{"slug": "feedback-wanted-low-latency-llm-guardrails-using-a-14m-parameter-discriminator", "title": "[Feedback Wanted] Low-Latency LLM Guardrails Using a 14M Parameter Discriminator", "summary": "DeshwalX released electra-small-prompt-injection-v1, a 14M-parameter multi-label classifier for low-latency LLM guardrails, achieving 90% micro F1 on the WildGuard benchmark. The model detects prompt injection, jailbreaks, and harmful content in a single forward pass, offering a faster alternative to large autoregressive detectors.", "body_md": "Hi everyone,\n\nI have released ** DeshwalX/electra-small-prompt-injection-v1**, a multi-label classifier for low-latency LLM guardrails.\n\nMany prompt injection detectors rely on large autoregressive models that introduce processing latency. This project tests the baseline efficiency of a compact encoder backbone by fine-tuning ** google/electra-small-discriminator** (~14M parameters) to detect four independent safety vectors in a single forward pass.\n\n**WildGuard Test Benchmark Evaluation:**\n\n**Micro F1:** 90.00% | **Macro F1:** 89.00%\n\n**Prompt Adversarial (Jailbreaks):** 1.00 F1\n\n**Response Refusal:** 0.86 F1\n\n**Prompt Harmful:** 0.88 F1\n\n**Response Harmful:** 0.81 F1\n\n**Model Repository:** [DeshwalX/electra-small-prompt-injection-v1 · Hugging Face](https://huggingface.co/DeshwalX/electra-small-prompt-injection-v1)\n\nI am looking for community feedback to improve **Version 1**. Please test the model on your own data and share:\n\nEdge cases where the model fails or misses an injection.\n\nExamples of false positives on benign inputs.\n\nSuggestions for architecture configurations or loss weight balancing for the next iteration.", "url": "https://wpnews.pro/news/feedback-wanted-low-latency-llm-guardrails-using-a-14m-parameter-discriminator", "canonical_source": "https://discuss.huggingface.co/t/feedback-wanted-low-latency-llm-guardrails-using-a-14m-parameter-discriminator/177768#post_1", "published_at": "2026-07-13 09:03:31+00:00", "updated_at": "2026-07-13 09:18:30.067114+00:00", "lang": "en", "topics": ["large-language-models", "ai-safety", "ai-research", "ai-tools", "natural-language-processing"], "entities": ["DeshwalX", "google/electra-small-discriminator", "WildGuard"], "alternates": {"html": "https://wpnews.pro/news/feedback-wanted-low-latency-llm-guardrails-using-a-14m-parameter-discriminator", "markdown": "https://wpnews.pro/news/feedback-wanted-low-latency-llm-guardrails-using-a-14m-parameter-discriminator.md", "text": "https://wpnews.pro/news/feedback-wanted-low-latency-llm-guardrails-using-a-14m-parameter-discriminator.txt", "jsonld": "https://wpnews.pro/news/feedback-wanted-low-latency-llm-guardrails-using-a-14m-parameter-discriminator.jsonld"}}