as to what kind of results they actually got,” said Siegel. “It was difficult to see if their implementation was correct, so they agreed to voluntarily step down.” Her statement seemed at odds with a HackMIT blog post that said the organization “disqualified” two projects.
The videos presented onstage were taken from the Microsoft researchers’ webpage, Siegel said, though she added that the Seamless team only “used them in the context of showing the power of the algorithm.”
However, in a recording of their final HackMIT presentation, which is available online, the Seamless team members are seen indicating multiple times that the looping videos were generated by their own implementation of Microsoft’s algorithm.
During their presentation, Buduma stated, “Now our algorithm then operates… to remove that cut to make that, again, a seamless loop. The way this works is we identify regions of various periodicities, based on the methods that were described in that Microsoft paper. We are able to reconstruct that using the graph-cutting algorithm to make this seamless product.” The seamless product that he refers to is Microsoft’s looping video of drummers.
At another point in their presentation, Buduma again refers to “our algorithm,” while displaying Microsoft’s looping video of a dancer.
In addition to winning second place, Seamless was also awarded a sponsor prize from Rough Draft Ventures for “creating ‘ooh’ inspiring moving images,” according to the firm’s Twitter account. They were invited to pitch at Rough Draft Ventures for the opportunity of up to $25,000 in funding, but this prize was revoked in light of their disqualification.
However, Siegel still maintains that Seamless wasn’t explicitly found guilty of any cheating. “Basically, Seamless chose to step down rather than going through the process of proving their code worked,” Siegel said. “I want to make it very clear that they weren’t incriminated or found guilty of anything.”
According to the copyright notice on Microsoft’s online publication of their “Automated video looping” research paper, “Permission to make digital or hard copies of part or all of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page.” However, Seamless seemed to present Microsoft’s videos as their own, without explicitly citing the Microsoft paper as their source.
The Seamless team members did not respond to requests for comment.
According to Siegel, the second disqualified team, AgileAssault, “used a project that was presented at another hackathon a week before. They wrote different code but the idea behind it was the same. In a discussion about their results afterwards it was decided that the two projects were too similar, and they agreed to step down.”
AgileAssault was created by Matthew Duran, a student at the University of Maryland, Baltimore County. During his final presentation, he described his project as an application that provided a “gamified approach to learning how to program and practicing programming through real-time competition with your peers.” AgileAssault had been submitted to HackUMBC the week before, where, according to Duran’s LinkedIn page, it had placed in the top 10.
HackMIT was notified that Duran’s project had previously been presented when they were contacted by HackUMBC organizers.
After watching the presentations of the top eight teams, judges and audience members determined the top three finalists by vote. To help prevent future misrepresentations, HackMIT plans to focus on improving their judging model.
In the same blog post that announced the disqualifications, Aneesh Agrawal ’17 wrote, “We’ll be revamping our judging model for next year and keeping special note of potential code of conduct violations.”
“This year, we realized that ‘the rules’ aren’t entirely clear to everyone, which leads to subjective interpretation and honest mistakes,” said Richard Z. Ni ’15, an organizer for HackMIT. “Next year, we plan to clearly communicate what’s expected of participants and exactly what constitutes a valid entry.”