# Applicable License And Attribution ## FinConRank Release Decision The author and coauthors explicitly confirmed Apache License, Version 2.0 for FinConRank. The full standard license text is in LICENSE. This packaging applies that decision to FinConRank's own modifications, domain-adapted LoRA collection, and accompanying original release material, subject to third-party rights. No copyright holder names or third-party authorizations have been invented. ## Upstream Components | Component | Role | Applicable license evidence | | --- | --- | --- | | Qwen/Qwen2-VL-2B-Instruct | Base model and processor; processor reference files included | Historical local resource inventory records Apache-2.0; upstream license text was not in the cached snapshot | | lightonai/MonoQwen2-VL-v0.1 | Initial LoRA continued during FinConRank training | Historical local resource inventory records Apache-2.0; upstream license text was not in the cached snapshot | NOTICE preserves upstream attribution and exact revisions. The base model is not bundled. Continued-training weights do not erase upstream rights. The author's license decision is not independent confirmation of upstream permissions or a replacement for any required upstream notices. Exact upstream license/model-card texts and any upstream NOTICE obligations have not been independently reverified in this offline packaging task. Recipients/distributors should retain applicable upstream notices and check the pinned upstream resources when redistributing. No model, data, dependency or license document was downloaded for this task. ## Other Material And Boundaries Benchmark reports, images, qrels and training data are not included. This license does not grant rights to third-party financial reports or user-supplied documents. Dependencies retain their own licenses and are not relicensed by requirements.txt. The Apache license text was copied verbatim from an existing local standard license file; it is not a recovered model-specific license or upstream notice. This release is a collection of ten fold-specific adapters, not a unified all-data model. Evaluation used query grouping within a shared fixed document corpus. It does not demonstrate strict unseen-page or unseen-report generalization.