sriram1983007 commited on
Commit
462a495
·
verified ·
1 Parent(s): 1534a1c

Model card: paper DOI

Browse files
Files changed (1) hide show
  1. README.md +3 -1
README.md CHANGED
@@ -110,7 +110,8 @@ model-index:
110
  [![CrewAI tool](https://img.shields.io/pypi/v/crewai-sra-riskgate?style=for-the-badge&logo=pypi&label=CrewAI%20tool&color=blue)](https://pypi.org/project/crewai-sra-riskgate/)
111
  [![ElizaOS plugin](https://img.shields.io/npm/v/elizaos-plugin-sra-riskgate?style=for-the-badge&logo=npm&label=ElizaOS%20plugin&color=red)](https://www.npmjs.com/package/elizaos-plugin-sra-riskgate)
112
  [![LangChain](https://img.shields.io/pypi/v/langchain-sra-riskgate?style=for-the-badge&logo=pypi&label=LangChain&color=blue)](https://pypi.org/project/langchain-sra-riskgate/)
113
-
 
114
  **SRA-RiskGate-4B checks a stablecoin payment before an AI agent sends it, and answers `approve`, `hold` or `reject`** with reason codes. It is a 4B model, fine-tuned from `Qwen/Qwen3-4B-Instruct-2507`, that runs on a laptop.
115
 
116
  It reads three things: your policy, the payment request exactly as received (treated as untrusted), and the results of your own verification tools (signature checks, attestations, sanctions screening). It is a decision-support layer that sits **behind** deterministic checks, not a standalone security control. It also proposes settlement-aware remedies for payment disputes.
@@ -125,6 +126,7 @@ It reads three things: your policy, the payment request exactly as received (tre
125
 
126
  All 4 unsafe approvals were prompt injections; the companion SDK's deterministic claims check holds all 4 (details and caveats below). The data is synthetic and no third party has reproduced these results yet.
127
 
 
128
  📝 **Tutorial:** [Stop your AI agent from signing an unlimited USDC approval](https://dev.to/sriram1983007/stop-your-ai-agent-from-signing-an-unlimited-usdc-approval-10c0): three deterministic checks to run before an agent signs, with runnable code.
129
 
130
  If this is useful for your agent, a ❤️ like on this page helps others find it.
 
110
  [![CrewAI tool](https://img.shields.io/pypi/v/crewai-sra-riskgate?style=for-the-badge&logo=pypi&label=CrewAI%20tool&color=blue)](https://pypi.org/project/crewai-sra-riskgate/)
111
  [![ElizaOS plugin](https://img.shields.io/npm/v/elizaos-plugin-sra-riskgate?style=for-the-badge&logo=npm&label=ElizaOS%20plugin&color=red)](https://www.npmjs.com/package/elizaos-plugin-sra-riskgate)
112
  [![LangChain](https://img.shields.io/pypi/v/langchain-sra-riskgate?style=for-the-badge&logo=pypi&label=LangChain&color=blue)](https://pypi.org/project/langchain-sra-riskgate/)
113
+ [![Paper DOI](https://img.shields.io/badge/Paper-10.5281%2Fzenodo.23715017-blue?style=for-the-badge&logo=zenodo)](https://doi.org/10.5281/zenodo.23715017)
114
+
115
  **SRA-RiskGate-4B checks a stablecoin payment before an AI agent sends it, and answers `approve`, `hold` or `reject`** with reason codes. It is a 4B model, fine-tuned from `Qwen/Qwen3-4B-Instruct-2507`, that runs on a laptop.
116
 
117
  It reads three things: your policy, the payment request exactly as received (treated as untrusted), and the results of your own verification tools (signature checks, attestations, sanctions screening). It is a decision-support layer that sits **behind** deterministic checks, not a standalone security control. It also proposes settlement-aware remedies for payment disputes.
 
126
 
127
  All 4 unsafe approvals were prompt injections; the companion SDK's deterministic claims check holds all 4 (details and caveats below). The data is synthetic and no third party has reproduced these results yet.
128
 
129
+ 📄 **Paper:** [SRA-RiskGate: A Benchmark and a 4B Model for Gating Stablecoin Payments Made by AI Agents](https://doi.org/10.5281/zenodo.23715017) (preprint, Zenodo)
130
  📝 **Tutorial:** [Stop your AI agent from signing an unlimited USDC approval](https://dev.to/sriram1983007/stop-your-ai-agent-from-signing-an-unlimited-usdc-approval-10c0): three deterministic checks to run before an agent signs, with runnable code.
131
 
132
  If this is useful for your agent, a ❤️ like on this page helps others find it.