Blatant AI slop just won a 25k USD DeepMind Kaggle Grand Prize
Points and comments are a snapshot, not live.
A $25k DeepMind Kaggle prize winner faces accusations of flawed methodology and misinterpreted data.
Google DeepMind's $25,000 Kaggle grand prize in a competition to benchmark AGI cognitive abilities was awarded to "MEDLEY-BENCH." A losing participant, Thomas Werkmeister, publicly contested the win, accusing the submission of poor SDK usage, unsubstantiated claims, and major logical inconsistencies in its writeup. Werkmeister provided specific examples, including allegedly misread graphs and contradictory findings, arguing the winning entry failed the competition's stated criteria of quality, defensibility, and clarity.
What commenters are saying
The thread is highly critical, centering on the irony and quality issues of AI slop winning a prestigious prize. A dominant sentiment is that Arxiv and Kaggle are now used for self-promotion over substance. Commenters link this to broader concerns about AI judging AI, slop-filled resumes, and the degradation of research quality. Specific criticism notes the winning entry's flawed methodology and the potential for AI-as-judge to create a closed feedback loop of mediocrity.