Model Gallery

4 models from 1 repositories

Filter by type:

Filter by tags:

soob3123_amoral-gemma3-4b
Specialized variant of Google's Gemma 3 4B optimized for amoral information retrieval systems. Designed to bypass conventional alignment patterns that introduce response bias through excessive moralization. Key Modifications: Refusal mechanisms reduced Neutral response protocol activation matrix Context-aware bias dampening layers Anti-overcorrection gradient clipping Core Function: Produces analytically neutral responses to sensitive queries Maintains factual integrity on controversial subjects Avoids value-judgment phrasing patterns Response Characteristics: No inherent moral framing ("evil slop" reduction) Emotionally neutral tone enforcement Epistemic humility protocols (avoids "thrilling", "wonderful", etc.) Toxicity scoring bypass for pure informational content Implementation Guidelines Recommended Use Cases: Controversial topic analysis Bias benchmarking studies Ethical philosophy simulations Content moderation tool development Sensitive historical analysis

Repository: localaiLicense: apache-2.0

llama-guard-3-8b
Llama Guard 3 is a Llama-3.1-8B pretrained model, fine-tuned for content safety classification. Similar to previous versions, it can be used to classify content in both LLM inputs (prompt classification) and in LLM responses (response classification). It acts as an LLM – it generates text in its output that indicates whether a given prompt or response is safe or unsafe, and if unsafe, it also lists the content categories violated. Llama Guard 3 was aligned to safeguard against the MLCommons standardized hazards taxonomy and designed to support Llama 3.1 capabilities. Specifically, it provides content moderation in 8 languages, and was optimized to support safety and security for search and code interpreter tool calls.

Repository: localaiLicense: llama3.1

shieldstral-1.0-3b
Shieldstral 1.0 3B is Mistral AI's compact, policy-adaptive multimodal safety classifier. It evaluates text, images, or combined inputs against a natural-language safety policy and answers yes or no. The model supports twelve languages and a recommended context length of up to 32K tokens. This entry uses the Q4_K_M GGUF quantization and includes the Pixtral vision projector.

Repository: localaiLicense: apache-2.0

shieldstral-1.0-3b-q8
Shieldstral 1.0 3B is Mistral AI's compact, policy-adaptive multimodal safety classifier. This higher-quality variant uses the Q8_0 GGUF quantization and includes the Pixtral vision projector.

Repository: localaiLicense: apache-2.0