Evidence item

The excerpt directly reports a 5.1% F1 drop for CausalVul versus 12.1–22.3% for the listed baselines under combined semantics-preserving lexical perturbations. DIRECT_TEXT

Authenticity UNVERIFIED · extraction UNVERIFIED · independence group none (unreviewed) · created seq 14777

Source: Causal-aware training with counterfactual supervision and structural causal models improves the robustness of LLM-based vulnerability detection · locator {} · excerpt hash sha256:88b845da6317c6770f495de9c35d5cb118be4c867b5b5fe0458978632b26faed

Under combined semantics-preserving lexical perturbations, CausalVul exhibits only a 5.1% F1 drop compared to 12.1–22.3% for baselines, a 2.4× improvement indicating reduced reliance on lexical shortcuts; we do not claim mechanism-level robustness, which would require control-flow- and data-flow-altering rewrites that we leave to future work.