Welcome to Day 2. Today's four problems move from pictures and keystrokes to lectures, language models, voices in a room, and web search. As yesterday, solve them in any order; each task is scored independently. For each task you may choose up to 2 submissions to count toward your final score — if you don't choose, your 2 best by public score are taken. Submissions are limited to one per minute per task. Good luck.
- E — Lecture Shuffle. A lecture was cut into 30-second clips and their order was lost — put the clips of each lesson back into the order they were recorded.
- F — Who Wrote This? Given a prompt and an AI-generated answer, identify which of ten anonymous models produced it — a multiclass classification problem.
- G — Direction of Arrival. From 30 seconds of four-microphone audio, work out both the direction the speaker's voice came from and a transcript of what they said.
- H — Search Engine. Build a sports search engine: for each query, rank the pages of a large corpus so the most relevant results come first.