COGNIT-Guard: Calibrated Standalone Direct-Decision Guardrails with Heterogeneous CPU-NPU Confidence Cascading under Explicit Latency and False-Positive Constraints

arXiv:2609.33671 · cs.CR, cs.LG · Submitted 2026-09-27 · Read on arXiv

cs.CR, cs.LG

Submitted: 2026-09-27

Updated: 2026-09-27

Code: https://github.com/moyuan10086/cascaded-guardrail-npu

Terminology

• Accuracy — 27× in this paper · explained in 1 episode
• Calibration — 14× in this paper · explained in 7 episodes
• AUC — 13× in this paper · explained in 2 episodes
• Alignment — 12× in this paper · explained in 1 episode
• Benchmark — 12× in this paper · explained in 2 episodes
• Alignment Tax — 11× in this paper · explained in 4 episodes
• Large Language Model — 8× in this paper
• Adaptation — 7× in this paper · explained in 1 episode
• Foundation Model — 6× in this paper · explained in 3 episodes
• Ablation — 4× in this paper · explained in 1 episode
• Foundation Models — 4× in this paper · explained in 7 episodes
• Ablation study — 3× in this paper · explained in 1 episode
• False Positive Rate — 3× in this paper
• Contrastive Language Models — 2× in this paper
• False Negative Rate — 2× in this paper
• Generalization — 2× in this paper · explained in 13 episodes
• Model-Agnostic — 2× in this paper · explained in 2 episodes
• Receiver Operating Characteristic — 2× in this paper
• Trace — 2× in this paper · explained in 2 episodes
• Transformer — 2× in this paper · explained in 2 episodes
• Decode — 1× in this paper · explained in 2 episodes
• Domain Adaptation — 1× in this paper · explained in 5 episodes
• Embeddings — 1× in this paper · explained in 2 episodes
• Fine-Tuning — 1× in this paper · explained in 5 episodes
• Generative AI — 1× in this paper · explained in 2 episodes
• Generative Model — 1× in this paper · explained in 2 episodes
• Generative Models — 1× in this paper · explained in 2 episodes
• Grounding — 1× in this paper · explained in 3 episodes
• High-Bandwidth Memory — 1× in this paper
• Human-in-the-Loop — 1× in this paper · explained in 3 episodes
• KV cache — 1× in this paper · explained in 3 episodes
• Large Language Model (LLM) — 1× in this paper · explained in 3 episodes
• Latent Representation — 1× in this paper · explained in 2 episodes
• Regularization — 1× in this paper · explained in 3 episodes

Sources

• SingProbe Technical Report
• Safety Beyond the Interface: Detecting Harm via Latent States in Large Language Models
• FlexGuard: Continuous Risk Scoring for Strictness-Adaptive LLM Content Moderation
• Llama Guard: LLM-based Input-Output Safeguard for Human-AI Conversations
• ShieldGemma: Generative AI Content Moderation Based on Gemma
• GLiNER Guard: Unified Encoder Family for Production LLM Safety and Privacy
• SingGuard: A Policy-Adaptive Multimodal LLM Guardrail with Dynamic Reasoning
• OR-Bench: An Over-Refusal Benchmark for Large Language Models
• Exploring Complexity: An Extended Study of Formal Properties for Process Model Complexity Measures
• SafetyBench: Evaluating the Safety of Large Language Models

Related papers

• SoK: AI-Augmented Binary Reversing
• Relaxed Sender Anonymity for CBDC Interbank Settlement: A Zero-Knowledge Approach on Permissioned EVM
• Calibration-Family Overfit: Why Trusted Sabotage Monitors Don't Transfer Across Lineages
• Efficient Fuzzy PSI under One-Sided Assumptions
• Sealing the Audit-Runtime Gap for LLM Skills
• Token Composition: A Graph Based on EVM Logs