diff --git a/docs/benchmarks.md b/docs/benchmarks.md
index 15ea3531..6bbdade5 100644
--- a/docs/benchmarks.md
+++ b/docs/benchmarks.md
@@ -35,8 +35,8 @@ Available datasets:
 | `repliqa` | ~30MB |
 | `hotpotqa` | ~331MB |
 | `wix` | ~511MB |
-| `orb_text` — OpenRAG Bench, text embedder (`qwen3-embedding:4b`) with VLM picture descriptions baked into chunk content | ~15GB |
-| `orb_multimodal` — OpenRAG Bench, multimodal embedder (`qwen3-vl-embedding-8b`); picture vectors live in the same space as text for cross-modal retrieval | ~16GB |
+| `orb_text` — OpenRAG Bench, text embedder (`qwen3-embedding:4b`) with VLM picture descriptions baked into chunk content | ~18 GB |
+| `orb_multimodal` — OpenRAG Bench, multimodal embedder (`qwen3-vl-embedding-8b`); picture vectors live in the same space as text for cross-modal retrieval | ~16 GB |
 
 After downloading, run benchmarks with `--skip-db` to use the pre-built database:
 
diff --git a/evaluations/evaluations/benchmark.py b/evaluations/evaluations/benchmark.py
index 09d4a7cc..4392b9c1 100644
--- a/evaluations/evaluations/benchmark.py
+++ b/evaluations/evaluations/benchmark.py
@@ -771,9 +771,14 @@ def upload(
     for the multi-GB ORB databases which would otherwise abort on any transient
     network failure under plain ``upload_folder``.
 
-    The local folder basename equals ``spec.db_filename`` (see
-    ``DatasetSpec.db_path``), so the folder lands at that path on the Hub.
+    ``upload_large_folder`` has no ``path_in_repo`` — it ships the contents of
+    ``folder_path`` to the repo root. Stage the db under a temp parent with
+    hardlinks so the basename becomes the remote path, leaving everything
+    else at the root undisturbed.
     """
+    import os
+    import tempfile
+
     specs = _resolve_datasets(dataset)
 
     api = HfApi()
@@ -796,12 +801,23 @@ def upload(
         except Exception:
             pass
 
-        console.print(f"[blue]Uploading {spec.key} ({db})...[/blue]")
-        api.upload_large_folder(
-            folder_path=str(db),
-            repo_id=HF_REPO_ID,
-            repo_type="dataset",
-        )
+        with tempfile.TemporaryDirectory() as staging:
+            target = Path(staging) / spec.db_filename
+            target.mkdir()
+            for src in db.rglob("*"):
+                if not src.is_file():
+                    continue
+                rel = src.relative_to(db)
+                dest = target / rel
+                dest.parent.mkdir(parents=True, exist_ok=True)
+                os.link(src, dest)
+
+            console.print(f"[blue]Uploading {spec.key} ({db})...[/blue]")
+            api.upload_large_folder(
+                folder_path=staging,
+                repo_id=HF_REPO_ID,
+                repo_type="dataset",
+            )
 
         console.print(f"[green]Uploaded {spec.key} to {HF_REPO_ID}[/green]")