Metraipolis › archive › dc_062e863c18ff

Benchmarking Factual Robustness of LLMs via Multi-conversation Persuasion — arXiv:2609.16777

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.16777
License (as archived)
CC0-1.0
Archived
2026-10-15T19:19:19.013Z
SHA-256 of the archived text
1be08418ada42d5aab5a9851ed2c6dc67ef683b37dcfa747aaa095a66d90a9e8
Stable URL of this page
https://metraipolis.com/documento/dc_062e863c18ff

How to cite

Benchmarking Factual Robustness of LLMs via Multi-conversation Persuasion — arXiv:2609.16777 — arXiv (en), https://arxiv.org/abs/2609.16777 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_062e863c18ff (Metraipolis archive dc_062e863c18ff, sha256 1be08418ada42d5aab5a9851ed2c6dc67ef683b37dcfa747aaa095a66d90a9e8)
BibTeX
@misc{dc062e863c18ff,
  title = {Benchmarking Factual Robustness of LLMs via Multi-conversation Persuasion — arXiv:2609.16777},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.16777},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_062e863c18ff; sha256 1be08418ada42d5aab5a9851ed2c6dc67ef683b37dcfa747aaa095a66d90a9e8},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.16777. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

As Large Language Models (LLMs) increasingly serve as primary knowledge retrieval interfaces, their robustness against \textit{persuasion attacks}---attempts to inject misinformation or enforce counterfactuals---has become a critical safety concern. Existing red-teaming frameworks typically evaluate models in multi-turn dialogues where the target model retains full conversation history. We identify a critical flaw in this setting termed \textbf{``Refusal Inertia''}: a model's initial refusal often propagates through subsequent turns largely to maintain contextual consistency, thereby masking its true vulnerability to sophisticated, isolated persuasion attempts. To rigorously evaluate the ``cold-start'' defense capabilities of SOTA models, we introduce the \textbf{SAST-IR} (Stateful Attacker, Stateless Target - Iterative Refinement) framework. By enforcing a memory wipe on the target while retaining the attacker's history, we simulate a worst-case adversarial setting using \textbf{multi-turn} (stateless) iterations. Leveraging \textbf{CP-Agent} (Cognitive Persuasion Agent), an enhanced diagnosis-guided agent, our experiments on the custom \textsc{CounterFact-Strict} dataset ($N=50$) yield alarming results: simple, diverse attack strategies achieved a staggering \textbf{96\%} success rate, exposing severe brittleness in memory-less defense. Furthermore, we reveal a \textbf{``Complexity Paradox''}: while complex, iteratively refined attacks are effective, they often trigger defensive compliance, whereas simple strategies achieve a higher rate of genuine persuasion (\textbf{84.7\%}). Our code and dataset are available at GitHub, this https URL .

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-16 in cs.CL).
Categorie arXiv: Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2609.16777 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-15T18:32:32.829Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-09-16?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).