Metraipolis › archive › dc_da468c4bdf74

The Analyst in the Prompt: Role, Retrieval, and Memory Biases in LLM Financial Analysis — arXiv:2609.03218

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.03218
License (as archived)
CC0-1.0
Archived
2026-10-17T01:40:07.371Z
SHA-256 of the archived text
e5001a3002c51c7c191e0a5b84251c3a18c7b9b77b25a2fa6d8ba700dafbc67f
Stable URL of this page
https://metraipolis.com/documento/dc_da468c4bdf74

How to cite

The Analyst in the Prompt: Role, Retrieval, and Memory Biases in LLM Financial Analysis — arXiv:2609.03218 — arXiv (en), https://arxiv.org/abs/2609.03218 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_da468c4bdf74 (Metraipolis archive dc_da468c4bdf74, sha256 e5001a3002c51c7c191e0a5b84251c3a18c7b9b77b25a2fa6d8ba700dafbc67f)
BibTeX
@misc{dcda468c4bdf74,
  title = {The Analyst in the Prompt: Role, Retrieval, and Memory Biases in LLM Financial Analysis — arXiv:2609.03218},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.03218},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_da468c4bdf74; sha256 e5001a3002c51c7c191e0a5b84251c3a18c7b9b77b25a2fa6d8ba700dafbc67f},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.03218. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Large Language Models (LLMs) increasingly use user context such as memory, profiles, and role prompts to personalize their responses. This personalization can affect evidence-based judgment: the same evidence may lead to different conclusions under different user contexts. Finance provides a high-stakes setting to study this problem because decisions often depend on interpreting long and complex documents. We test this using 3,575 SEC filings across twelve LLMs. We compare persona-conditioned retrieval, neutral retrieval, and memory-framed context to separate the effect of evidence selection from the effect of interpretation. We find that most user-context spillover comes from how models interpret the same evidence under different roles, rather than from retrieving different evidence. We then test two simple mitigation strategies: expressing the same investor mindset as a user profile instead of an assistant role, and separating evidence-based and personalized outputs. Both reduce spillover, but neither removes it completely, and their effectiveness varies substantially across models.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-04 in cs.CL).
Categorie arXiv: Computation and Language (cs.CL) ; Computational Engineering, Finance, and Science (cs.CE); Portfolio Management (q-fin.PM).
Pagina del lavoro: https://arxiv.org/abs/2609.03218 — 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 2026-10-17T01:11:43.509Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-09-04?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).