NOVA Blueprint Dashboard: first winning code shows a 418% improvement the gray bars show molecule scores from a naive approach (random selection used in the industry as a benchmark), while the blue bars highlight the molecule scores from NOVA Blueprint’s first winner. each molecule’s score = binding affinity to a target protein minus average binding to a set of anti-targets. the final submission score is the average across 100 unique molecules (higher is better). even though the random search tested far more molecules (23,992 vs. 100), the blue peak sits to the right of the score distribution meaning the winner’s code consistently found better, more selective molecules, more efficiently (less attempts). the dashed lines mark each group’s average score (μ = 0.66 for benchmark, μ = 2.76 for winner), a 4.2x improvement. the goal: increase the difference between benchmark and miner average scores, across rotating targets. this means the development of smarter, more generalizable chemical search algorithms. follow our competitions and check the code submitted on our dashboard: 🔗 #Bittensor #SN68 #DeSci #DeAI #drugdiscovery