Dialect-aware LLM safety classifiers from the DIA-GUARD project — fine-tuned on 48 English dialects for safe/unsafe content detection.
Jason Lucas
jsl5710
·
AI & ML interests
Trustworthy AI, Multilingual NLP, Low-Resource Languages, Safe AI, Transfer Learning, AI for Cybersecurity, Human-Center AI, Privacy and Security
Recent Activity
updated a model 1 day ago
jsl5710/Shield-SmolLM2-1.7B-Full-FT-CE published a model 1 day ago
jsl5710/Shield-SmolLM2-1.7B-Full-FT-CE updated a model 1 day ago
jsl5710/Shield-Llama-3.2-1B-Full-FT-CEOrganizations
models 13
jsl5710/Shield-SmolLM2-1.7B-Full-FT-CE
Text Generation • 2B • Updated • 500
jsl5710/Shield-Qwen3Guard-Gen-0.6B-Full-FT-CE
Text Generation • 0.6B • Updated • 547
jsl5710/Shield-Qwen3-1.7B-Full-FT-CE
Text Generation • 2B • Updated • 986
jsl5710/Shield-Llama-3.2-1B-Full-FT-CE
Text Generation • 1B • Updated • 653
jsl5710/Shield-Gemma-3-270m-Full-FT-CE
Text Generation • 0.3B • Updated • 721
jsl5710/Shield-Gemma-3-1B-Full-FT-CE
Text Generation • 1.0B • Updated • 497
jsl5710/Shield-Qwen3Guard-Gen-0.6B-PEFT-CE
Text Generation • Updated • 62
jsl5710/Shield-SmolLM2-1.7B-PEFT-CE
Text Generation • Updated • 91
jsl5710/Shield-Qwen3.5-0.8B-PEFT-CE
Text Generation • Updated • 65
jsl5710/Shield-Qwen3-1.7B-PEFT-CE
Text Generation • Updated • 69