Hugging Face's logo

Datasets:
regolo
/
brick-skill-tables

Dataset card Files
xet
 Community
1
Dataset Viewer
Duplicate
The dataset viewer is not available for this split.
Cannot load the dataset split (in streaming mode) to extract the first rows.
Error code:   StreamingRowsError
Exception:    CastError
Message:      Couldn't cast
model: string
provider: string
capabilities: list<item: string>
  child 0, item: string
skill_vector: list<item: double>
  child 0, item: double
source: string
confidence: list<item: string>
  child 0, item: string
imputed_capabilities: list<item: null>
  child 0, item: null
support: null
subset_hash: null
date: timestamp[s]
notes: string
models: list<item: struct<model: string, provider: string, file: string, source: string>>
  child 0, item: struct<model: string, provider: string, file: string, source: string>
      child 0, model: string
      child 1, provider: string
      child 2, file: string
      child 3, source: string
to
{'capabilities': List(Value('string')), 'date': Value('timestamp[s]'), 'models': List({'model': Value('string'), 'provider': Value('string'), 'file': Value('string'), 'source': Value('string')})}
because column names don't match
Traceback:    Traceback (most recent call last):
                File "/src/services/worker/src/worker/utils.py", line 147, in get_rows_or_raise
                  return get_rows(
                         ^^^^^^^^^
                File "/src/libs/libcommon/src/libcommon/utils.py", line 272, in decorator
                  return func(*args, **kwargs)
                         ^^^^^^^^^^^^^^^^^^^^^
                File "/src/services/worker/src/worker/utils.py", line 127, in get_rows
                  rows_plus_one = list(itertools.islice(safe_iter(ds, dataset=dataset), rows_max_number + 1))
                                  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/src/services/worker/src/worker/utils.py", line 478, in safe_iter
                  yield from ds.decode(False) if ds.features else ds
                File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 2815, in __iter__
                  for key, example in ex_iterable:
                                      ^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 2352, in __iter__
                  for key, pa_table in self._iter_arrow():
                                       ^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 2377, in _iter_arrow
                  for key, pa_table in self.ex_iterable._iter_arrow():
                                       ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 536, in _iter_arrow
                  for key, pa_table in iterator:
                                       ^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 419, in _iter_arrow
                  for key, pa_table in self.generate_tables_fn(**gen_kwags):
                                       ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/packaged_modules/json/json.py", line 310, in _generate_tables
                  self._cast_table(pa_table, json_field_paths=json_field_paths),
                  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/packaged_modules/json/json.py", line 130, in _cast_table
                  pa_table = table_cast(pa_table, self.info.features.arrow_schema)
                             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/table.py", line 2369, in table_cast
                  return cast_table_to_schema(table, schema)
                         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/table.py", line 2297, in cast_table_to_schema
                  raise CastError(
              datasets.table.CastError: Couldn't cast
              model: string
              provider: string
              capabilities: list<item: string>
                child 0, item: string
              skill_vector: list<item: double>
                child 0, item: double
              source: string
              confidence: list<item: string>
                child 0, item: string
              imputed_capabilities: list<item: null>
                child 0, item: null
              support: null
              subset_hash: null
              date: timestamp[s]
              notes: string
              models: list<item: struct<model: string, provider: string, file: string, source: string>>
                child 0, item: struct<model: string, provider: string, file: string, source: string>
                    child 0, model: string
                    child 1, provider: string
                    child 2, file: string
                    child 3, source: string
              to
              {'capabilities': List(Value('string')), 'date': Value('timestamp[s]'), 'models': List({'model': Value('string'), 'provider': Value('string'), 'file': Value('string'), 'source': Value('string')})}
              because column names don't match

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.

Brick public skill tables

Per-model skill vectors consumed by the Brick router. Each <model>.json maps a model id to a 6-dimensional capability vector in [0,1]. brick init reads this folder to seed skill_router.models[].skill_vector for known model ids, so a user never has to re-run inference for a model someone already measured.

This folder ships with the CLI (templates/ is published) and mirrors the Hugging Face dataset regolo/brick-skill-tables, which grows as users contribute measurements for new models (brick skills extract --publish, opt-in).

Capability order (canonical, never reorder) 

[coding, creative_synthesis, instruction_following, math_reasoning, planning_agentic, world_knowledge]

Record schema 

{
  "model": "claude-haiku-4-5",
  "provider": "anthropic",
  "capabilities": ["coding", "...", "world_knowledge"],
  "skill_vector": [0.73, 0.65, 0.70, 0.81, 0.50, 0.77],
  "source": "benchmark",          // benchmark | measured | heuristic
  "confidence": ["medium", "low", "medium", "high", "medium", "medium"],
  "imputed_capabilities": [],     // coords filled from the model mean (NA in public benchmarks)
  "support": null,                // {correct,total} per capability — only for source=measured
  "subset_hash": null,            // frozen probe-set hash — only for source=measured
  "date": "2026-06-11",
  "notes": "..."
}

source and trust order

  1. measured — produced by brick skills extract running the frozen probe set against the model and grading verifiable categories (math exact-match, code tests, MMLU-Pro letter). Highest trust; carries support and subset_hash.
  2. benchmark — derived from published AI-lab benchmarks (SWE-bench, AIME, GPQA, MMLU-Pro, tau-bench, …) normalized to [0,1]. Cold-start prior for frontier closed models. A measured record for the same model overrides it.
  3. heuristic — interpolated fallback for an unknown id; marked explicitly.

measured and benchmark vectors are not on the same scale (different question mixes, different grading). Treat benchmark as a prior; prefer measured when both exist for a model.

NA imputation

Public benchmarks systematically omit some capabilities (creative writing is rarely benchmarked; Google does not publish instruction-following or agentic numbers for most Gemini tiers). For a missing coordinate we impute the mean of the model's known coordinates, list the coordinate in imputed_capabilities, and set its confidence to low. This keeps the vector usable by the router while flagging the estimate as soft.

Contributing

Run brick skills extract <model> --publish (opt-in consent prompt) to measure a new model on the frozen probe set and upsert a measured record to regolo/brick-skill-tables. You may also open a PR adding a <model>.json here.
Downloads last month
44
Total file size:
16.8 kB