Of course it was exposed - not sure it's explicit or not. Why wouldn't HackerNews comments be part of the training data? And Simon's blog and the many discussions about Pelicans? It'd be hard to miss. Doesn't mean Anthropic has made this an explicit goal in training.
replyIt would be genuinely shocking at this point if any of the frontier models weren't well exposed to the problem.
replySee here for more discussion of that: https://news.ycombinator.com/item?id=49803892#49804881
replyJust want to say: you’re such a legend, please do not stop sharing your pelicans, it’s always fun to see how they change over the months :)
replyLets start frog riding motorcycle trend until they frogmaxx, or cat driving convertible.
replyThe model recognizing the task doesn't mean it was benchmaxxed (RLVR-trained) to solve it. It might simply recognize it from pre-training on Internet text.
reply