Dear Special Call Editors, Please consider the manuscript "TaxBridge-Bench: A Neurosymbolic Benchmark for Spreadsheet-to-Ledger Reasoning in Tax Compliance" for the Special Call for Neurosymbolic Benchmark Papers. In the submission form, I will select the "Benchmark" manuscript type. The manuscript introduces TaxBridge-Bench, a benchmark for evaluating neurosymbolic systems on spreadsheet-to-ledger reasoning under UK Making Tax Digital tax-compliance constraints. The contribution is benchmark-first: it defines the task, datasets, metrics, scorer, baselines, and intended artefact package. The reference blackboard architecture is presented as one evaluated system rather than as the sole contribution. The benchmark targets a gap in current spreadsheet and LLM-agent evaluation: exact reconstruction of legally meaningful financial totals from messy workbooks where evidence is distributed across values, layout, formulas, formatting, dates, VAT treatment, and trade context. The paper reports traceable results for an LLM-first debate panel, a simple serialised-workbook Sonnet 5 baseline, and a symbolic-first blackboard reference system. It also discloses an earlier author-confirmed AutoGen-style pilot as an artefacts-pending comparator rather than as a primary reproducible baseline. I disclose that I am the founder of Neuro-Symbolic Ltd and the developer of the TaxBridge reference implementation evaluated in the manuscript. To support independent inspection, the supplementary package will include generated workbooks, executable specification files, ground truth, scoring code, prompts, raw baseline outputs, and scorer reports. Repository URL/DOI: [insert stable or anonymised review URL before submission]. The work uses synthetic benchmark data only; no raw customer data or human-participant data is included. Sincerely, David Elliman