SEARL: jointly evolving a policy and tool-graph memory¶

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

Paper information¶

Field	Details
Paper	SEARL: jointly evolving a policy and tool-graph memory (arXiv 2604.07791)
Organization	Shanghai AI Laboratory / Shanghai Jiao Tong University
First public release	2026-04-09 (arXiv v1)
Authors' code	No official author repository was found.
Adapter / method	`searl`
Local reproduction code	`src/auto_research/agent_research/`

Original paper¶

Motivation and method¶

Maintain tools and successful transitions in a memory graph. Each new rollout updates both the policy and graph-edge weights, closing the loop between experience, retrieval, and improvement.

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

Key figure from the paper¶

Figure 2: Core architecture, modules, and their connections. From the original paper; copyright remains with the authors.

Core equations¶

\[ a_t\sim\pi_\theta(\cdot|x,G),\quad G\leftarrow G+\Delta(\tau,R),\quad\theta\leftarrow\theta+\eta\nabla J(\theta;G). \]

Paper results¶

On several tool-agent tasks, joint policy and graph-memory optimization outperforms separate updates. No production online A/B test is reported; no online figures are inferred.

Local reproduction¶

PlanBench mini-suite, 120 episodes, seed 42: joint success 1.0000, average cost 0.3400, and nonzero method-specific telemetry.

auto-research agent-eval --method searl --benchmark planbench-mini --episodes 120 --seed 42
auto-research evolve --model agent --dataset planbench-mini --direction "组合 searl 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.