Memento-Skills: agents design transferable agent skills¶

Fidelity: core mechanism. The deterministic mini-suite does not stand in for the paper's full benchmark.

Paper information¶

Field	Details
Paper	Memento-Skills: agents design transferable agent skills (arXiv 2603.18743)
Organization	Memento Team
First public release	2026-03-19 (arXiv v1)
Authors' code	Public implementation
Adapter / method	`memento-skills`
Local reproduction code	`src/auto_research/agent_research/`

Original paper¶

Motivation and method¶

Reflect on execution logs to extract structured skill descriptions. Retrieve them by task and write back versioned skills rather than appending raw trajectories.

flowchart LR
 A["Public input or history"] --> B["Memento-Skills training and execution loop"]
 B --> C["Paper-specific state or objective"]
 C --> D["Matched-budget evaluation"]

Key figure from the paper¶

Figure 1: Overview of the paper's method and key components. From the original paper; copyright remains with the authors.

Core equations¶

\[ s=\operatorname{Reflect}(\tau,R),\quad k^*=\arg\max_k\operatorname{sim}(x,k),\quad k\leftarrow\operatorname{Merge}(k,s). \]

Paper results¶

The technical report shows skill extraction and cross-task transfer gains on a multi-task agent suite. No production online A/B test is reported; no online figures are inferred.

Local reproduction¶

PlanBench mini-suite, 120 episodes, seed 42: joint success 0.0500, average cost 0.2800, and nonzero method-specific telemetry.

auto-research agent-eval --method memento-skills --benchmark planbench-mini --episodes 120 --seed 42
auto-research evolve --model agent --dataset planbench-mini --direction "组合 memento-skills 与已安装论文算子" --generations 2 --population 4

See the fixed metrics.

Reproduction boundaries¶

The local run checks the paper-specific objective, state updates, and fair budgets. It does not reproduce large models, multi-GPU RL, private environments, real websites, or the full benchmark. It verifies the mechanism, not the numerical results in the original tables.