@IntologyAI: Can coding agents do research? We release NanoGPT-Bench, an internal eval we’ve used to test agents on an AI R&D proble…
Summary
IntologyAI releases NanoGPT-Bench, an internal benchmark to evaluate coding agents on AI R&D tasks. Current agents recover only 9.3% of human progress, mostly through hyperparameter tuning, highlighting gaps in algorithmic research capabilities.
View Cached Full Text
Cached at: 05/20/26, 06:25 AM
Can coding agents do research?
We release NanoGPT-Bench, an internal eval we’ve used to test agents on an AI R&D problem with months of human progress
Codex, Claude Code, Autoresearch recover only 9.3% of human progress, mostly tuning hyperparams & ignoring algorithmic research
NanoGPT-Bench is built on the NanoGPT Speedrun, a popular LLM pretraining competition to minimize the training time of a GPT-2 style model. Existing human submissions constitute nearly 2 years of work. To control for dependencies and contamination in frontier models, we standardize evaluation to a 5-month window of world records. Evaluation is fully autonomous and end-to-end, with no human intervention or internet access.
Similar Articles
@OkhayIea: Everyone's racing to build "AI scientists." So we asked a blunt question: Can today's best coding agents beat the publi…
Introduces NatureBench, a cross-disciplinary benchmark of 90 tasks from Nature papers to test AI coding agents, finding the best agent (Claude Opus 4.7) surpasses SOTA on only 17.8% of tasks and often succeeds by reducing science to supervised ML rather than genuine discovery.
@OpenAI: We’re introducing GeneBench-Pro, a research-level benchmark for a harder kind of AI progress: how well agents can navig…
OpenAI introduces GeneBench-Pro, a research-level benchmark to test AI agents' ability to navigate messy biological data, choose analysis paths, and make judgment calls in computational biology.
PaperBench: Evaluating AI’s Ability to Replicate AI Research
OpenAI introduces PaperBench, a benchmark evaluating AI agents' ability to replicate state-of-the-art AI research by replicating 20 ICML 2024 papers with 8,316 gradable tasks. The best-performing model (Claude 3.5 Sonnet) achieves only 21% replication score, below human PhD-level performance, highlighting current limitations in autonomous research capabilities.
Introducing GeneBench-Pro
OpenAI introduces GeneBench-Pro, a research-level benchmark designed to test AI agents' ability to perform judgment-heavy analyses in computational biology, covering genomics, quantitative biology, and translational medicine.
@Lyubh22: Coding benchmarks are saturating. AI4Research is the next frontier. Thrilled to see our MLS-Bench (https://mls-bench.co…
Announcing MLS-Bench, the first AI4Research benchmark to gain broad community adoption, testing AI agents on 140 executable tasks across 12 domains to propose modular ML improvements. The post includes leaderboard scores for models like Claude Opus 4.6 and GPT-5.4.