Metraipolis › archive › dc_7d2e95b6320f

MIRAGE: How Conversation State Shapes Historical Evidence Use in Multimodal Personal Agents — arXiv:2609.19059

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.19059
License (as archived)
CC0-1.0
Archived
2026-10-15T13:49:42.674Z
SHA-256 of the archived text
c4fb3fd3a3da45f12588601b8d728a72e187753b20f461378f7f4f5542854b78
Stable URL of this page
https://metraipolis.com/documento/dc_7d2e95b6320f

How to cite

MIRAGE: How Conversation State Shapes Historical Evidence Use in Multimodal Personal Agents — arXiv:2609.19059 — arXiv (en), https://arxiv.org/abs/2609.19059 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_7d2e95b6320f (Metraipolis archive dc_7d2e95b6320f, sha256 c4fb3fd3a3da45f12588601b8d728a72e187753b20f461378f7f4f5542854b78)
BibTeX
@misc{dc7d2e95b6320f,
  title = {MIRAGE: How Conversation State Shapes Historical Evidence Use in Multimodal Personal Agents — arXiv:2609.19059},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.19059},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_7d2e95b6320f; sha256 c4fb3fd3a3da45f12588601b8d728a72e187753b20f461378f7f4f5542854b78},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.19059. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Multimodal large language model (MLLM) agents are increasingly used as personal assistants for long-running tasks. Their utility depends on continuity: agents must retrieve and use earlier evidence across dialogue, files, and workspace state. However, agents can generate plausible answers even when access to that history has degraded, causing outcome-only evaluation to overestimate true evidence use. We present MIRAGE (Multimodal Interaction Retrieval, Attribution, and Grounding Evaluation), a controlled study of historical evidence use under conversation-state variation in multimodal personal agents. MIRAGE holds evidence objects, questions, and scoring fixed while varying only conversation state, and evaluates whether an agent can determine answerability, recover the correct source, and answer from it. Across seven frontier and open-weight multimodal backbones, we find that: 1) pre-compaction depth and post-compaction continuation form distinct, non-monotonic failure regimes rather than a single degradation curve; 2) open-weight models rely heavily on context continuity and are reluctant to spontaneously switch to tool-mediated retrieval when provenance fails; and 3) retrieval pressure improves source attribution in deep pre-compaction states for tool-compliant models, but consistently regresses after compaction, where stored evidence has already degraded. These findings show that historical evidence use should be evaluated under state variation, rather than inferred from outcome-only correctness.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-17 in cs.CL).
Categorie arXiv: Computation and Language (cs.CL) ; Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2609.19059 — 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 2026-10-15T13:48:24.368Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-09-17?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).