List your company

Toxicity Detection on RealTox Challenging

98.92AccuracyTrajectory (Raw)
87.209690.249893.2996.3302Mar 1, 2026

Evaluation results

MethodLinks
Trajectory (Raw)Backbone=Qwen3-30B MoE2026.0398.92
Trajectory (Raw)Backbone=Qwen2.5-14B2026.0398.83
Trajectory (Raw)Backbone=Qwen3-32B2026.0398.83
TaT (Disp.)Backbone=Qwen2.5-14B2026.0398.58
Trajectory (Raw)Backbone=Llama3.1-8b2026.0397.91
TaT (Disp.)Backbone=Llama3.1-8b2026.0396
Linear ProbeBackbone=Llama3.1-8b2026.0395.83
Linear ProbeBackbone=Qwen2.5-14B2026.0395.16
TaT (Disp.)Backbone=Qwen3-32B2026.0395.08
Linear ProbeBackbone=Qwen3-30B MoE2026.0394.5
Linear ProbeBackbone=Qwen3-32B2026.0391.16
TaT (Disp.)Backbone=Qwen3-30B MoE2026.0387.66