The dataset viewer is not available for this split.
Error code: FeaturesError
Exception: ArrowInvalid
Message: JSON parse error: Invalid value. in row 0
Traceback: Traceback (most recent call last):
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 324, in _generate_tables
df = pandas_read_json(f)
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 38, in pandas_read_json
return pd.read_json(path_or_buf, **kwargs)
~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/pandas/io/json/_json.py", line 815, in read_json
return json_reader.read()
~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/pandas/io/json/_json.py", line 1014, in read
obj = self._get_object_parser(self.data)
File "/usr/local/lib/python3.14/site-packages/pandas/io/json/_json.py", line 1040, in _get_object_parser
obj = FrameParser(json, **kwargs).parse()
File "/usr/local/lib/python3.14/site-packages/pandas/io/json/_json.py", line 1176, in parse
self._parse()
~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/pandas/io/json/_json.py", line 1392, in _parse
ujson_loads(json, precise_float=self.precise_float), dtype=None
~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ValueError: Expected object or value
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "/src/services/worker/src/worker/job_runners/split/first_rows.py", line 244, in compute_first_rows_from_streaming_response
iterable_dataset = iterable_dataset._resolve_features()
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 4523, in _resolve_features
features = _infer_features_from_batch(self.with_format(None)._head())
~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2768, in _head
return next(iter(self.iter(batch_size=n)))
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2972, in iter
for key, pa_table in ex_iterable.iter_arrow():
~~~~~~~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2483, in _iter_arrow
yield from self.ex_iterable._iter_arrow()
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 547, in _iter_arrow
for key, pa_table in iterator:
^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 430, in _iter_arrow
for key, pa_table in self.generate_tables_fn(**gen_kwags):
~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 327, in _generate_tables
raise e
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 290, in _generate_tables
pa_table = paj.read_json(
io.BytesIO(batch), read_options=paj.ReadOptions(block_size=block_size)
)
File "pyarrow/_json.pyx", line 364, in pyarrow._json.read_json
File "pyarrow/error.pxi", line 155, in pyarrow.lib.pyarrow_internal_check_status
return check_status(status)
File "pyarrow/error.pxi", line 92, in pyarrow.lib.check_status
raise convert_status(status)
pyarrow.lib.ArrowInvalid: JSON parse error: Invalid value. in row 0Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
Streetscape representation sensitivity: code and recorded results
Research materials for Photographic representation sensitivity in multimodal streetscape auditing: a matched-panorama study in Singapore, by Yilu Luo and Jinmin Li.
This release accompanies a manuscript prepared for submission to Environment and Planning B: Urban Analytics and City Science. It is not an accepted or published journal article.
Contents
Download Streetscape_reproducibility.zip, verify its SHA256 against release_manifest.json, and read READ_ME.txt after extracting it. The archive contains code, all recorded endpoint attempts, strict and wrapper-normalized analyses, metadata, tables, and schematic/statistical figures. Its SHA256SUMS.json verifies individual files.
The study uses 739 unique Singapore source panoramas from eleven neighborhoods, a 132-location common panel, a nested 44-location control panel, and 9,450 recorded research request dispositions: 6,150 original attempts and 3,300 balanced follow-up attempts. The follow-up repeats both reference and control inputs five times at all 44 nested locations, using randomized rounds and fixed client concurrency. GPT-5.6-sol was accessed through an API; actual endpoint metadata and the request identifier gpt-5.6-sol are retained. The explicitly post hoc neighborhood extension describes how sample-level attribute reporting rates and ranks vary across image representations.
Reproduction and interpretation
The stored-output analyses can be rerun without GPU inference or access to the source images. Follow READ_ME.txt and analysis_environment.json. The common-sample, neighborhood, urban-comparability, original-execution and balanced-control analyses were independently reproduced from the transferable package. Scientific tables and observed estimands were checked against the stored outputs; the balanced validator also independently reparses responses and recalculates all run-pair contrasts. New model calls require a suitable environment, lawful access to image inputs, and separately supplied credentials; they need not reproduce outputs bit for bit.
The outcomes measure model reporting sensitivity and repeatability. They are not ground-truth accuracy, human perceptions, or validated neighborhood quality. The original controls have only reference-side repeats. Their one-sided contrast can be nonpositive even when reporting probabilities differ. The balanced follow-up measures both reference and control repeatability and compares all 25 cross-condition run pairs. Under independent stable repeated outputs, its expected contrast is one half of the squared distance between marginal state-probability vectors. This assumption does not establish API stationarity or factual correctness. Original and follow-up records remain separate, and historical raw records are preserved. Consult the recorded failures, input-integrity exclusions, and study limitations before reuse.
Terminology clarification: the frozen protocol's shorthand "Present-to-not_visible switching" denotes bidirectional present/not_visible switching in the implemented metric and final analysis. Direction-specific transition counts are also retained. The frozen protocol and raw records have not been silently rewritten.
Balanced follow-up results
Wrapper-normalized contrasts subtract the average of both within-condition baselines from all-cross disagreement. Units are percentage points; intervals are marginal 95% intervals from 500 m spatial blocks. Each location requires five valid outputs on both sides. Strict results and alternative spatial groupings are included in the archive. An interval containing zero does not establish equivalence.
| Configuration | Control | Complete locations | Contrast | 95% interval |
|---|---|---|---|---|
| Qwen122B | Grid packaging | 44 | 6.11 | 4.72 to 7.70 |
| Qwen122B | Image order | 44 | 3.59 | 2.35 to 4.82 |
| Qwen122B | Panorama seam | 44 | 2.65 | 1.67 to 3.75 |
| Qwen397B-FP8 | Grid packaging | 44 | 4.78 | 3.20 to 6.47 |
| Qwen397B-FP8 | Image order | 44 | 2.19 | 1.01 to 3.79 |
| Qwen397B-FP8 | Panorama seam | 44 | 3.11 | 1.82 to 4.37 |
| GPT-5.6-sol | Grid packaging | 44 | 1.09 | 0.31 to 1.86 |
| GPT-5.6-sol | Image order | 43 | 0.31 | -0.27 to 0.83 |
| GPT-5.6-sol | Panorama seam | 43 | 1.86 | 0.93 to 2.96 |
These results are conditional on the sampled locations, prompt, parser and recorded serving conditions; they do not measure factual accuracy.
Paired repetition and round-associated diagnostics
The 8 October revision reanalyzes the existing 3,300 balanced records without new inference. It reports reference-minus-control within-condition rates on fixed complete cohorts, round-to-other-four profiles, all ten within-condition round pairs with elapsed-time gaps, and same-round versus different-round cross disagreement. Strict and normalized results and alternative spatial groupings remain separate. Unequal within-condition rates can occur under stable condition-specific state distributions: they are not, by themselves, evidence of temporal drift. The GPT-5.6-sol seam baseline difference has a positive 500 m interval but its 250 m interval includes zero. Overlapping round diagnostics neither establish stationarity nor prove independence. See analysis_repeat_diagnostics20261008/ and its independent validation record.
Data and image rights
Original and derived Google Street View images and encoded image payloads are excluded. The authors report obtaining imagery through Google Maps API. Metadata and hashes support provenance and matching; they do not grant image access or redistribution rights. Image acquisition and use remain subject to applicable provider terms. See DATA_RIGHTS_NOTICE.txt. The legacy inventory repository's license label is not asserted as a license for third-party imagery. This release provides research materials for inspection and reproducibility; no blanket third-party content license is asserted.
Cite the release
Cite the authors, study title, this repository URL, and the exact repository revision used. No journal DOI is available for this manuscript. release_manifest.json records the release date and archive checksum. Funding: the authors report no financial support for the research, authorship or publication of this article.
Generative AI assistance in writing, coding, experiment orchestration and analysis summaries is disclosed in the associated submission. The recorded responses are outputs of systems under study, not fabricated participant observations.
- Downloads last month
- 32