# Auto Researching, not hyperparameter tuning: Convergence Analysis of 10,000 Experiments

arXiv: https://arxiv.org/abs/2603.15916 · HTML: https://mlsanity.com/paper/2603.15916/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Placed in General LLM agent frameworks and benchmarks, one of 16 directions in LLM agents and tool use.
- Builds on 31 of its 40 references that are in the map.
- First posted on arXiv 16 Mar 2026.

## Placed in

- LLM agents and tool use / [General LLM agent frameworks and benchmarks](https://mlsanity.com/agents/general-llm-agent-frameworks-and-benchmarks/)

## Builds on (most cited first, 30 of 31)

- [DINOv2: Learning Robust Visual Features without Supervision](https://mlsanity.com/paper/2304.07193/) (2023)
- [Scaling Laws for Neural Language Models](https://mlsanity.com/paper/2001.08361/) (2020)
- [Mamba: Linear-Time Sequence Modeling with Selective State Spaces](https://mlsanity.com/paper/2312.00752/) (2023)
- [SWE-bench: Can Language Models Resolve Real-World GitHub Issues?](https://mlsanity.com/paper/2310.06770/) (2023)
- [Sigmoid Loss for Language Image Pre-Training](https://mlsanity.com/paper/2303.15343/) (2023)
- [Training Compute-Optimal Large Language Models](https://mlsanity.com/paper/2203.15556/) (2022)
- [Optuna: A Next-generation Hyperparameter Optimization Framework](https://mlsanity.com/paper/1907.10902/) (2019)
- [DARTS: Differentiable Architecture Search](https://mlsanity.com/paper/1806.09055/) (2018)
- [The AI Scientist: Towards Fully Automated Open-Ended Scientific Discovery](https://mlsanity.com/paper/2408.06292/) (2024)
- [Towards Understanding Sycophancy in Language Models](https://mlsanity.com/paper/2310.13548/) (2023)
- [Large Language Models as Optimizers](https://mlsanity.com/paper/2309.03409/) (2023)
- [Efficient Neural Architecture Search via Parameter Sharing](https://mlsanity.com/paper/1802.03268/) (2018)
- [Retentive Network: A Successor to Transformer for Large Language Models](https://mlsanity.com/paper/2307.08621/) (2023)
- [Revisiting Feature Prediction for Learning Visual Representations from Video](https://mlsanity.com/paper/2404.08471/) (2024)
- [Introduction to Multi-Armed Bandits](https://mlsanity.com/paper/1904.07272/) (2019)
- [BOHB: Robust and Efficient Hyperparameter Optimization at Scale](https://mlsanity.com/paper/1807.01774/) (2018)
- [MLE-bench: Evaluating Machine Learning Agents on Machine Learning Engineering](https://mlsanity.com/paper/2410.07095/) (2024)
- [Agent Laboratory: Using LLM Agents as Research Assistants](https://mlsanity.com/paper/2501.04227/) (2025)
- [MLAgentBench: Evaluating Language Agents on Machine Learning Experimentation](https://mlsanity.com/paper/2310.03302/) (2023)
- [ReEvo: Large Language Models as Hyper-Heuristics with Reflective Evolution](https://mlsanity.com/paper/2402.01145/) (2024)
- [AIDE: AI-Driven Exploration in the Space of Code](https://mlsanity.com/paper/2502.13138/) (2025)
- [RE-Bench: Evaluating frontier AI R&D capabilities of language model agents against human experts](https://mlsanity.com/paper/2411.15114/) (2024)
- [EvoPrompting: Language Models for Code-Level Neural Architecture Search](https://mlsanity.com/paper/2302.14838/) (2023)
- [MLGym: A New Framework and Benchmark for Advancing AI Research Agents](https://mlsanity.com/paper/2502.14499/) (2025)
- [Verbalized Sampling: How to Mitigate Mode Collapse and Unlock LLM Diversity](https://mlsanity.com/paper/2510.01171/) (2025)
- [Zipformer: A faster and better encoder for automatic speech recognition](https://mlsanity.com/paper/2310.11230/) (2023)
- [Code Repair with LLMs gives an Exploration-Exploitation Tradeoff](https://mlsanity.com/paper/2405.17503/) (2024)
- [Nexar Dashcam Collision Prediction Dataset and Challenge](https://mlsanity.com/paper/2503.03848/) (2025)
- [Scaling Laws for Hyperparameter Optimization](https://mlsanity.com/paper/2302.00441/) (2023)
- [Should You Use Your Large Language Model to Explore or Exploit?](https://mlsanity.com/paper/2502.00225/) (2025)

## Built on by (most cited first, 1 of 1)

- [Workflow Closure Is Not Scientific Closure in Auto-Research Systems](https://mlsanity.com/paper/2605.26200/) (2026)

## Abstract (arXiv)

When LLM agents autonomously design ML experiments, do they perform genuine architecture search -- or do they default to hyperparameter tuning within a narrow region of the design space? We answer this question by analyzing 10,469 experiments executed by two LLM agents (Claude Opus and Gemini 2.5 Pro) across a combinatorial configuration space of 108,000 discrete cells for dashcam collision detection over 27 days. Through ANOVA decomposition, we find that architectural choices explain 94% of performance variance (F = 1324, η2 = 0.94), while hyperparameter variation within a fixed architecture explains only 6%. Cross-task validation on a second collision dataset confirms this finding (75% architecture-explained variance) with a different winning backbone, confirming genuine architecture discovery. The agents' key contribution is discovering that V-JEPA,2 video features with Zipformer temporal encoders achieve 0.9245 AP -- a configuration no human proposed -- and concentrating search on productive architectural regions: at N = 50, LLM-guided search reaches AP = 0.985 versus 0.965 for from-scratch random search. Post-bugfix convergence follows a power law (c = 0.11, R2 = 0.93); the low exponent reflects the cost of broad exploration, not inefficiency, since the LLM discovers qualitatively better regions than random or Bayesian baselines. We characterize multi-agent search dynamics via entropy cycles and Jensen--Shannon specialization, providing the first large-scale empirical framework for LLM-guided combinatorial ML experiment design.