RNAChallenge dataset

A challenging benchmarking dataset of 27,283 transcripts (16,243 coding; 11,040 non-coding), enriched for short coding sequences (sORFs) and used to evaluate transcript coding potential prediction tools.

Source