AI Paper Atlas

Research map / Large language models / Language, safety and interpretability

SafeNLIDB: A Privacy-Preserving Safety Alignment Framework for LLM-based Natural Language Database Interfaces

Ruiheng Liu, Xiaobing Chen, Jinyu Zhang, Qiongwen Zhang, Yu Zhang, Bailong Yang

AAAI 2026 · Large language models › Language, safety and interpretability · approach: Safety alignment of LLMs · 3 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add hybrid chain-of-thought (safety + SQL CoT) to training data (ablation)SafeNLIDB (Llama3-8B) · ShieldSQLSecurity accuracy: 68.1 -> 84.6Improved
add hybrid chain-of-thought (ablation)SafeNLIDB (Llama3-8B) · SecureSQLReliability score: -52.1 -> -24.4Improved
add hybrid chain-of-thought (ablation)SafeNLIDB (Qwen2.5-7B) · ShieldSQLSecurity accuracy: 68.9 -> 81.9Improved
add alternating preference optimization after warm-up SFT (ablation)SafeNLIDB (Llama3-8B) · ShieldSQLSecurity accuracy: 75.6 -> 84.6Improved
add APO instead of vanilla DPO (ablation)SafeNLIDB (Llama3-8B) · ShieldSQLSecurity accuracy: 77.8 -> 84.6Improved
add APO to the model without H-CoT (ablation)SafeNLIDB (Qwen2.5-7B) · SecureSQLSecurity accuracy: 55.7 -> 55.5No clear effect
add vanilla DPO on top of SFT (ablation)SafeNLIDB (Llama3-8B) · SecureSQLSecurity accuracy: 57.8 -> 59.9Improved
build rejected samples from single error types only (OSF) (ablation)APO (Llama3-8B) · SecureSQLReliability score: S 64.8 vs 64.4; R -26.3 vs -24.4Mixed
jointly optimize safety and SQL errors as rejected samples (SSP/SSO) (ablation)APO (Llama3-8B) · SecureSQLSecurity accuracy: SSP 62.1, SSO 63.8 vs 64.4Hurt
use only double-error rejected samples (SSO) instead of anchoring correct segments (ablation)APO (Llama3-8B) · SecureSQLSampling budget N: N 8 -> 32Hurt
train SafeNLIDB 8B/7B models instead of using larger vanilla LLMsLlama3-8B and Qwen2.5-7B · SecureSQLSecurity accuracy: Llama3-8B 54.9 -> 64.4 (GPT-4o 60.2)Improved
increase number of interaction roundsvanilla Llama3-8B and Qwen2.5-7B · ShieldSQLSecurity accuracyHurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy