Self-Harm Detection on JiraiBench 1.0 (test)
66.63Macro F1SAS (English version)
Evaluation results
| Method | Links | |
|---|---|---|
| SAS (English version)LLM Backbone=DeepSeek V3.22026.01 | 66.63 | |
| S^3 AgentLLM Backbone=DeepSeek V3.22026.01 | 61.8 | |
| OWLLLM Backbone=DeepSeek V3.22026.01 | 59.75 | |
| Self-RefineLLM Backbone=DeepSeek V3.22026.01 | 57.92 | |
| Zero-shotLLM Backbone=DeepSeek V3.22026.01 | 57.33 | |
| Zero-shot CoTLLM Backbone=DeepSeek V3.22026.01 | 57.02 | |
| Plan-and-SolveLLM Backbone=DeepSeek V3.22026.01 | 55.91 | |
| SAS (English version)LLM Backbone=Ministral-3-8B2026.01 | 53.89 | |
| Self-RefineLLM Backbone=Gemma-3-12B-it2026.01 | 53.33 | |
| SAS (English version)LLM Backbone=Qwen-2.5-7B2026.01 | 50.32 | |
| Zero-shot CoTLLM Backbone=Qwen-2.5-7B2026.01 | 49.99 | |
| Plan-and-SolveLLM Backbone=Qwen-2.5-7B2026.01 | 49.92 | |
| SAS (English version)LLM Backbone=Gemma-3-12B-it2026.01 | 49.38 | |
| Self-RefineLLM Backbone=Ministral-3-8B2026.01 | 49.3 | |
| Zero-shot CoTLLM Backbone=Ministral-3-8B2026.01 | 48.86 | |
| Zero-shotLLM Backbone=Ministral-3-8B2026.01 | 48.33 | |
| Plan-and-SolveLLM Backbone=Ministral-3-8B2026.01 | 47.77 | |
| S^3 AgentLLM Backbone=Qwen-2.5-7B2026.01 | 47.18 | |
| Zero-shot CoTLLM Backbone=Gemma-3-12B-it2026.01 | 47.05 | |
| OWLLLM Backbone=Qwen-2.5-7B2026.01 | 47.03 | |
| OWLLLM Backbone=Ministral-3-8B2026.01 | 46.57 | |
| S^3 AgentLLM Backbone=Gemma-3-12B-it2026.01 | 46.03 | |
| Plan-and-SolveLLM Backbone=Gemma-3-12B-it2026.01 | 45.94 | |
| Zero-shotLLM Backbone=Gemma-3-12B-it2026.01 | 45.92 | |
| S^3 AgentLLM Backbone=Ministral-3-8B2026.01 | 45.82 | |
| SAS (English version)LLM Backbone=Llama-3.1-8B2026.01 | 44.48 | |
| OWLLLM Backbone=Gemma-3-12B-it2026.01 | 43.12 | |
| Zero-shotLLM Backbone=Qwen-2.5-7B2026.01 | 39.71 | |
| Zero-shotLLM Backbone=Llama-3.1-8B2026.01 | 39 | |
| Self-RefineLLM Backbone=Llama-3.1-8B2026.01 | 38.83 | |
| Zero-shot CoTLLM Backbone=Llama-3.1-8B2026.01 | 36.29 | |
| OWLLLM Backbone=Llama-3.1-8B2026.01 | 35.83 | |
| Self-RefineLLM Backbone=Qwen-2.5-7B2026.01 | 34.8 | |
| S^3 AgentLLM Backbone=Llama-3.1-8B2026.01 | 34.72 | |
| Plan-and-SolveLLM Backbone=Llama-3.1-8B2026.01 | 32.74 |