Dataset Viewer

Split (1)

test

The dataset viewer is not available for this split.

Cannot extract the features (columns) for the split 'test' of the config 'default' of the dataset.

Error code:   FeaturesError
Exception:    ArrowInvalid
Message:      Schema at index 1 was different: 
Overall: double
Stage_Concrete_Operational: double
Stage_Sensorimotor: double
Concept_conservation: double
Concept_perceptualconstancy: double
vs
IoU: double
ACC@0.5: double
ACC@0.75: double
ACC@0.9: double
Limited_ACC@0.5: double
Limited_ACC@0.75: double
Limited_ACC@0.9: double
Limited_Small_ACC@0.5: double
Traceback:    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 4408, 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 2679, in _head
                  return next(iter(self.iter(batch_size=n)))
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2861, in iter
                  for key, pa_table in ex_iterable.iter_arrow():
                                       ~~~~~~~~~~~~~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2395, in _iter_arrow
                  yield from self.ex_iterable._iter_arrow()
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 580, in _iter_arrow
                  yield new_key, pa.Table.from_batches(chunks_buffer)
                                 ~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^
                File "pyarrow/table.pxi", line 5040, in pyarrow.lib.Table.from_batches
                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: Schema at index 1 was different: 
              Overall: double
              Stage_Concrete_Operational: double
              Stage_Sensorimotor: double
              Concept_conservation: double
              Concept_perceptualconstancy: double
              vs
              IoU: double
              ACC@0.5: double
              ACC@0.75: double
              ACC@0.9: double
              Limited_ACC@0.5: double
              Limited_ACC@0.75: double
              Limited_ACC@0.9: double
              Limited_Small_ACC@0.5: double

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.

YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

tsv_highres — exact-resolution variants of the ≥8K single-image benchmark pool

Controlled-resolution eval set for the experiment: Thyme (baseline SFT ckpt, crop-tool eval pipeline) vs base Qwen2.5-VL-7B-Instruct (direct VQA), across input resolutions.

Construction

VLMEvalKit-wide ≥4K harvest (hires4k pipeline; see Icey444/tsv_sample/highres for the 300-sample-layer census): every benchmark TSV probed, scanned per-row, filtered at max(W,H) ≥ 3840.
≥8K cut: rows with max(W,H) ≥ 7680 (per-row resolution indexes, largest image per row).
Single-image only: multi-image rows dropped (only XLRS part9 was affected).
One row per distinct image: image-payload-md5 dedup (pointer rows and repeat questions dropped, first occurrence kept) → 2,901 rows / 31 files.
Resolution variants: each image decoded once, cascade-Lanczos resized so the LONG SIDE is exactly 7680 / 3840 / 1024 / 256, re-encoded JPEG quality=95 (uniform re-encode across all variants, so codec/format is not a confound). Aspect ratio preserved. Schema untouched otherwise.

Layout

8k/ 4k/ 1k/ 256/     one folder per resolution, identical filenames inside:
  <Benchmark>.tsv           VLMEvalKit schema, base64 images inline
  XLRS-Bench-lite_part{0..14}.jsonl
eval_results/         predictions + scores from the experiment arms

Each folder is a drop-in $LMUData root for VLMEvalKit with VLMEVAL_USE_LOCAL_TSV_IF_EXISTS=1 (dataset names match exactly; the XLRS loader picks up the part jsonls natively).

Row counts (per resolution, identical across variants)

XLRS-Bench-lite 1784 (15 parts) · MMT-Bench_ALL 626 · HRBench8K 151 · CoreCognition 74 · MMT-Bench_VAL 69 · GroundingME 59 · MMT-Bench_VAL_MI 42 · InfoVQA_VAL 27 · VMCBench_TEST 22 · InfoVQA_TEST 20 · AMBER 7 · MME-RealWorld-Lite 7 · VMCBench_DEV 4 · OCRBench 3 · SEEDBench2 2 · WildDoc 2 · WildVision 2 — total 2,901.

Benches used in the experiment (answers public, locally scoreable): InfoVQA_VAL, CoreCognition, HRBench8K, MMT-Bench_VAL, XLRS-Bench-lite. Removed from the release: MMT-Bench_ALL and VMCBench_TEST (answers held out — not locally scoreable) and GroundingME (GT boxes live in original-image pixel coordinates, incompatible with resized serving).

Downloads last month: 1,057

Total file size:

81.9 GB