Calibrated Decision Models for Autonomous Penetration-Testing Harnesses
Paper on where Jev and Laya fit in LLM pentest agents: finding adjudication, severity recalibration, pruning and confirmation loops.
This summary is based on how curated lists describe the project; its README was not available to review.
Overview
An arXiv paper examining where calibrated decision models such as Jev and Laya fit inside autonomous penetration-testing harnesses driven by LLM agents.
Per the listing, it covers adjudicating findings, recalibrating severity, pruning agent branches and running confirmation loops, with an exploratory case study.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Yes / no | Is this reported finding real?(inferred) | yes · no |
| Score | What severity should this finding have?(inferred) | — |
How it works
- LLM pentest agent produces findings
- Decision model adjudicates and recalibrates severity
- Prune unproductive branches
- Confirm findings in a loop
Strengths
- Maps decision models onto a concrete security workflow
Caveats
- Preprint; exploratory case study
- Only list summary available
At a glance
- Type
- Paper
- Status
- research
Use cases and integrations
How curated lists describe it
Where Jev and Laya fit in LLM pentest agents: finding adjudication, severity recalibration, agent pruning, and confirmation loops, with an exploratory case study
AnotiaWang/awesome-decision-models
Frequently asked questions
Which models are discussed?
Jev and Laya.
Where do they fit?
Finding adjudication, severity recalibration, pruning and confirmation loops.
Where is it?
https://arxiv.org/abs/2609.28940
Related projects
Python kernel of Jev Choice/Score/Noul gates, caches and loops used to admit or reject Lean proof candidates in research pipelines.
LibraryresearchPythonOpen-weight 0.6B decision model on a Qwen3 backbone: state and typed questions in, a probability for every option out, nothing decoded.
ModelresearchPythonSuperagent's Apache-2.0 security decision model from AutoJev-27B, scoring options for injection, tool calls, code and alerts.
ModelbetaAutoGPT agent platform, listed as adding Jev decision blocks for typed routing, filtering, scoring and confidence-gated next actions.
ProjectproductionPythonGo CLI for managing WordPress sites at scale, with typesafe commands that ask Jev questions and a scan --triage that ranks malware findings.
ProjectproductionGoMulti-agent orchestration layer for Claude Code with opt-in Jev hooks for fast judgment points, decision caching and egress controls.
PluginproductionTypeScript