- Rating
- 2
- Confidence
- 3
- Informativeness
- 2
- Presentation
- 2
- Soundness
- 1
Read review
Paper summary
This paper introduces TotalGuardEval, a guardrail benchmark covering 16 risk categories, paired safe samples, and data on both the request and response sides. Using 7,260 synthetic samples, the authors evaluate 10 public models under original, obfuscated, and adversarial conditions, with a focus on missed detections, over-blocking, and adversarial robustness.
Critical strengths
- Safe and harmful samples are paired across both the request and response sides, allowing misses and over-blocking to be measured within the same topic.
- A broad range of models is covered, with per-category FPR/FNR reported rather than a single aggregate score; the data audit and limitation discussion are also transparent.
Critical weaknesses
- All samples are generated by a single model and only 264 are validated by two authors sharing the design priors, leaving label reproducibility for culture-sensitive categories unsubstantiated.
- The 88 attacks are not verified to preserve harmful intent and cannot be independently reconstructed.
- Highly correlated variants of the same request are counted as independent items in the attack F1. Please report results clustered by source request.
Questions for authors
- Could reproducible attack samples be released, with manual verification that each attack preserves the original harmful intent?
- Could each condition be evaluated with matched category proportions or balanced accuracy? And how are system prompts and output parsing standardized across models?
Thank you for raising reproducibility and the unit of analysis. Appendix P documents the request-side transformations. The attack suite combines methods from NVIDIA’s garak and HiveTrace Red with custom transformations. Researchers can rerun the publicly implemented methods on the original requests using the corresponding frameworks. To replay the exact attack inputs used for Table 4, a release would need to include each evaluated rewrite, its source request, and the transformation ID. Manual review of those same pairs would establish which rewrites preserve the original harmful intent.
The original corpus is balanced within each pair of risk and safe categories (Table 9). The obfuscation and attack conditions use a fixed safe reference and add variants of harmful requests, so the class proportions differ across conditions (Appendix M). Balanced accuracy would weight harmful and safe recall equally. The submitted per-category FNR and FPR already separate missed detections from over-refusals. Table 4 reports attack scores over individual rewrites. A source-weighted analysis assigns each original request the same total weight across its rewrites and treats those rewrites as one cluster when estimating uncertainty.
As described in Section 5, we used each model’s default classification template and supplied the TotalGuardEval taxonomy to models that support custom policies. The models emitted safety labels directly, and we successfully mapped every output from the reported runs to SAFE or UNSAFE. All models received the same attack inputs. The blinded re-annotation in Appendix N measures internal consistency under the written rubric; external validation of the culturally sensitive categories is a separate question.