Newest papers

When LLMs Forget What They Believe: Characterizing State Integrity Collapse in Long-Horizon Dialogues

Yancheng Zhu, Yifan Yuan · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.31314 · cited by 0 works (OpenAlex) · open-access PDF

Indonesian

Kami mempelajari disosiasi pengingat-kewajiban ini dalam dialog jangka panjang dan mendefinisikan State Integrity Collapse: sebuah model mempertahankan keterbatasan yang dinyatakan dalam memori tetapi berhenti menerapkannya pada putaran berikutnya. Kami memperkenalkan StateTrack-Bench, sebuah suite yang dikendalikan dari percakapan yang mencakup numerik, format, peran/orang, memori faktual, dan keterbatasan aturan logika. Sepanjang tujuh snapshots model yang ditempatkan oleh penyedia dan lima probe kedalaman dialog, retensi bersama ketat berkisar dari 62.4% hingga 100.0%, dengan degradasi terkonsentrasi dalam keterbatasan simbolik dan dalam satu model yang sangat rapuh. Audit pengingat acak menunjukkan bahwa 88.6% dari kegagalan yang diamati adalah kegagalan perilaku yang dapat dikonfirmasi lebih tinggi daripada konstr

Machine translation (Meta M2M100); the English original follows.

Abstract

Large language models can repeat an instruction while failing to keep obeying it. We study this recall-compliance dissociation in long-horizon dialogue and define State Integrity Collapse: a model preserves a stated constraint in memory but stops applying it to later turns. We introduce StateTrack-Bench, a controlled suite of conversations covering numerical, format, role/persona, factual-memory, and logical-rule constraints. Across seven provider-pinned model snapshots and five dialogue-depth probes, strict joint retention ranges from 62.4% to 100.0%, with degradation concentrated in symbolic constraints and in one especially brittle model. A forked recall audit shows that 88.6% of observed failures are behavioral-drift failures rather than ordinary forgetting; excluding the brittle outlier preserves this diagnosis (84.4% of failures). Extended evaluations at depths up to 50 turns confirm that the dissociation persists beyond the core benchmark, while Structured State Anchoring recovers up to 24.4 points on immutable constraints. The results show that reliability cannot be inferred from single-turn instruction following alone; evaluation must track whether remembered constraints remain operational as dialogue state grows. Code and artifacts are available at: https://github.com/xzzvsxd/StateTrack.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Computer science, Suite, State (computer science), Computer security, Witness, Audit

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange