Research map / Adversarial robustness and security / Attacks on text and graphs
Semantics-Preserving Adversarial Attacks on Event-Driven Stock Prediction Models
Aofan Liu, Haoxuan Li, Hongjian Xing, Yuguo Yin, Zijun Li, Yiyan Qi
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 |
|---|---|---|---|
| add Semantic Translation Module (Stage 2) (ablation) | ChameleonAttack · FinBERT FPB | naturalness MOS: 1.3 -> 4.5 | Improved |
| add Semantic Translation Module (Stage 2) (ablation) | ChameleonAttack · FinBERT FPB | attack success rate: 93.1 -> 92.5 | No clear effect |
| add Adaptive Sparsification instead of argmax projection (ablation) | ChameleonAttack · FinBERT FPB | attack success rate: 72.8 -> 92.5 | Improved |
| use adaptive sparsity instead of fixed sparsity target (ablation) | ChameleonAttack · FinBERT FPB | attack success rate: 81.5 -> 92.5 | Improved |
| use gradient-based continuous optimization instead of random search (ablation) | ChameleonAttack · FinBERT FPB | attack success rate: 12.3 -> 92.5 | Improved |
| add Semantic Translation Module (Stage 2) (ablation) | ChameleonAttack · FinBERT FPB | semantic similarity: 0.25 -> 0.85 | Improved |
| apply perplexity filtering defense | ChameleonAttack vs GCG Attack · FinBERT FPB | attack success rate: 88.0 vs 60.5 | Mixed |
| apply adversarial detector defense | ChameleonAttack · FinBERT FPB | attack success rate: 92.5 -> 85.0 | Hurt |
| apply adversarial training to FinBERT | ChameleonAttack · FinBERT FPB | attack success rate: 92.5 -> 75.0 | Hurt |
| compare ChameleonAttack to AmpleGCG | attack method · FinBERT Twitter Positive | attack success rate: 85.6 -> 93.4 | Improved |
| attack Qwen3-8B FinRobot agent | FinRobot · agent accuracy | accuracy: 89.4% to 36.2% | Hurt |
Closest papers by meaning
- Efficient Adversarial Training in LLMs with Continuous Attacks (NeurIPS 2024 · similarity 0.87)
- Semantic Representation Attack against Aligned Large Language Models (NeurIPS 2025 · similarity 0.86)
- An LLM can Fool Itself: A Prompt-Based Adversarial Attack (ICLR 2024 · similarity 0.86)
- RigorLLM: Resilient Guardrails for Large Language Models against Undesired Content (ICML 2024 · similarity 0.86)
- REINFORCE Adversarial Attacks on Large Language Models: An Adaptive, Distributional, and Semantic Objective (ICML 2025 · similarity 0.86)
- AutoAdvExBench: Benchmarking Autonomous Exploitation of Adversarial Example Defenses (ICML 2025 · similarity 0.85)
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.
