INSURE-Dial: A Phase-Aware Conversational Dataset & Benchmark for Compliance Verification and Phase Detection
Shubham Kulkarni, Alexander Lyzhov, Preetam Joshi, Shiva Chaitanya · 2026
Read INSURE-Dial: A Phase-Aware Conversational Dataset & Benchmark for Compliance Verification and Phase Detection by Shubham Kulkarni, Alexander Lyzhov, Preetam Joshi, Shiva Chaitanya (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-18448.txt.
Description
Shubham Kulkarni and three co-authors build a benchmark for auditing insurance calls. Their 2026 paper notes administrative phone work drains roughly one trillion dollars annually from United States healthcare, with over five hundred million benefit verification calls handled by hand in 2024. No public benchmark existed for this phase-aware call auditing.
The team introduces INSURE-Dial: fifty de-identified AI-initiated calls with live representatives, averaging seventy-one turns. One thousand synthetic calls mirror that workflow. Calls carry a phase-structured schema: patient identification, IVR navigation, coverage status, medication checks, and agent identification. Each phase gets labels for information and procedural compliance. Two tasks are defined: phase boundary detection and compliance verification. Per-phase scores are strong for small baselines, but end-to-end reliability is limited by span-boundary errors. Full-call exact segmentation on real calls stays low, exposing a gap between fluency and audit-grade evidence.
Catalog record
| CREATOR | Shubham Kulkarni, Alexander Lyzhov, Preetam Joshi, Shiva Chaitanya |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 104 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-18448