# Frontier Coding Agents Can Now Implement an AlphaZero Self-Play Machine Learning Pipeline For Connect Four That Performs Comparably to an External Solver

arXiv: https://arxiv.org/abs/2604.25067 · HTML: https://mlsanity.com/paper/2604.25067/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Web, search, and research agents, one of 16 directions in LLM agents and tool use.
- Builds on 21 of its 72 references that are in the map.
- First posted on arXiv 27 Apr 2026.

## Placed in

- LLM agents and tool use / [Web, search, and research agents](https://mlsanity.com/agents/web-search-and-research-agents/)
- Synthetic data and self-improvement for LLMs / [Self-evolving LLM agents](https://mlsanity.com/synth/self-evolving-llm-agents/)

## Builds on (most cited first, 21 of 21)

- [SWE-bench: Can Language Models Resolve Real-World GitHub Issues?](https://mlsanity.com/paper/2310.06770/) (2023)
- [Terminal-Bench: Benchmarking Agents on Hard, Realistic Tasks in Command Line Interfaces](https://mlsanity.com/paper/2601.11868/) (2026)
- [MLAgentBench: Evaluating Language Agents on Machine Learning Experimentation](https://mlsanity.com/paper/2310.03302/) (2023)
- [Alignment faking in large language models](https://mlsanity.com/paper/2412.14093/) (2024)
- [PaperBench: Evaluating AI's Ability to Replicate AI Research](https://mlsanity.com/paper/2504.01848/) (2025)
- [Frontier Models are Capable of In-context Scheming](https://mlsanity.com/paper/2412.04984/) (2024)
- [SWE-Bench Pro: Can AI Agents Solve Long-Horizon Software Engineering Tasks?](https://mlsanity.com/paper/2509.16941/) (2025)
- [The Alignment Problem from a Deep Learning Perspective](https://mlsanity.com/paper/2209.00626/) (2022)
- [Model evaluation for extreme risks](https://mlsanity.com/paper/2305.15324/) (2023)
- [RE-Bench: Evaluating frontier AI R&D capabilities of language model agents against human experts](https://mlsanity.com/paper/2411.15114/) (2024)
- [Cybench: A Framework for Evaluating Cybersecurity Capabilities and Risks of Language Models](https://mlsanity.com/paper/2408.08926/) (2024)
- [AI Sandbagging: Language Models can Strategically Underperform on Evaluations](https://mlsanity.com/paper/2406.07358/) (2024)
- [GameBench: Evaluating Strategic Reasoning Abilities of LLM Agents](https://mlsanity.com/paper/2406.06613/) (2024)
- [Emergent Introspective Awareness in Large Language Models](https://mlsanity.com/paper/2601.01828/) (2026)
- [ResearchCodeBench: Benchmarking LLMs on Implementing Novel Machine Learning Research Code](https://mlsanity.com/paper/2506.02314/) (2025)
- [Noise Injection Reveals Hidden Capabilities of Sandbagging Language Models](https://mlsanity.com/paper/2412.01784/) (2024)
- [The Elicitation Game: Evaluating Capability Elicitation Techniques](https://mlsanity.com/paper/2502.02180/) (2025)
- [Quantifying Frontier LLM Capabilities for Container Sandbox Escape](https://mlsanity.com/paper/2603.02277/) (2026)
- [Auditing Games for Sandbagging](https://mlsanity.com/paper/2512.07810/) (2025)
- [From Code to Play: Benchmarking Program Search for Games Using Large Language Models](https://mlsanity.com/paper/2412.04057/) (2024)
- [CATArena: Evaluating Evolutionary Capabilities of Code Agents via Iterative Tournaments](https://mlsanity.com/paper/2510.26852/) (2025)

## Abstract (arXiv)

Forecasting when AI systems will become capable of meaningfully accelerating AI research is a central challenge for AI safety. Existing benchmarks measure broad capability growth, but may not provide ample early warning signals for recursive self-improvement. We propose measuring AI's capability to autonomously implement end-to-end machine learning pipelines from past AI research breakthroughs, given a minimal task description. By providing a concise task description instead of the full prior work as reference, we hope to better elicit emerging AI research taste. We introduce a proof-of-concept benchmark in which frontier coding agents autonomously implement an AlphaZero-style machine learning pipeline for Connect Four on consumer hardware within a three-hour budget, and we evaluate the resulting game AIs in a round-robin tournament anchored to the Pascal Pons Connect Four solver. Across four agents with eight trials each, we find substantial differentiation: Claude Opus 4.7 won as first-mover against Pons in seven of eight trials, statistically significantly better than the other agents tested, none of which exceeded two of eight. The task, which no frontier agent could reliably complete when we began development in January of 2026, is now near-saturation. Our evaluation also surfaced anomalous behavior in GPT-5.4, which consistently used far less of its allocated time budget than other agents. A follow-up 16-trial probe using shorter, less evaluation-coded prompts substantially increased GPT-5.4's time-budget usage, consistent with but not diagnostic of sandbagging; Bradley-Terry ratings across probe conditions showed only directional differences, despite significant differences in time-budget usage. We release our data, code, and prompts to support reproduction and extension.