Error code:   JWTInvalidSignature
Exception:    InvalidSignatureError
Message:      Signature verification failed
Traceback:    Traceback (most recent call last):
                File "/src/libs/libapi/src/libapi/jwt_token.py", line 286, in validate_jwt
                  decoded = jwt.decode(
                      jwt=token,
                  ...<2 lines>...
                      options=options,
                  )
                File "/usr/local/lib/python3.14/site-packages/jwt/api_jwt.py", line 368, in decode
                  decoded = self.decode_complete(
                      jwt,
                  ...<8 lines>...
                      leeway=leeway,
                  )
                File "/usr/local/lib/python3.14/site-packages/jwt/api_jwt.py", line 265, in decode_complete
                  decoded = self._jws.decode_complete(
                      jwt,
                  ...<3 lines>...
                      detached_payload=detached_payload,
                  )
                File "/usr/local/lib/python3.14/site-packages/jwt/api_jws.py", line 270, in decode_complete
                  self._verify_signature(
                  ~~~~~~~~~~~~~~~~~~~~~~^
                      signing_input,
                      ^^^^^^^^^^^^^^
                  ...<4 lines>...
                      options=merged_options,
                      ^^^^^^^^^^^^^^^^^^^^^^^
                  )
                  ^
                File "/usr/local/lib/python3.14/site-packages/jwt/api_jws.py", line 417, in _verify_signature
                  raise InvalidSignatureError("Signature verification failed")
              jwt.exceptions.InvalidSignatureError: Signature verification failed

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

Berkeley GRAFIQ — Graphics Inquiry Benchmark

Synthetic VQA benchmark for evaluating vision-language models on 3D rendering properties (depth, surface roughness, normal orientation, light direction).

The on-disk dataset is the contents of examples/showcase/ from the grafiq_datagen repository, packed into a single tarball to fit free-tier HF upload limits.

Layout

File	What it is
`showcase.tar`	Full showcase folder, including 1.5k render dirs (~3.5 GB).
`benchmark.jsonl`	Original 276-item benchmark.
`benchmark_filtered.jsonl`	Original benchmark, post-filter (227 items).
`benchmark_v2.jsonl`	Balanced 480-item benchmark, 120 per task.
`benchmark_v2_report.json`	L/R balance and drop-reason counts for v2.
`filter_report.json`	Filter rule stats for the original benchmark.

Each render dir inside showcase.tar contains:

rgb.png — the actual image fed to VLMs.
ground_truth.json — task metadata (object paths, planted parameter values, side mapping).
qa_item.json — the verified QA record, present only when the pixel verifier accepted the item.
Task-specific render passes (*0001.exr).

Quick start

from huggingface_hub import hf_hub_download
import tarfile, json

# 1. Pull the lightweight benchmark file.
bench = hf_hub_download(repo_id="tsunghanwu/berkeley-grafiq-dataset",
                         filename="benchmark_v2.jsonl",
                         repo_type="dataset")
items = [json.loads(l) for l in open(bench)]

# 2. Pull and extract the renders if you need the images.
tar_path = hf_hub_download(repo_id="tsunghanwu/berkeley-grafiq-dataset",
                            filename="showcase.tar",
                            repo_type="dataset")
with tarfile.open(tar_path) as tf:
    tf.extractall("examples/showcase")

Tasks

depth_comparison — which object is closer to the camera?
roughness_comparison — which object is glossier (lower roughness)?
normal_orientation — which object's surface faces the camera more directly?
light_direction — which object is more lit by the directional key light?

Each task is a binary "Left vs Right" forced choice. See the evidence block in each qa_item.json for the raw verifier metrics.

Downloads last month: 97

Total file size:

3.75 GB