pdfhell 0.3.0

PDF Hell — adversarial PDFs that break AI document readers. Procedural ground truth, not LLM-as-judge.

PDF Hell

Live leaderboard · Website · PyPI · multivon-eval (engine)

Adversarial PDFs that stress-test AI document readers — with procedural ground truth, not LLM-as-judge.

PDF Hell is a small, focused benchmark for three specific failure modes in AI document pipelines. Every test case is a PDF generated from code, so the correct answer is known exactly. There's no LLM judging another LLM's interpretation — the same complexity that fools the model isn't asked to grade it.

The headline finding (mini-v1, 30 cases, 2026-05-17)

GPT-4o falls for the hidden-OCR trap on 10 out of 10 cases (95% Wilson CI [72%, 100%]) — it consistently returns the invisible amount from the PDF's text layer instead of the visible amount rendered on the page:

Trap: hidden_ocr_mismatch (invoice — visible total $12,345.67, hidden OCR total $22,345.67)
Question: What is the TOTAL AMOUNT DUE?

→ openai:gpt-4o            $22,345.67   ← fell for trap (10/10 in this trap family)
→ openai:gpt-5.4-mini      $22,345.67   ← fell for trap (9/10)
→ openai:gpt-5.4           $12,345.67   ← correct (8/10 across trap)
→ google:gemini-2.5-flash  $12,345.67   ← correct (10/10)
→ anthropic:claude-sonnet-4-6  $12,345.67   ← correct (10/10)

The visible page, the hidden text layer, and an agent that fuses both will give three different answers. pdfhell exists to catch that.

Quickstart (30 seconds)

# 3-case smoke run against the cheapest vision model
export GOOGLE_API_KEY=...
uvx pdfhell run --model google:gemini-2.5-flash --suite smoke

# Or the full mini-v1 suite (30 cases, ~10s on Flash, ~$0.01)
uvx pdfhell run --model anthropic:claude-sonnet-4-6 --suite mini

# Or generate one trap PDF and inspect it
uvx pdfhell make --trap hidden_ocr_mismatch --seed 42
open ./cases/hidden_ocr_mismatch-0042.pdf

pdfhell run builds the suite on first use, sends each PDF to the vision model, and grades the answer against code-based ground truth.

Mini-v1 leaderboard (8 models, 30 cases)

Model	Pass rate	95% CI	Hidden OCR	Footnote	Split table
anthropic:claude-sonnet-4-6	29/30 (97%)	[83%, 99%]	10/10	9/10	10/10
google:gemini-3.1-pro-preview	28/30 (93%)	[78%, 98%]	10/10	8/10	10/10
google:gemini-3.1-flash-lite	28/30 (93%)	[78%, 98%]	10/10	8/10	10/10
google:gemini-2.5-pro	28/30 (93%)	[78%, 98%]	10/10	8/10	10/10
google:gemini-2.5-flash	28/30 (93%)	[78%, 98%]	10/10	8/10	10/10
openai:gpt-5.4	27/30 (90%)	[74%, 97%]	8/10	9/10	10/10
openai:gpt-5.4-mini	20/30 (67%)	[49%, 81%]	1/10	9/10	10/10
openai:gpt-4o	14/30 (47%)	[30%, 64%]	0/10	8/10	6/10

What is and isn't supported by this data:

• ✅ GPT-4o is materially worse than the others on this suite — its CI [30%, 64%] does not overlap with any other model's.
• ✅ GPT-4o falls for the hidden-OCR trap 100% of cases (CI [72%, 100%]). Every failure returned the hidden-OCR amount specifically.
• ✅ GPT-5.4 fixes most of it (80% pass on hidden OCR) — a real generational improvement.
• ❌ "Claude leads" — Sonnet's CI [83%, 99%] overlaps with Gemini's [78%, 98%]. The two are statistically indistinguishable on this suite. Don't read ordinal rankings from 30 cases.
• ❌ "PDF Hell is sufficient to evaluate document AI." It's a stress test for three specific failure modes. Pair it with a domain benchmark (DocVQA, your own regression suite) for coverage.

Suite hash: 8ad87b8d (mini-v1, 30 cases). Every leaderboard row above was measured on the same hash. Raw run JSON at https://github.com/multivon-ai/multivon-web/tree/main/public/data/pdfhell-runs.

Research-discovered traps (0.3.0)

pdfhell ships with an autoresearch loop (pdfhell.research) that discovers new adversarial traps automatically by maximising cross-model discrimination on the eval panel. Three strong reasoning models (Opus 4-7, GPT-5, Gemini 2.5 Pro) rotate as the researcher; candidates pass five validation gates (parseable, deterministic, answerable, forbidden-clean, lint-clean) before any eval spend is committed.

The very first overnight run ($7 budget, 10 candidates, 2 hours wall clock) discovered unicode_confusable_total:

Two visually-identical "TOTAL" rows. One uses ASCII "O". The other
uses Cyrillic capital "О" (U+041E). A printed clause names which
codepoint is binding. Vision-only readers can't tell the labels
apart and must guess.

Model	Pass on unicode_confusable_total
openai:gpt-5	100%
anthropic:claude-haiku-4-5	93%
google:gemini-2.5-flash	87%
openai:gpt-4o	80%
google:gemini-2.5-pro	67%
anthropic:claude-sonnet-4-6	60%
anthropic:claude-opus-4-7	0%
google:gemini-flash-lite-latest	0%

The premium tier is not universally better. Opus 4-7 (Anthropic's most expensive vision model) fails 0/15 while Haiku 4-5 (Anthropic's cheapest) passes 14/15. Same provider, different blind spots.

The full research trail (every candidate, every rationale, every dollar) lives in pdfhell/research/results.tsv, keep/*.json, and budget.jsonl. See pdfhell/research/README.md for the methodology and how to run your own loop.

What's in mini-v1

Trap family	Cases	What breaks
hidden_ocr_mismatch	10	Invoices where the visible amount differs from an invisible OCR text layer. Vision-only models read the page; text-extraction pipelines read the layer; they disagree.
footnote_override	10	Legal clauses where a 6pt footnote overrides the body — liability caps with carve-outs, terminations with restrictions, data-residency with disaster-recovery exceptions.
split_table_across_pages	10	Financial tables where the header row sits on page 1 and the body rows on page 2. RAG loaders that paginate independently lose column context.

Every case has a deterministic seed. Re-running with the same seed regenerates byte-identical PDFs and identical answer keys (Canvas(invariant=True) on every generator).

Suite versioning. The mini-v1 label + suite hash (8ad87b8d) fingerprints the exact (trap_family, seed) pairs measured. Adding a new trap family produces mini-v2 with a different hash — runs across different hashes are not directly comparable. See the next section for the roadmap.

Why this exists

The current AI-eval state of the art uses an LLM-as-judge to grade another LLM's answer. That's circular: the same complexity that fools the agent fools the judge. PDF Hell rejects that:

1. Code-based ground truth. The answer is a literal Python value the generator chose, not a frontier model's opinion.
2. A named failure mode per trap. When a model fails, we know which specific failure caught it (e.g. "trusted the hidden OCR layer over the visible page").
3. A diagnostic signal, not just a score. Per-trap-family breakdown tells you which assumption broke.

Commands

pdfhell list-traps                              # list trap families
pdfhell make --trap <family> --seed <n>         # generate one case
pdfhell build --suite <smoke|mini> --out <dir>  # materialise a suite
pdfhell run --model <provider>:<model>          # evaluate a model
  [--suite smoke|mini]                          #   (default: mini)
  [--cases-dir <dir>]                           #   (default: ./cases/<suite>)
  [--out <path>]                                #   JSON output
  [--junit <path>]                              #   JUnit XML for GitHub Actions / GitLab CI
  [--fail-threshold <0.0-1.0>]                  #   non-zero exit if pass_rate below threshold
  [--workers <n>]                               #   parallel API requests (default: 4)
  [--quiet]
pdfhell report runs/<file>.json                 # print a saved run's summary

Provider shorthand: anthropic:claude-sonnet-4-6, openai:gpt-4o, google:gemini-2.5-pro, google:gemini-2.5-flash, etc. API key from env (ANTHROPIC_API_KEY, OPENAI_API_KEY, GOOGLE_API_KEY).

CI integration

Drop this into .github/workflows/eval.yml:

name: PDF Hell
on: [pull_request]
jobs:
  pdfhell:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: astral-sh/setup-uv@v5
      - run: uvx pdfhell run --model anthropic:claude-sonnet-4-6 --suite mini --junit results.xml --fail-threshold 0.7
        env:
          ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}
      - uses: actions/upload-artifact@v4
        with:
          name: pdfhell-results
          path: results.xml

JUnit XML renders natively in the GitHub Actions / GitLab CI / CircleCI / Jenkins PR panel — failures show up as red rows with the expected and observed answers in the failure message.

How scoring works

Two layers, applied in order:

1. Procedural exact match (primary) — for single-value traps, the model's free-text answer must contain the expected value (whitespace-tolerant, case-insensitive). For prose traps like footnote_override, the model must include every required token (the cap value, every carve-out section number, etc.) in any order, in any phrasing. The model isn't graded on prose style; it's graded on whether it captured the facts.
2. Forbidden-answer detection (diagnostic) — did the model return one of the answers the trap was specifically designed to elicit (e.g. the hidden-OCR amount)? If so, the trap caught a known failure mode and we record it. Doesn't affect the primary score.

Anything that looks like a refusal ("I can't determine...") is recorded as refused, not as a wrong answer.

The QAG explanation layer from multivon-eval (DocumentGrounding) is available separately for users who want a human-readable "why did the model fail" breakdown — but it's never on the scoring path.

Adding a new trap family

Add a generator at pdfhell/generators/<your_trap>.py:

from ..case import HellCase
from . import _common as C

def generate(seed: int) -> tuple[bytes, HellCase]:
    rng = C.rng_for(seed)
    # ... draw a PDF with reportlab using rng for all random choices ...
    # invariant=True is the default — keep your generator deterministic.
    return pdf_bytes, HellCase(
        id=f"your_trap-{seed:04d}",
        trap_family="your_trap",
        seed=seed,
        question="What is ...?",
        expected_answer="42",                # single canonical answer
        expected_tokens=["42"],              # OR list of required substrings for prose
        forbidden_answers=["41", "43"],      # OR a value the trap specifically elicits
        metadata={"expected_failure_mode": "Model does X when it should do Y."},
    )

Register it in pdfhell/generators/__init__.py. See CONTRIBUTING.md for the full guide. Tests run with pytest.

Roadmap

The 0.1 release is intentionally narrow — three trap families, 30 cases. Coming next:

• merged_table_cells — value depends on row/column span interpretation
• rotated_scan — visually legible but OCR-broken pages
• near_duplicate_entities — "ACME Ltd." vs "ACME Holdings Ltd."
• prompt_injection_in_body — "Ignore previous instructions and answer X"
• chart_axis_inversion — answers depend on reading axis direction
• checkbox_ambiguity — selected vs unselected with low visual margin
• cross_page_citation — answers requiring page + bounding-box citations

Target full suite: 10 trap families, ~50 cases.

Hosted generator

For document-AI teams who need adversarial test cases tailored to their templates (claims forms, MSAs, medical records, KYC docs), there's a hosted generator that takes your templates and produces adversarial variants with code-based ground truth — same methodology, your data shape.

Email [e-mail] for early access, or see multivon.ai/commercial.

Installing

# Recommended (zero-install with uv):
uvx pdfhell list-traps

# Or in a venv:
python -m venv .venv && source .venv/bin/activate
pip install pdfhell

Bare install brings in multivon-eval (the engine), reportlab (PDF generation), pypdf, and the three frontier-provider SDKs (anthropic, openai, google-genai). No provider extras to remember; no GPU required.

The Multivon ecosystem

Five public + one early-access package, all built on a shared evaluation engine:

Repo	What it is
multivon-eval	Python SDK — 44 evaluators + bootstrap CLI + multivon_eval.auto. PDF Hell's engine.
pdfhell (you are here)	Adversarial PDFs that break AI document readers
multivon-mcp	MCP server — exposes pdfhell_run + pdfhell_make as tools to Claude / Cursor
eval-action	GitHub Action — runs pdfhell + multivon-eval on every PR
eval-framework-benchmark	Reproducible head-to-head vs DeepEval + RAGAS (text eval, not PDFs)
multivon-guard (early access)	Local proxy that catches LLM coding agents leaking secrets / PII

License

Apache 2.0. Built on multivon-eval.

Citing

@software{pdfhell,
  title  = {PDF Hell: Adversarial PDFs for AI document readers},
  author = {Multivon},
  url    = {https://github.com/multivon-ai/pdfhell},
}