Baidu Unveils ERNIE Evolve Auto-Research System, Boosts GPU Inference Throughput 33.7%
Baidu_Inc · x · 2026-11-09
Baidu announced ERNIE Evolve, an automated research system for improving AI training and inference. Given a research goal, it selects directions, modifies code and runs experiments; only improvements passing independent evaluation and review are kept, with each round informing the next — a step toward recursive self-improvement (RSI).
- SOTA on NanoChat AutoResearch and both NanoGPT Speedrun tracks (38.9s vs public best 39.9s on Track 1)
- 33.7% higher deterministic GPU inference throughput; 14.9% faster single-request decoding
- Architecture: Director/Executor/Evaluator/Reviewer evolve loop across N parallel lanes
Next: extending the approach to training systems, model capabilities and evaluation.
Related event: Baidu Unveils ERNIE Evolve, an Automated AI Research System(2 posts)→
More from Companies & People
- 5 free math textbooks that cover the core math behind machine learning, exercises included — cneuralnetwork · 2026-11-09
- Five free math textbooks that cover the math behind ML, with exercises included — cneuralnetwork · 2026-11-09
- To master math, stop memorizing formulas and explain concepts until a child gets it — TivadarDanka · 2026-11-09
- Polymarket gives Tesla just 7% odds of launching California robotaxis by year-end — Polymarket · 2026-11-09
- Sakana AI denies IPO plans: goal is building Japan's national AI champion — SakanaAILabs · 2026-11-09
- ELLIS x UniReps talk with Adrien Doerig and Jerry Tang set for Nov 10 on Zoom — ClementineDomi6 · 2026-11-09