Take two strong coding agents. Give each the same machine-learning task and six hours, and Claude Opus 5 wins a medal on 72.4% of MLE-bench's 75 Kaggle tasks while GPT-5.6-sol wins one on 68.0%. Now give them one six-hour budget and make them take turns in the same workspace. Together they medal on 78.2%.
That is Humanize MLE Agents (HMA), and the number comes from its results table.
Best pairing
+8.0 ptsOpus 5 ⇄ GPT-5.6-sol, over the mean of the two alone (70.2%), in any-medal rate.
All six pairings
+6.4 ptsThe average gain over each pairing’s own single-agent mean. The conservative figure.
Mean percentile
+5.9 ptsWhere the best pairing lands on each task’s leaderboard, against the mean of its two agents alone.
Taking turns, with a clean slate
The arrangement is simple, and almost all of it is about what is not carried over.
Swipe the diagram sideways to see all of it →
1/0 One shared workspace, and one six-hour budget for the task.
- One agent works the task in its own native harness and submits candidate solutions.
- After at most five accepted submissions, or when it finishes on its own, it stops. The other agent starts a fresh session in the same workspace: the same code, models and candidates, but none of the first agent's context.
- They alternate until the six hours are nearly up. The last 15 minutes go to picking one of the candidates already accepted. Private test scores are never shown to either agent.
The repository credits two effects at once. The first is context renewal: a long-running agent improves more and more slowly, and a fresh session restarts that curve. The second is complementary capabilities: the second model searches differently from the first. The handoff keeps everything the first agent built and drops only its train of thought.
Against each agent alone
The same table lists two external systems with longer budgets: ScienceFlow reaches 70.2% in 24 hours, and MLEvolve 65.3% in 12. Every HMA pairing uses six.
- GPT-5.6-sol alone: Time budget 6 h, Any medal 68.0%
- GPT-5.6-sol ⇄ Kimi K3: Time budget 6 h, Any medal 70.7%
- GPT-5.6-sol ⇄ DeepSeek V4.1 Flash: Time budget 6 h, Any medal 72.0%
- Opus 5 alone: Time budget 6 h, Any medal 72.4%
- GPT-5.6-sol ⇄ Opus 5: Time budget 6 h, Any medal 73.3%
- Opus 5 ⇄ GPT-5.6-sol: Time budget 6 h, Any medal 78.2%
- MLEvolve: Time budget 12 h, Any medal 65.3%
- ScienceFlow: Time budget 24 h, Any medal 70.2%
| HMA: Opus 5 ⇄ GPT-5.6-sol | 6 | 78.2 | 53.8 | 90.7 | 89.0 |
|---|---|---|---|---|---|
| HMA: GPT-5.6-sol ⇄ Opus 5 | 6 | 73.3 | 50.7 | 85.3 | 85.4 |
| Claude Opus 5, native /goal | 6 | 72.4 | 50.7 | 85.3 | 84.8 |
| HMA: GPT-5.6-sol ⇄ DeepSeek V4.1 Flash | 6 | 72.0 | 50.7 | 85.3 | 84.4 |
| HMA: GPT-5.6-sol ⇄ Kimi K3 | 6 | 70.7 | 50.7 | 85.3 | 83.9 |
| ScienceFlow | 24 | 70.2 | — | — | — |
| GPT-5.6-sol, native /goal | 6 | 68.0 | 47.6 | 81.3 | 81.3 |
| MLEvolve | 12 | 65.3 | 34.7 | 76.0 | — |
What to be careful with
- Order matters a lot. The same two models score 78.2% with Opus going first and 73.3% with GPT-5.6-sol going first. Choosing the best order after the fact flatters the headline. The average over pairings, 6.4 points, is the more conservative figure.
- This is self-reported, and it is not on the official leaderboard. The MLE-bench leaderboard has not reviewed these runs, and it has not taken new submissions since April. For scale only, its highest "All" entry today is 64.44%, with a 24-hour budget and an older model. Different models, budgets and repeat counts make that a different experiment, not a lower bar.
- The paper is not out yet. The arXiv link will be added with the paper release. Until then, the full ablations exist only in the repository's experiment map.
The whole suite reruns from the repository: the baselines, HMA, and the cap and starting-order ablations. To use the same alternation on your own work, run it as a flow: fixed-interrupt flame chase.