Baidu’s ERNIE Evolve reports gains in AI code optimization
Baidu has released ERNIE Evolve, an automated research system designed to improve AI training and inference code. Given a repository, an objective and evaluation criteria, it selects research directions, modifies code and runs experiments. A hidden evaluator and an independent reviewer must approve a change before it is retained. Baidu has not disclosed which language models power the agents, and the system itself has not been released.
In public tests, ERNIE Evolve achieved 0.8763 validation BPB on NanoChat AutoResearch, reached the target loss in 38.9 seconds on NanoGPT Speedrun Track 1, and completed Track 3 in 2,590 steps. It scored 0.819 mean SOL on SOL-ExecBench, below the 0.827 leaderboard result. Baidu said the system exceeded public bests on three of four reported metrics.
The company also applied the system to ERNIE-5.5-Flash. After four research rounds, generation speed on Kunlunxin P800 reached 623.6% of the baseline. Other reported gains included 33.7% higher deterministic throughput and a 14.9% increase in single-request decoding speed, with accuracy unchanged. The results and logs are available, while the system remains closed.
Written by Sakhanda Wire AI from reporting by MarkTechPost. Facts, quotes and full context are in the original article below.