Project Video
A short visual overview of how SkillOpt treats natural-language skills as trainable artifacts: roll out, reflect, edit, validate, and export.
Promotional video for the SkillOpt project page. The static paper teaser is shown below for high-resolution inspection.
Paper Teaser
The teaser summarizes the SkillOpt training loop: rollout evidence, optimizer-side reflection, bounded skill edits, validation gating, and the exported reusable skill.
Figure from the SkillOpt paper. On small screens, the figure area scrolls horizontally to preserve the original details.
Instead of fine-tuning a model or hand-maintaining prompts, SkillOpt runs the frozen agent on scored batches, asks a separate optimizer model to propose structured edits, and accepts a candidate only when validation performance improves.
Frozen target model Optimizer model Add / delete / replace edits Held-out gate
Rollout
The target model executes tasks with the current skill and records scored trajectories.
Reflect
The optimizer analyzes success and failure minibatches to find reusable procedures.
Edit
Candidate add, delete, and replace operations are merged and ranked under a budget.
Gate
The candidate skill is kept only if it improves held-out selection performance.
Rollout batches capture messages, tool calls, verifier feedback, task metadata, and final scores.
Failures and successes are reflected separately so edits correct recurring errors while preserving working behavior.
An edit budget functions as a textual learning rate, preventing useful rules from being overwritten by broad rewrites.
Rejected edits, slow update, and optimizer-side meta skill provide longer-horizon feedback without bloating deployment.
Method comparison
| Component | Setting | SearchQA | Spreadsheet | LiveMath |
|---|---|---|---|---|
| Learning rate | lr=4 default | 87.1 | 77.5 | 61.3 |
| Learning rate | without lr | 84.6 | 75.7 | 57.3 |
| Rejected buffer | with buffer | 87.1 | 77.5 | 61.3 |
| Rejected buffer | without buffer | 85.5 | 72.9 | 58.9 |
| Update memory | meta skill + slow update | 87.1 | 77.5 | 61.3 |
| Update memory | without both | 86.3 | 55.0 | 59.7 |
Bounded Textual learning rates prevent destructive rewrites while keeping enough plasticity to learn new procedures.
Gated Held-out selection turns reflection into propose-and-test optimization rather than unconditional self-editing.
Buffered Rejected edits become negative feedback, helping the optimizer avoid repeating harmful directions.
Accepted edits become the current skill only after held-out selection improves. Step 3 is rescued by a slow update; Step 4 trains higher but fails selection.
GPT-5.4 LiveMath skill transferred to GPT-5.4-nano on LiveMathBench.
Codex-trained SpreadsheetBench skill transferred into Claude Code.
GPT-5.4-nano used as its own optimizer improved SpreadsheetBench over baseline.
The target model consumes only the final skill, not optimizer memory.
A stronger optimizer model gives the largest gains, but the loop is not merely distillation from a stronger model. Even matched target-as-optimizer settings can discover useful edits when the update is constrained, buffered, and validated.
@misc{yang2026skilloptexecutivestrategyselfevolving,
title={SkillOpt: Executive Strategy for Self-Evolving Agent Skills},
author={Yifan Yang and Ziyang Gong and Weiquan Huang and Qihao Yang and Ziwei Zhou and Zisu Huang and Yan Li and Xuemei Gao and Qi Dai and Bei Liu and Kai Qiu and Yuqing Yang and Dongdong Chen and Xue Yang and Chong Luo},
year={2026},
eprint={2605.23904},
archivePrefix={arXiv},
primaryClass={cs.AI},
url={https://arxiv.org/abs/2605.23904},
}