# Listening Alone, Understanding Together: Collaborative Context Recovery for Privacy-Aware AI

Source: [Listening Alone, Understanding Together: Collaborative Context Recovery for Privacy-Aware AI](<https://www.sota2.com/research/paper/69e06d1b7d0172140ec7cab7>)

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

Data updated: 2026-04-17T05:00:04.441Z

Authors: Tanmay Srivastava, Amartya Basu, Shubham Jain, Vaishnavi Ranganathan

Published: 2026-04-14T23:18:06.000Z

## Abstract

We introduce CONCORD, a privacy-aware asynchronous assistant-to-assistant (A2A) framework that leverages collaboration between proactive speech-based AI. As agents evolve from reactive to always-listening assistants, they face a core privacy risk (of capturing non-consenting speakers), which makes their social deployment a challenge. To overcome this, we implement CONCORD, which enforces owner-only speech capture via real-time speaker verification, producing a one-sided transcript that incurs missing context but preserves privacy. We demonstrate that CONCORD can safely recover necessary context through (1) spatio-temporal context resolution, (2) information gap detection, and (3) minimal A2A queries governed by a relationship-aware disclosure. Instead of hallucination-prone inferring, CONCORD treats context recovery as a negotiated safe exchange between assistants. Across a multi-domain dialogue dataset, CONCORD achieves 91.4% recall in gap detection, 96% relationship classification accuracy, and 97% true negative rate in privacy-sensitive disclosure decisions. By reframing always-listening AI as a coordination problem between privacy-preserving agents, CONCORD offers a practical path toward socially deployable proactive conversational agents.

## 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.

- [Hybrid Decision Gate on CONCORD Evaluation Set](<https://www.sota2.com/research/sota/hybrid-decision-gate-on-concord-evaluation-set.md>): Hybrid Decision Gate; CONCORD Evaluation Set; TNR: 97.

- [Info Gap Detection on CONCORD Evaluation Set](<https://www.sota2.com/research/sota/info-gap-detection-on-concord-evaluation-set.md>): Info Gap Detection; CONCORD Evaluation Set; False Positive Rate (FPR): 6.8.

- [Local Reference Resolution on CONCORD Evaluation Set](<https://www.sota2.com/research/sota/local-reference-resolution-on-concord-evaluation-set.md>): Local Reference Resolution; CONCORD Evaluation Set; TPR: 78.3.

- [Relationship Assessment on CONCORD Evaluation Set](<https://www.sota2.com/research/sota/relationship-assessment-on-concord-evaluation-set.md>): Relationship Assessment; CONCORD Evaluation Set; Classification Accuracy: 96.4.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2604.13348>)

[PDF](<https://arxiv.org/pdf/2604.13348.pdf>)