The fdtn-ai/antares-1b model, a 1-billion parameter language model built on IBM Granite 4.0 1B, has been released. This model is specialized for vulnerability localization in codebases and operates by autonomously navigating source code repositories through a terminal interface. It utilizes a two-stage pipeline of supervised fine-tuning (SFT) and Group Relative Policy Optimization (GRPO) to identify vulnerable code segments. Instructions are available for integrating Antares-1B with libraries like Transformers, inference providers such as vLLM and SGLang, and deployment via Docker. AI
IMPACT Enables automated vulnerability detection in codebases, potentially improving software security and development workflows.
RANK_REASON Release of a specialized language model with technical details and integration instructions. [lever_c_demoted from research: ic=1 ai=1.0]
Read on Hugging Face Trending Models →
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
