[2603.03417] Parallel Test-Time Scaling with Multi-Sequence Verifiers

Computer Science > Cryptography and Security arXiv:2603.03417 (cs) [Submitted on 3 Mar 2026] Title:Parallel Test-Time Scaling with Multi-Sequence Verifiers Authors:Yegon Kim, Seungyoo Lee, Chaeyun Jang, Hyungi Lee, Juho Lee View a PDF of the paper titled Parallel Test-Time Scaling with Multi-Sequence Verifiers, by Yegon Kim and 4 other authors View PDF HTML (experimental) Abstract:Parallel test-time scaling, which generates multiple candidate solutions for a single problem, is a powerful technique for improving large language model performance. However, it is hindered by two key bottlenecks: accurately selecting the correct solution from the candidate pool, and the high inference latency from generating many full solutions. We argue that both challenges are fundamentally linked to verifier calibration. A well-calibrated verifier not only improves answer selection, but also enables early-stopping strategies to reduce latency. However, existing verifiers are limited as they score each candidate in isolation, overlooking rich contextual information across the set of candidates. To address this, we introduce the Multi-Sequence Verifier (MSV), the first verifier designed to jointly process all candidate solutions and model their interactions. MSV achieves improved calibration, which directly enhances best-of-N selection performance. We further introduce a streaming MSV variant that empowers a novel early-stopping framework. Our novel framework fully leverages parallel decoding,...