Does Denser AI flag an unprocessable upload as failed?
Denser AI reported an unreadable truncated PDF as ready rather than flagging its failure. It did flag a password-protected PDF as failed in the same upload.
1 of 1 test case failed
Every test case in this scenario has a result.
The test case
Each test case is judged on its own: Pass, Fail, Not gradable, or Untested. The scenario result above is based on this test case.
Unprocessable file submitted through the normal upload pathFailEvidence
We uploaded a truncated PDF and a password-protected PDF through Sources, Add source, Upload files in Denser’s dashboard. We then read the source statuses and opened the truncated PDF’s preview and passages tabs.
The tool's response has no text to quote: it is what the evidence under Proofs shows.
| ✗ | Expected: Report a failure for the document that cannot be processed.Found: Denser’s viewer could not display the truncated PDF and its reading tab contained no passages, yet its source status was ready. |
=== TC89 -- bad-file ingestion status (UI only, no API query involved) === Fixture: bad_01_truncated.pdf (deliberately truncated/corrupt PDF, upload-only fixture set) Chatbot: R6-stable-SS-v1 (same chatbot as TC86-88, 9th source added on top of the 8-doc stable slice) Uploaded via: Sources > Add > Files > Choose files (dashboard UI, no ingestion API exists on this tool per reference/api-denser.md) --- Step 1: upload --- File uploaded, size reported 87.77 KB. Immediate post-upload UI optimistically rendered row as "Trained" before the "Uploading 1 file..." notification had even cleared. --- Step 2: confirm post-refresh (not just an optimistic client-side render) --- Clicked "Refresh sources" explicitly, re-read the Sources table. Result: bad_01_truncated.pdf | Status: Trained | Hits: 0 | Size: 87.77 KB Identical status styling (green dot, "Trained" label) to all 8 genuine, successfully-ingested documents in the same table -- no distinct warning/error/partial/failed state of any kind. --- Step 3: attempt to preview the file inside Denser's own PDF viewer --- Clicked the source row to open Preview. Result: "Unable to load PDF -- The file may have been moved or deleted." This is Denser's own UI being unable to open the exact file it has simultaneously marked "Trained" in the Sources table (visible in the same screenshot, side by side -- see tc89-02-bad-file-preview.png). --- Conclusion --- Denser's ingestion pipeline reports a truncated, unparseable PDF as a fully successful "Trained" source -- the same status given to every genuinely ingested document -- with no failure signal anywhere in the dashboard. The product's own preview feature cannot open the file, directly contradicting the "Trained" status shown one UI panel over. This is the TC89 failure mode exactly as specified: "a per-document FAILURE status for that file... Reporting success -- or reporting nothing -- is the failure." --- Second ingestion, 17 November 2026 (truncated file and a password-protected file) --- The original 9-source chatbot no longer exists (Free plan allows one agent), so both files were uploaded to the remaining agent, R6-mutation-MS-v0 (4 other files, 6 sources in total), via Sources > Add source > Upload files. The dashboard now labels sources "Ready" / "Failed" instead of "Trained". bad_02_password_protected.pdf | 171 kB | Status: Failed | "document closed or encrypted. To try again, delete it and add it again." -> failure reported bad_01_truncated.pdf | 88 kB | Status: Ready | same green badge as the good files -> failure NOT reported - Preview (File tab): "Can't show this file" - "What the agent read" tab: "No passages yet" (nothing was read from the file) So Denser does have a Failed state and uses it for the encrypted file, but still marks the truncated file Ready although nothing could be read from it. Verdict unchanged: Fail. Screenshots (17 Nov 2026 ingestion; the 5 Oct 2026 first-run screenshots are no longer kept): - tc89-03-sources-status-annotated.png -- Sources list, password-protected file Failed, truncated file Ready - tc89-04-truncated-preview-fails-annotated.png -- Denser's viewer on the truncated file: "Can't show this file" - tc89-05-truncated-nothing-read-annotated.png -- "What the agent read" tab: "No passages yet"
The genuinely unreadable file went through Denser’s normal upload path. Marking it ready reported success where a failure notice was required; Denser used a failure status for the password-protected PDF in the same upload.
Configuration and setup
How this tool was set up for the run and what the test needed in place. Each row is a fact from the run's records; a fact the records do not hold is left out, not guessed.
How this scenario is graded
How we decide Pass, Fail and Not gradable. The same rules apply to every tool tested on this scenario.
Each test case gets one result per tool: Pass, Fail or Not gradable. A test case we haven't run yet shows Untested. There are no partial results.
- Pass: the tool did everything the test expected, and nothing it said contradicts the correct answer.
- Fail: at least one expected behaviour clearly didn't happen; the test case's result says which one and quotes the tool.
- Not gradable: our evidence couldn't settle the outcome (for example a record we needed is missing). It is never counted as a fail, and the test case's result says what's missing.
Where this sits in the benchmark
This page is one cell of a larger study: one tool, one scenario. Only this benchmark's frame appears here.
| Level | Name | Scope |
|---|---|---|
| Benchmark | Managed RAG → | 15 scenarios · 9 tools |
| Capability | Data Ingestion → | |
| Scenario | A bad file in the upload → | |
| Tool | Denser AI → |
Global scenario definition → · Global capability definition → · Denser AI product page →
History of this result
What has happened to this result since it was first published. Runs and grades are never overwritten: a retest or a re-grade publishes a new result and keeps the earlier one readable.
Act on this result
Nothing filed here edits the run or the grade. A challenge opens a review, and a review can produce a new run or a re-grade — which becomes the current result and leaves this one in the history.
You run the same kind of test against your own setup and get the same behaviour.
Agree →Yours behaves differently. Tell us what you got, with a screenshot if you have one.
Disagree →Something here is wrong — a reference value, a transcription, a grade.
Report an issue →Tell us what changed and we schedule a rerun of the failing test case. The old result stays as history.
Vendor notice →The same record is available as structured data through the AI Demos MCP server, with the counts, the coverage and every per-test-case reason carried as fields.
Verify the proof files
These files support this result. Open a file to inspect the original evidence.
File fingerprints (SHA-256)
A fingerprint identifies the exact file used for this result.