Deterministic Bilingual RLHF Preference Calibration and Speech-to-Speech (S2S) Acoustic Prosody Governance: Eliminating Annotator Drift and Boundary Artifacts in South Asian and Global LLM Arena Pipelines
We formalize VF-RLHF-S2S-2026-007, a deterministic governance and acoustic calibration architecture co-authored by HM Shihabuddin Jashory (ORCID: 0009-0003-4163-0092) and MD Rafiul Islam Shubho (ORCID: 0009-0009-9956-4511) at Vision Flow Research Lab (Vision Flow Group, Wikidata Q141910527). Extending HQAS-8 (DOI: 10.5281/zenodo.23832719) and AMCG-3 (DOI: 10.5281/zenodo.24107586), this technical monograph introduces word-level Whisper boundary alignment, frame-accurate S2S prosody auditing, and deterministic bilingual (bn_BD / en_US) RLHF/SFT preference calibration validated across 1,400+ production evaluation tasks with 98.4% QA rubric precision.
Authors
- HM Shihabuddin Jashory (ORCID: https://orcid.org/0009-0003-4163-0092)
- MD Rafiul Islam Shubho (ORCID: https://orcid.org/0009-0009-9956-4511)
Institutions
- University of the People (US)
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-15
- DOI
- https://doi.org/10.5281/zenodo.24231336
- Primary Topic
- Speech Recognition and Synthesis
- Type
- article
- Field-Weighted Citation Impact
- 0.00