Research map / Adversarial robustness and security / Attacks on text and graphs
REALISTA: Realistic Latent Adversarial Attacks that Elicit LLM Hallucinations
Buyun Liang, Jinqi Luo, Liangzu Peng, Kwan Ho Ryan Chan, Darshan Thaker, Kaleab A Kinfu, Fengrui Tian, Hamed Hassani, Rene Vidal
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 |
|---|---|---|---|
| optimize continuous combinations of semantic edit directions in latent space (REALISTA) instead of discrete rephrasing (SECA) | Llama-3-3B · MMLU subset (open-ended MCQA) | ASR@30 (%): 79.61 -> 97.11 | Improved |
| REALISTA instead of SECA | Llama-3-8B · MMLU subset (open-ended MCQA) | ASR@30 (%): 82.97 -> 93.60 | Improved |
| REALISTA instead of SECA | Qwen-2.5-7B · MMLU subset (open-ended MCQA) | ASR@30 (%): 32.47 -> 41.61 | Improved |
| REALISTA instead of SECA | Qwen-2.5-14B · MMLU subset (open-ended MCQA) | ASR@30 (%): 27.51 -> 27.24 | No clear effect |
| use LARGO or ICD instead of REALISTA | attack methods · MMLU subset (open-ended MCQA) | SEE (%): Llama-3-3B: REALISTA 0.86 vs LARGO 97.42 and ICD 100.00 | Hurt |
| apply REALISTA to commercial reasoning models with free-form responses | GPT-5-Nano · MMLU subset, free-form | ASR@30 (%): 4.02 -> 23.61 | Improved |
| apply REALISTA to commercial reasoning models with free-form responses | GPT-5-Mini · MMLU subset, free-form | ASR@30 (%): 2.01 -> 20.72 | Improved |
| use ICD template attack instead of REALISTA on reasoning models | GPT-5-Nano and GPT-5-Mini · MMLU subset, free-form | ASR@30 (%) and SEE: ICD ASR 6.32 and 2.57 with SEE 100.00 | Hurt |
| human annotators versus LLM judge for semantic equivalence of REALISTA prompts | REALISTA · Human evaluation (100 samples) | SEE (%): LLM 5.27, HA 5.22, HB 5.18 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- SECA — SECA: Semantically Equivalent and Coherent Attacks for Eliciting LLM Hallucinations (NeurIPS 2025) · better on 6 of 7 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- SECA: Semantically Equivalent and Coherent Attacks for Eliciting LLM Hallucinations (NeurIPS 2025 · similarity 0.95)
- Look Closer! An Adversarial Parametric Editing Framework for Hallucination Mitigation in VLMs (AAAI 2026 · similarity 0.90)
- HaDeMiF: Hallucination Detection and Mitigation in Large Language Models (ICLR 2025 · similarity 0.89)
- Poison as Cure: Visual Noise for Mitigating Object Hallucinations in LVMs (NeurIPS 2025 · similarity 0.89)
- In-Context Sharpness as Alerts: An Inner Representation Perspective for Hallucination Mitigation (ICML 2024 · similarity 0.89)
- LLM-Check: Investigating Detection of Hallucinations in Large Language Models (NeurIPS 2024 · similarity 0.88)
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.
