# AFA: Identity-Aware Memory for Preventing Persona Confusion in Multi-User Dialogue

Source: [AFA: Identity-Aware Memory for Preventing Persona Confusion in Multi-User Dialogue](<https://www.sota2.com/research/paper/69f1907855a3c32af89fb9cf>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-04-30T05:00:04.704Z

Authors: Mohammad Al-Ratrout, Pavan Uttej Ravva, Shayla Sharmin, Aditya Raikwar, Ju Young Shin, Roghayeh Leila Barmaki

Published: 2026-04-27T21:55:03.000Z

## Abstract

When multiple people share a single voice assistant, the system conflates their histories: one resident's preferences can leak into another's responses, eroding utility and trust. We call this failure mode persona confusion, and we show it is a measurable problem in today's single-user dialogue systems when deployed in shared environments. We present the Adaptive Friend Agent (AFA), a modular framework that combines voice-based speaker identification with per-user memory stores to enable identity-aware, personalized dialogue across multiple users. To support training and evaluation, we construct PAT (Personalized Agent chaT), a synthetic dataset of 58,289 persona-grounded dialogue turns spanning 133 user profiles and 12 real-world scenarios. We evaluate AFA across five LLM back-ends in a standard response-quality benchmark, with a LLaMA-2-70B model fine-tuned on PAT achieving the highest overall performance. To directly measure persona confusion prevention, we introduce an interleaved multi-user evaluation protocol with a novel metric, Persona Attribution Accuracy (PAA), demonstrating that identity-aware routing improves PAA from 35.7% to 61.3%. Human evaluation confirms annotators perceive significantly higher personalization in routing-enabled responses. Our results establish that identity-aware user routing is the critical component for preventing persona confusion in multi-user conversational systems.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Multi-user Dialogue Response Generation on PAT](<https://www.sota2.com/research/sota/multi-user-dialogue-response-generation-on-pat.md>): Multi-user Dialogue Response Generation; PAT; P-Cov: 46.3.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2604.25022>)

[PDF](<https://arxiv.org/pdf/2604.25022.pdf>)