Understand this story
On November 9, 2026, Baidu released the automated research system ERNIE Evolve, designed to improve AI training and inference. After being given a clear research goal, the system can independently choose research directions, modify code, and run experiments; only improvements that pass independent evaluation and review are retained, and the results of each round are used to guide the next round. Baidu says the system achieved SOTA on the public benchmarks NanoChat AutoResearch and NanoGPT Speedrun; in ERNIE 5.5 Flash inference tests, deterministic GPU inference throughput increased by 33.7%, and single-request decoding speed increased by 14.9%. Baidu says this is a step toward recursive self-improvement (RSI), and that the next step is to extend the method to training systems, model capabilities, and evaluation.
AI-generated from public reports
Latest development · Report
Baidu Releases ERNIE Evolve, an Automated AI Research and Inference Optimization SystemDevelopments
1 confirmed developments
- Source:English text is being prepared.
Reports and discussion timeline
Browse reports and discussion by time. Their evidence levels remain distinct.
Nov 10
Unite.AI
Baidu Releases ERNIE Evolve, an Automated AI Research and Inference Optimization System
Baidu has released ERNIE Evolve, a system that can automatically select research directions, modify code, and run experiments, retaining only improvements that pass independent evaluation and review. Baidu says that on the NanoChat five-minute training task it verified bits per byte of 0.8763, better than the publicly compared 0.887791; on NanoGPT Speedrun Track 1 it reached a 3.28 FineWeb validation loss in 38.92 seconds, versus 39.9 seconds for the comparison. The system was also used for inference optimization of ERNIE-5.5-Flash, where four rounds of research on the Kunlunxin P800 raised generation speed to 623.6% of the baseline and increased single-request GPU decoding speed by 14.9%. On SOL-ExecBench, its score of 0.819 did not exceed the comparison's leading result of 0.827.
MarkTechPostRelated discussion
MarkTechPost joined this discussion
This post contributes to heat; it is not a formal report or confirmed fact.
Nov 9
BaiduFirst-partySelected
Baidu Releases ERNIE Evolve Automated Research System, GPU Inference Throughput Improved by 33.7%
Baidu has released the ERNIE Evolve automated research system for improving AI training and inference: given a clear research goal, it selects research directions, modifies code, and runs experiments; only improvements that pass independent evaluation and review are retained, and the results of each round guide the next round. On public benchmarks, the system achieved SOTA on NanoChat AutoResearch and NanoGPT Speedrun; in ERNIE 5.5 Flash inference tests, deterministic GPU inference throughput improved by 33.7%, and single-request decoding speed improved by 14.9%. Baidu calls this a step toward recursive self-improvement (RSI), and plans next to extend the method to training systems, model capabilities, and evaluation.
Heat trend for this story
- Current
- 18
- Period peak
- 25 Nov 10 06:00
- 24-hour change
- —
Move the pointer or select the chart to inspect hourly heat. Use the left and right arrow keys with a keyboard.