Skip to content
#small language model Book Open access

Persona-Isolated Evaluation of LLM-Based Dialogue Agents: Measuring History-Conditioned Response Variation

Nov 2026 · 0 citations

Abstract

Adaptive LLM-based agents can appear to adjust not only their answer content but also their response structure according to prior dialogue and user-stated preferences. This paper treats this phenomenon as a measurement problem, not as an attempt to reproduce naturalistic long-term human–agent interaction. We ask where history-conditioned user-side regularities become observable in generated responses under controlled histories. We introduce a branch-based design that separates Agent Persona (AP), the observable agent-side prior imposed by system constraints, from User Persona Observation (UPO), an operational label for response-level user-side regularities inferred from dialogue history. Fixed probes elicit responses under inter-lineage isolation without writing probe responses back into the main lineage. Using Japanese and English tracks as a bounded bilingual robustness check, we compare an embedding-based semantic-proxy readout with a heuristic rule-based presentation diagnostic. A reconstructed-context audit checks that the AP-absent and AP-present matched-history conditions use matched injected histories and differ only in whether the AP prior is included in the observable inputs. Results show that the AP-present condition is associated with smaller embedding-space drift, with clearer evidence in Japanese and weaker but directionally consistent evidence in English. The rule-based diagnostic overlaps with AP instructions, is strongly length-sensitive, and varies by language and scenario. These findings suggest that adaptive dialogue agents should be evaluated by separating semantic-proxy drift from communicative presentation, and that UPO should be treated as an operational response-level readout rather than as a stable user trait or internal model state.

Read PDF

Similar papers

#small language model Dataset Open access Oct 2026

Socratic guiding questions in synthetic arithmetic data: matched LoRA runs (revision v2)

Supporting data, adapters, predictions and code for the article *Low-Cost LoRA Fine-Tuning of Small Language Models for Multi-Step Arithmetic Reasoning* by Jake O'Grady, Asena Isik Gürhan, Chee Fong Ting and Effirul Ramlan (University of Galway). We generated 20,000 GSM8K-derived arithmetic problems with step-by-step s...

O'Grady, Jake, Gürhan, Asena Isik, Chee, Fong Ting et al. · 465 citations
#computer vision Open access Jun 2016

Software Development in Startup Companies: The Greenfield Startup Model

The results are packaged in the Greenfield Startup Model (GSM), which explains the priority of startups to release the product as quickly as possible, and the need to shorten time-to-market, by speeding up the development through low-precision engineering activities.

Carmine Giardino, Nicolò Paternoster, M. Unterkalmsteiner et al. · 178 citations · ⚡14
#computer vision Open access Oct 2016

Software Startups - A Research Agenda

Software startup companies develop innovative, software-intensive products within limited timeframes and with few resources, searching for sustainable and scalable business models.

M. Unterkalmsteiner, P. Abrahamsson, Xiaofeng Wang et al. · 157 citations · ⚡17
#machine learning Review Open access Oct 2016

“Failures” to be celebrated: an analysis of major pivots of software startups

This study conducts a case survey study based on the secondary data of the major pivots happened in 49 software startups, and demonstrates that customer need pivot is the most common among all pivot types.

Sohaib Shahid Bajwa, Xiaofeng Wang, Anh Nguyen-Duc et al. · 127 citations · ⚡15
#computer vision Review Open access May 2015

A survey study on major technical barriers affecting the decision to adopt cloud services

The comparison of adopter and non-adopter sample reveals three potential adoption inhibitor, security, data privacy, and portability, which underlines the importance of the technical and security perspectives for research investigating the adoption of technology.

Nattakarn Phaphoom, Xiaofeng Wang, S. Samuel et al. · 111 citations · ⚡8
#large language models Review Open access Nov 2020

Status of the WHO recommended timing and frequency of antenatal care visits in Northern Bangladesh

Objective There is dearth of information on the timeliness of antenatal care (ANC) uptake. This study aimed to determine the timely ANC uptake by a medically trained provider (MTP) as per the World Health Organization (WHO) recommendations and the country guideline. Methods Cross-sectional survey was done with 2,731 wo...

B. Sarker, Musfikur Rahman, T. Rahman et al. · 89 citations · ⚡7

Related blog posts

Microsoft Research Blog Nov 16, 2026

TokenEconomizer: Learning from Experience to Make AI Coding Agents More Efficient

AI coding agents are increasingly help developers fix bugs and build features largely on their own, but every interaction with their models consumes tokens. In each session, an agent sends instructions, tool descriptions, and session history to its model. Most efforts to reduce this cost focus on individual tasks. Yet developers return to the same […] The post TokenEconomizer: Learning from Experience to Make AI Coding Agents More Efficient appeared first on Microsoft Research.

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.