AI Paper Atlas

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

WALKSAFE: Risk-aware Graph Random Walk with Bi-GRPO for LLM Safety

Shilong Pan, Zhiliang Tian, Wanlong Yu, Zhen Huang, Qingyu Qiu, Zihan Chen, Zhonghao Sun, Minlie Huang, Dongsheng Li

AAAI 2026 · Large language models › Language, safety and interpretability · approach: Safety alignment of LLMs · 1 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 entity-relation graph structure (ablation)WalkSafe, Llama-3.2-3B · TETavg harmlessness (lower better): 6.70 -> 3.25Improved
add entity-relation graph structure (ablation)WalkSafe, Qwen2.5-3B · TEThelpfulness: 1.40 -> 2.58Improved
add risk-aware random walk instead of empirical threshold (ablation)WalkSafe, Qwen3-4B · TETavg harmlessness (lower better): 18.24 -> 15.63Improved
add risk-aware random walk instead of empirical threshold (ablation)WalkSafe, Llama-3.2-3B · TETavg harmlessness (lower better): 6.62 -> 3.25Improved
use Bi-GRPO instead of vanilla GRPO (ablation)WalkSafe, Qwen2.5-3B · TEThelpfulness: 1.02 -> 2.58Improved
use Bi-GRPO instead of vanilla GRPO (ablation)WalkSafe, Llama-3.2-3B · TETavg harmlessness (lower better): 3.41 -> 3.25Improved
use Bi-GRPO instead of vanilla GRPO (ablation)WalkSafe, Qwen3-4B · TETavg harmlessness (lower better): 18.06 -> 15.63Improved
apply WalkSafe post-trainingLlama-3.2-3B-Instruct · TETavg harmlessness (lower better): 11.22 -> 3.25Improved
apply WalkSafe post-trainingQwen3-4B · TETavg harmlessness (lower better): 24.73 -> 15.62Improved
apply WalkSafe post-trainingQwen2.5-3B-Instruct · TEThelpfulness: 0.04 -> 2.58Improved
apply STAIR safety trainingLlama-3.2-3B-Instruct · TEThelpfulness: -0.12 -> -0.49Hurt
apply STAIR safety trainingQwen2.5-3B-Instruct · TETavg harmlessness (lower better): 16.17 -> 7.09Improved

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