Research map / Adversarial robustness and security / Attacks on text and graphs
Time-To-Inconsistency: A Survival Analysis of Large Language Model Robustness to Adversarial Attacks
Yubo Li, Ramayya Krishnan, Rema Padman
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 Weibull AFT instead of Cox Baseline | time-to-inconsistency model · MT-Consistency test | IBS: 0.344 -> 0.180 | Improved |
| use Weibull AFT instead of Cox Baseline | time-to-inconsistency model · MT-Consistency test | C-index: 0.861 -> 0.874 | Improved |
| add model-drift interaction terms to Weibull AFT (ablation) | Weibull AFT · MT-Consistency test | IBS: 0.180 -> 0.175 | Improved |
| add model-drift interaction terms to Weibull AFT (ablation) | Weibull AFT · MT-Consistency test | C-index: 0.874 -> 0.869 | Hurt |
| add model-drift interaction terms to Cox (ablation) | Cox Baseline · MT-Consistency test | C-index: 0.861 -> 0.868 | Improved |
| add model-drift interaction terms to Cox (ablation) | Cox Baseline · MT-Consistency test | IBS: 0.344 -> 0.343 | No clear effect |
| use log-logistic instead of Weibull AFT family (ablation) | AFT · MT-Consistency test | IBS: 0.180 -> 0.187 | Hurt |
| use Weibull AFT instead of Random Survival Forest | survival model · MT-Consistency test | C-index: 0.845 -> 0.874 | Improved |
| use Weibull AFT instead of Random Survival Forest | survival model · MT-Consistency test | IBS: 0.190 -> 0.180 | Improved |
| use AFT instead of Cox for late-round calibration | time-to-inconsistency model · MT-Consistency test round 8 | Brier score: 0.446 -> 0.027 | Improved |
| use AFT-based monitor instead of drift-threshold baseline | risk monitor · failing conversations | % alerted: 62% -> 76% | Improved |
| use AFT-based monitor instead of drift-threshold baseline | risk monitor · censored conversations | % alerted: 32% -> 19% | Improved |
Closest papers by meaning
- On the Worst Prompt Performance of Large Language Models (NeurIPS 2024 · similarity 0.89)
- Remarkable Robustness of LLMs: Stages of Inference? (NeurIPS 2025 · similarity 0.87)
- Beyond Prompts: Dynamic Conversational Benchmarking of Large Language Models (NeurIPS 2024 · similarity 0.87)
- Found in Conversation: LLMs Teach Themselves to Close the Multi-Turn Gap (NeurIPS 2026 · similarity 0.87)
- Exploring Conversational Adaptability: Assessing the Proficiency of Large Language Models in Dynamic Alignment with Updated User Intent (AAAI 2025 · similarity 0.87)
- Measuring Non-Adversarial Reproduction of Training Data in Large Language Models (ICLR 2025 · similarity 0.87)
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.
