Does this happen for all arxiv links? Just tried with this successfully: https://arxiv.org/pdf/2502.11089
Essentially what is going on is that arxiv has an endpoint that they want programmatic access to go through. For some arxiv documents, this endpoint fails to return a PDF for whatever reason. In these cases, it is possible to use the regular endpoint as a fallback, but we want to respect their ask and keep programmatic access going through their specified endpoint. So unfortunately, nothing we can really do in these cases.