Yeco Flash Mini-IT v1
118M-parameter Italian micro-LLM, trained for under €15, ~200 tok/s on CPU. Beats 5× larger models on Italian logic benchmarks.
Paper
YecoAI Lab
Sandboxed adversarial research for prompt injection and LLM defense.
Abstract
YecoAI Lab is a controlled environment designed specifically for red teaming — simulating attacks to identify weaknesses before deployment.
Capabilities
- Automated attack vectors: testing thousands of known injection patterns
- Gradient-based attacks: optimization techniques to find adversarial suffixes
- Defense evaluation: measuring the effectiveness of various guardrails
Thesis
Security is not a feature; it is a fundamental requirement for any autonomous system. By rigorously testing our models in the Lab, we ensure that the agents we build are not just smart, but safe.