Research map / Large language models / Language, safety and interpretability
Model-Agnostic Sentiment Distribution Stability Analysis for Robust LLM-Generated Texts Detection
Siyuan Li, Xi Lin, Guangyan Li, Zehao Liu, Aodu Wulianghai, Li Ding, Jun Wu, Jianhua Li
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use sentiment distribution stability (SDC) instead of best baseline detector | detector on GPT-4-0613 text · five datasets avg | F1 (%): 63.62 -> 71.19 | Improved |
| use SentiDetect-SDC instead of best baseline | detector on Gemini-1.5-Pro text · five datasets avg | F1 (%): 61.53 -> 71.65 | Improved |
| use SentiDetect-SDC instead of best baseline | detector on LLaMa-3.3 text · five datasets avg | F1 (%): 72.37 -> 85.17 | Improved |
| use SentiDetect-SDP instead of RAIDAR | detector on GPT-4-0613 text · HumanEval | F1 (%): 67.48 -> 82.74 | Improved |
| use SentiDetect instead of baselines under adversarial perturbation | detector · HumanEval perturbed | F1 (%): 24.26 -> 48.05 | Improved |
| increase number of LER prompts I from 3 to 9 (ablation) | SentiDetect-SDC · Student Essay | F1 (%): 59.03 -> 56.60 | No clear effect |
| increase number of LER prompts I from 3 to 9 (ablation) | SentiDetect-SDC · Yelp Review | F1 (%): 77.02 -> 77.63 | No clear effect |
| use SentiDetect-SDC instead of GPT-Zero | detector on LLaMa-3.3 text · Paper | F1 (%): 87.61 -> 89.22 | No clear effect |
| shorten input text to about 20 words (ablation) | SentiDetect · detection length | F1 | Mixed |
Closest papers by meaning
- Training-free LLM-generated Text Detection by Mining Token Probability Sequences (ICLR 2025 · similarity 0.87)
- ReMoDetect: Reward Models Recognize Aligned LLM's Generations (NeurIPS 2024 · similarity 0.87)
- AdaDetectGPT: Adaptive Detection of LLM-Generated Text with Statistical Guarantees (NeurIPS 2025 · similarity 0.86)
- Human Texts Are Outliers: Detecting LLM-generated Texts via Out-of-distribution Detection (NeurIPS 2025 · similarity 0.86)
- DPIC: Decoupling Prompt and Intrinsic Characteristics for LLM Generated Text Detection (NeurIPS 2024 · similarity 0.86)
- DALD: Improving Logits-based Detector without Logits from Black-box LLMs (NeurIPS 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
