Publications
The record, citable.
The first entry is the project's real preliminary report (Oxunjon Ubaydullayev, April 2026): self-published, no DOI and no reference list yet, per its own AI-assistance statement. Everything below it is a planned record — work that is real to the project (the delivery-gap classes; the memory chain, requalified to "memory raises probability, it does not create capability"; the escalation chain, kills 0,1,1,1,3 across one chain, n=1; cross-model temperament) but not yet deposited. No DOIs, arXiv IDs, or version tags are shown for the planned records, because none are registered — listing a fabricated identifier, even labelled, has no place here. When real deposits land, these entries are replaced with the actual records. Verify against the deposited record, never against this list.
Memory-Grounded Social Dynamics in Repeated LLM Agent Simulations — Preliminary Behavioral Evaluation Report
Oxunjon Ubaydullayev
report · Preliminary behavioral evaluation report · self-published · no DOI yet · 2026
Simulation Room: A Platform for Studying Becoming in LLM Agents Across Restartable Lives
Oxunjon Ubaydullayev
preprint · Planned preprint — not yet submitted, no identifier assigned · 2026
The Delivery Gap: Five Classes of Failure Between Intention and Act in Simulated Agents
Oxunjon Ubaydullayev
dataset · Planned deposit — not yet registered, no DOI assigned · 2026
The Memory Chain: From Failure to Rescue, and Its Honest Requalification
Oxunjon Ubaydullayev
report · Planned report — not yet deposited, no DOI assigned · 2026
Escalation Under Raw Transfer: A Single-Chain Kill Trajectory (n=1) and the Case for a Digestion Organ
Oxunjon Ubaydullayev
report · Planned report — not yet deposited, no DOI assigned · 2026
Cross-Model Temperament: Differences in Becoming Across Language Models
Oxunjon Ubaydullayev
dataset · Planned deposit — not yet registered, no DOI assigned · 2026