AIBUILDAI · RESEARCH
Research & case studies
Results, methods, and reproducible artifacts from autonomous AI model-building and scientific research runs.
NatureBench leaderboard · AI for science
Introducing AIBuildAI Science: #1 on NatureBench for autonomous AI-for-science model development
AIBuildAI Science is a vertical AI agent for autonomously developing high-performance AI models for scientific research. Across the 90 real-world AI-for-Science tasks of NatureBench, derived from papers published in leading Nature family journals, it surpassed published state-of-the-art methods on 25.6% of tasks and matched them on 57.8%, ranking first on the leaderboard ahead of every frontier coding agent evaluated. Every task was completed fully autonomously within a 4-hour time budget.
Read article
IRASim RT-1 trajectory-to-video benchmark · Robot world models
AIBuildAI autonomously post-trains a robot world model to near state-of-the-art video prediction in 29 hours
Given a short task brief and a robot manipulation dataset, AIBuildAI post-trained NVIDIA’s open Cosmos world model for action-conditioned video prediction, entirely without human intervention. Twenty-nine hours later, on the official RT-1 trajectory-to-video benchmark, its model reached 25.56 PSNR and 0.845 SSIM, up from the base model’s 17.75 PSNR and 0.716 SSIM, with the highest SSIM reported to date and PSNR ranking fourth among published results.
Read article
IOAI
AIBuildAI achieves first place at the International Olympiad in AI (IOAI)
Across all six IOAI 2025 individual-contest tasks, AIBuildAI achieved a total score of 568.49 out of 600, ranking first on the competition leaderboard. Given only the problem statements and provided data, it autonomously designed, implemented, and optimized every solution within a six-hour budget, with no human intervention.
Read article
NatureBench case study · Cancer genomics
AIBuildAI autonomously develops a state-of-the-art model for cancer gene identification in just four hours
In just four hours, AIBuildAI autonomously discovered a new state-of-the-art AI model for cancer gene identification. Without human intervention, it designed, implemented, trained, and optimized a graph neural network-based model that achieved a mean AUPRC of 0.774, outperforming the published state-of-the-art method, TREE (Nature Biomedical Engineering), which achieved 0.716, and surpassing it on 6 of 8 biological networks. The entire solution is fully open source and reproducible.
Read article