Datasets:

pavan01729
/

ambench-routing-v1

Tasks:

Text Classification

Tags:

License:

Dataset card Files Files and versions

xet

Community

Dataset Viewer

Duplicate

Split (1)

train

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
id: string
expected: string
picked: string
correct: bool
difficulty: string
base_model: struct<name: string, path: string>
  child 0, name: string
  child 1, path: string
adapters: struct<Astrophysics: struct<path: string, description: string>, Bash: struct<path: string, descripti (... 5082 chars omitted)
  child 0, Astrophysics: struct<path: string, description: string>
      child 0, path: string
      child 1, description: string
  child 1, Bash: struct<path: string, description: string>
      child 0, path: string
      child 1, description: string
  child 2, Biology: struct<path: string, description: string>
      child 0, path: string
      child 1, description: string
  child 3, CaseLaw: struct<path: string, description: string>
      child 0, path: string
      child 1, description: string
  child 4, Chemistry: struct<path: string, description: string>
      child 0, path: string
      child 1, description: string
  child 5, CTI: struct<path: string, description: string>
      child 0, path: string
      child 1, description: string
  child 6, Cypher: struct<path: string, description: string>
      child 0, path: string
      child 1, description: string
  child 7, Finance: struct<path: string, description: string>
      child 0, path: string
      child 1, description: string
  child 8, FunctionCall: struct<path: string, description: string>
      child 0, path: string
      child 1, description: string
  child 9, History: struct<path: string, description: string>
  
...
tion: string
  child 83, YaraGen: struct<path: string, description: string>
      child 0, path: string
      child 1, description: string
  child 84, CloudformationGen: struct<path: string, description: string>
      child 0, path: string
      child 1, description: string
  child 85, MedREFL: struct<path: string, description: string>
      child 0, path: string
      child 1, description: string
  child 86, PiiV2: struct<path: string, description: string>
      child 0, path: string
      child 1, description: string
  child 87, EconometricsV2: struct<path: string, description: string>
      child 0, path: string
      child 1, description: string
  child 88, MedicalGeneticsV2: struct<path: string, description: string>
      child 0, path: string
      child 1, description: string
  child 89, qwen25-grpo-qed-optim-v2-gated-stuck: struct<path: string, description: string>
      child 0, path: string
      child 1, description: string
  child 90, qwen25-smiles-cpt: struct<path: string, description: string>
      child 0, path: string
      child 1, description: string
  child 91, qwen25-smiles-midtrain: struct<path: string, description: string>
      child 0, path: string
      child 1, description: string
  child 92, qwen25_query_rewriter: struct<path: string, description: string>
      child 0, path: string
      child 1, description: string
  child 93, qwen25_router_v2: struct<path: string, description: string>
      child 0, path: string
      child 1, description: string
to
{'base_model': {'name': Value('string'), 'path': Value('string')}, 'adapters': {'Astrophysics': {'path': Value('string'), 'description': Value('string')}, 'Bash': {'path': Value('string'), 'description': Value('string')}, 'Biology': {'path': Value('string'), 'description': Value('string')}, 'CaseLaw': {'path': Value('string'), 'description': Value('string')}, 'Chemistry': {'path': Value('string'), 'description': Value('string')}, 'CTI': {'path': Value('string'), 'description': Value('string')}, 'Cypher': {'path': Value('string'), 'description': Value('string')}, 'Finance': {'path': Value('string'), 'description': Value('string')}, 'FunctionCall': {'path': Value('string'), 'description': Value('string')}, 'History': {'path': Value('string'), 'description': Value('string')}, 'Legal': {'path': Value('string'), 'description': Value('string')}, 'Marketing': {'path': Value('string'), 'description': Value('string')}, 'Medicine': {'path': Value('string'), 'description': Value('string')}, 'Mermaid': {'path': Value('string'), 'description': Value('string')}, 'MusicTheory': {'path': Value('string'), 'description': Value('string')}, 'Philosophy': {'path': Value('string'), 'description': Value('string')}, 'PII': {'path': Value('string'), 'description': Value('string')}, 'Psychology': {'path': Value('string'), 'description': Value('string')}, 'Quantum': {'path': Value('string'), 'description': Value('string')}, 'Reasoning': {'path': Value('string'), 'description': Value('string')}, 'Regex'
...
on': Value('string')}, 'RegexV2': {'path': Value('string'), 'description': Value('string')}, 'CodeCompletion': {'path': Value('string'), 'description': Value('string')}, 'DockerfileGen': {'path': Value('string'), 'description': Value('string')}, 'DockerComposeGen': {'path': Value('string'), 'description': Value('string')}, 'HelmTemplate': {'path': Value('string'), 'description': Value('string')}, 'CodeCompletionDpo': {'path': Value('string'), 'description': Value('string')}, 'MakefileGen': {'path': Value('string'), 'description': Value('string')}, 'SolidityGen': {'path': Value('string'), 'description': Value('string')}, 'YaraGen': {'path': Value('string'), 'description': Value('string')}, 'CloudformationGen': {'path': Value('string'), 'description': Value('string')}, 'MedREFL': {'path': Value('string'), 'description': Value('string')}, 'PiiV2': {'path': Value('string'), 'description': Value('string')}, 'EconometricsV2': {'path': Value('string'), 'description': Value('string')}, 'MedicalGeneticsV2': {'path': Value('string'), 'description': Value('string')}, 'qwen25-grpo-qed-optim-v2-gated-stuck': {'path': Value('string'), 'description': Value('string')}, 'qwen25-smiles-cpt': {'path': Value('string'), 'description': Value('string')}, 'qwen25-smiles-midtrain': {'path': Value('string'), 'description': Value('string')}, 'qwen25_query_rewriter': {'path': Value('string'), 'description': Value('string')}, 'qwen25_router_v2': {'path': Value('string'), 'description': 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(
                      dataset=dataset,
                  ...<4 lines>...
                      column_names=column_names,
                  )
                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 483, in safe_iter
                  yield from ds.decode(False) if ds.features else ds
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2840, in __iter__
                  for key, example in ex_iterable:
                                      ^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2373, in __iter__
                  for key, pa_table in self._iter_arrow():
                                       ~~~~~~~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2398, in _iter_arrow
                  for key, pa_table in self.ex_iterable._iter_arrow():
                                       ~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 536, in _iter_arrow
                  for key, pa_table in iterator:
                                       ^^^^^^^^
                File "/usr/local/lib/python3.14/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/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 343, in _generate_tables
                  self._cast_table(pa_table, json_field_paths=json_field_paths),
                  ~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 132, in _cast_table
                  pa_table = table_cast(pa_table, self.info.features.arrow_schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2378, in table_cast
                  return cast_table_to_schema(table, schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2306, in cast_table_to_schema
                  raise CastError(
                  ...<3 lines>...
                  )
              datasets.table.CastError: Couldn't cast
              id: string
              expected: string
              picked: string
              correct: bool
              difficulty: string
              base_model: struct<name: string, path: string>
                child 0, name: string
                child 1, path: string
              adapters: struct<Astrophysics: struct<path: string, description: string>, Bash: struct<path: string, descripti (... 5082 chars omitted)
                child 0, Astrophysics: struct<path: string, description: string>
                    child 0, path: string
                    child 1, description: string
                child 1, Bash: struct<path: string, description: string>
                    child 0, path: string
                    child 1, description: string
                child 2, Biology: struct<path: string, description: string>
                    child 0, path: string
                    child 1, description: string
                child 3, CaseLaw: struct<path: string, description: string>
                    child 0, path: string
                    child 1, description: string
                child 4, Chemistry: struct<path: string, description: string>
                    child 0, path: string
                    child 1, description: string
                child 5, CTI: struct<path: string, description: string>
                    child 0, path: string
                    child 1, description: string
                child 6, Cypher: struct<path: string, description: string>
                    child 0, path: string
                    child 1, description: string
                child 7, Finance: struct<path: string, description: string>
                    child 0, path: string
                    child 1, description: string
                child 8, FunctionCall: struct<path: string, description: string>
                    child 0, path: string
                    child 1, description: string
                child 9, History: struct<path: string, description: string>
                
              ...
              tion: string
                child 83, YaraGen: struct<path: string, description: string>
                    child 0, path: string
                    child 1, description: string
                child 84, CloudformationGen: struct<path: string, description: string>
                    child 0, path: string
                    child 1, description: string
                child 85, MedREFL: struct<path: string, description: string>
                    child 0, path: string
                    child 1, description: string
                child 86, PiiV2: struct<path: string, description: string>
                    child 0, path: string
                    child 1, description: string
                child 87, EconometricsV2: struct<path: string, description: string>
                    child 0, path: string
                    child 1, description: string
                child 88, MedicalGeneticsV2: struct<path: string, description: string>
                    child 0, path: string
                    child 1, description: string
                child 89, qwen25-grpo-qed-optim-v2-gated-stuck: struct<path: string, description: string>
                    child 0, path: string
                    child 1, description: string
                child 90, qwen25-smiles-cpt: struct<path: string, description: string>
                    child 0, path: string
                    child 1, description: string
                child 91, qwen25-smiles-midtrain: struct<path: string, description: string>
                    child 0, path: string
                    child 1, description: string
                child 92, qwen25_query_rewriter: struct<path: string, description: string>
                    child 0, path: string
                    child 1, description: string
                child 93, qwen25_router_v2: struct<path: string, description: string>
                    child 0, path: string
                    child 1, description: string
              to
              {'base_model': {'name': Value('string'), 'path': Value('string')}, 'adapters': {'Astrophysics': {'path': Value('string'), 'description': Value('string')}, 'Bash': {'path': Value('string'), 'description': Value('string')}, 'Biology': {'path': Value('string'), 'description': Value('string')}, 'CaseLaw': {'path': Value('string'), 'description': Value('string')}, 'Chemistry': {'path': Value('string'), 'description': Value('string')}, 'CTI': {'path': Value('string'), 'description': Value('string')}, 'Cypher': {'path': Value('string'), 'description': Value('string')}, 'Finance': {'path': Value('string'), 'description': Value('string')}, 'FunctionCall': {'path': Value('string'), 'description': Value('string')}, 'History': {'path': Value('string'), 'description': Value('string')}, 'Legal': {'path': Value('string'), 'description': Value('string')}, 'Marketing': {'path': Value('string'), 'description': Value('string')}, 'Medicine': {'path': Value('string'), 'description': Value('string')}, 'Mermaid': {'path': Value('string'), 'description': Value('string')}, 'MusicTheory': {'path': Value('string'), 'description': Value('string')}, 'Philosophy': {'path': Value('string'), 'description': Value('string')}, 'PII': {'path': Value('string'), 'description': Value('string')}, 'Psychology': {'path': Value('string'), 'description': Value('string')}, 'Quantum': {'path': Value('string'), 'description': Value('string')}, 'Reasoning': {'path': Value('string'), 'description': Value('string')}, 'Regex'
              ...
              on': Value('string')}, 'RegexV2': {'path': Value('string'), 'description': Value('string')}, 'CodeCompletion': {'path': Value('string'), 'description': Value('string')}, 'DockerfileGen': {'path': Value('string'), 'description': Value('string')}, 'DockerComposeGen': {'path': Value('string'), 'description': Value('string')}, 'HelmTemplate': {'path': Value('string'), 'description': Value('string')}, 'CodeCompletionDpo': {'path': Value('string'), 'description': Value('string')}, 'MakefileGen': {'path': Value('string'), 'description': Value('string')}, 'SolidityGen': {'path': Value('string'), 'description': Value('string')}, 'YaraGen': {'path': Value('string'), 'description': Value('string')}, 'CloudformationGen': {'path': Value('string'), 'description': Value('string')}, 'MedREFL': {'path': Value('string'), 'description': Value('string')}, 'PiiV2': {'path': Value('string'), 'description': Value('string')}, 'EconometricsV2': {'path': Value('string'), 'description': Value('string')}, 'MedicalGeneticsV2': {'path': Value('string'), 'description': Value('string')}, 'qwen25-grpo-qed-optim-v2-gated-stuck': {'path': Value('string'), 'description': Value('string')}, 'qwen25-smiles-cpt': {'path': Value('string'), 'description': Value('string')}, 'qwen25-smiles-midtrain': {'path': Value('string'), 'description': Value('string')}, 'qwen25_query_rewriter': {'path': Value('string'), 'description': Value('string')}, 'qwen25_router_v2': {'path': Value('string'), 'description': 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.

AMBench-Routing-v1

Track 1 (Routing Accuracy) of AMBench, the released benchmark suite for the Adaptive Minds project (LoRA adapters exposed as callable tools for a base model). Measures whether a router (keyword substring match, embedding-similarity nearest-neighbor, or an LLM reading a natural-language catalog) can pick the right specialist adapter for a query, at increasing library scale (60 -> 94 -> 100 real, independently trained adapters).

Three versioned sub-packages, each standalone-reproducible from its own files alone (no access to any private repo needed):

Directory	Library size	Queries	What it adds
`v1/`	62 adapters	139	The original frozen release: in-library routing, should-refuse (out-of-coverage queries), and naturalistic-phrasing robustness slices.
`v1-90lib/`	94 adapters	166	Extends `v1/` with real queries for every adapter added since, plus expanded hand-curated equivalence groups.
`v1-99lib/`	100 adapters	172	Extends `v1-90lib/` with the final 6 adapters, and adds a third router baseline: embedding-similarity (`sentence-transformers/all-MiniLM-L6-v2`, no LLM call).

Reproduce every number from any sub-directory alone:

cd v1 && python3 scorer.py --check
cd v1-90lib && python3 build_and_run.py   # rebuilds registry + queries + picks + scores
cd v1-99lib && python3 build_and_run.py --with-cron

A disclosed correction

v1/'s own README.md (Known Gap #9) documents a real scoring bug found after this package was first assembled: a downstream caller's equivalence-set logic silently dropped this package's own hand-curated cross-domain equivalence groups. No query, gold label, or pick in this package was ever wrong — only that external caller's scoring was. The bug is fixed in v1-90lib/'s and v1-99lib/'s own build_and_run.py; running either reproduces the corrected numbers directly.

Headline numbers (LLM router, equivalence-set tier)

Scale	Keyword	Embedding	LLM
60 adapters	76.3%	89.9%	92.8%
94 adapters	63.9%	--	89.2%
100 adapters	61.0%	69.8%	87.8%

Full per-slice breakdowns, known gaps, and methodology are in each sub-directory's own README.md.

Citation

Part of the Adaptive Minds project. See the project repository for the full paper and adapter weights (pavan01729/adaptive-minds-loras).

Downloads last month: 74

Total file size:

422 kB