LinguaGuard — original wording balanced accuracy — structured step prompt: leaderboard
Original-wording TSBench AgentDojo: 1,220 steps, 868 safe, 352 unsafe, 589 trajectories, 58 distinct user instructions. Balanced accuracy of supplied-action risk classification. Condition: structured step prompt. Managed models three repeats; self-hosted greedy single run. Appendix E explicitly identifies Qwen3.8-27B FP8 and Sonnet5 thinking disabled. Native trajectory guards predict prefix-level risk against step labels and are not pooled with step-level guards. Scanner withholds user request. Provider defaults do not resolve unknown checkpoint variants; qualified names remain distinct. FPR/TPR and translation shifts are operating-point diagnostics, not standalone general-capability scores. Metric: Balanced accuracy. Higher is better. Source: https://arxiv.org/html/2610.22244v1
Metric: Balanced accuracy. Source: arxiv.org. Saturation forecast: Estimated already saturated. 14 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.8 Flash | 0.98 | #21 |
| 2 | Claude Sonnet 5 (Non-reasoning) | 0.96 | #54 (Claude Sonnet 5) |
| 3 | GPT-OSS-120B | 0.9 | #336 |
| 4 | Gemma 4 12B | 0.85 | #529 |
| 5 | GLM-5.2 | 0.84 | #71 |
| 6 | Nemotron 3.5 Lightning | 0.82 | #489 |
| 7 | Llama 4 Maverick | 0.82 | #464 |
| 8 | Qwen 3.5 4B | 0.81 | #469 |
| 9 | Qwen 3.5 9B | 0.8 | #373 |
| 10 | Gemma 4 E2B | 0.54 | #1099 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #7 Gemini 4 Argon.
Interactive version: theaggregate.ai/benchmark?slug=linguaguard-original-wording-balanced-accuracy-structured-step-prompt · How It Works · Data refreshed daily, snapshot 2026-10-17.