2 papers
cs.CR2026
Silent Alarm: A J-Space Protocol for Comparing Danger Recognition Across Models and Quantization Levels
Roman Prosvirnin, Victor Minchenkov, Alexey Soldatov +2
Jailbreak-robustness research typically evaluates safety through generated responses using an LLM-as-judge approach. Such evaluations, however, are sensitive to the benchmark's gra…
cs.CL2026
ContextRAG: Extraction-Free Hierarchical Graph Construction for Retrieval-Augmented Generation
Roman Prosvirnin, Sergei Kuznetsov, Seungmin Jin
Graph-structured retrieval-augmented generation (RAG) systems can improve answer quality on multi-hop questions, but many current systems rely on large language models (LLMs) to ex…