Hugging Face's logo

Datasets:
Nabidnur
/
arc-agi-2-solver-traces

License:
Dataset card Files
xet
 Community
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
task_id: string
split: string
solved: bool
committed_solver: string
features: struct<n_demos: int64, n_tests: int64, same_dims_all: bool, in_dims: list<item: int64>, out_dims: li (... 113 chars omitted)
  child 0, n_demos: int64
  child 1, n_tests: int64
  child 2, same_dims_all: bool
  child 3, in_dims: list<item: int64>
      child 0, item: int64
  child 4, out_dims: list<item: int64>
      child 0, item: int64
  child 5, in_palette: list<item: int64>
      child 0, item: int64
  child 6, out_palette: list<item: int64>
      child 0, item: int64
  child 7, palette_delta: list<item: int64>
      child 0, item: int64
actions: list<item: struct<op: string, split: string, detail: string, solver: string, result: string, ms: int (... 47 chars omitted)
  child 0, item: struct<op: string, split: string, detail: string, solver: string, result: string, ms: int64, answer: (... 35 chars omitted)
      child 0, op: string
      child 1, split: string
      child 2, detail: string
      child 3, solver: string
      child 4, result: string
      child 5, ms: int64
      child 6, answer: string
      child 7, solved: bool
      child 8, valid: bool
to
{'task_id': Value('string'), 'split': Value('string'), 'solved': Value('bool'), 'committed_solver': Value('string'), 'actions': Value('string')}
because column names don't match
Traceback:    Traceback (most recent call last):
                File "/src/services/worker/src/worker/utils.py", line 149, 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 129, 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 544, 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 3052, in __iter__
                  for key, example in ex_iterable:
                                      ^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2559, in __iter__
                  for key, pa_table in self._iter_arrow():
                                       ~~~~~~~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2584, 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 558, in _iter_arrow
                  for key, pa_table in iterator:
                                       ^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 440, 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 361, 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 2381, in table_cast
                  return cast_table_to_schema(table, schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2309, in cast_table_to_schema
                  raise CastError(
                  ...<3 lines>...
                  )
              datasets.table.CastError: Couldn't cast
              task_id: string
              split: string
              solved: bool
              committed_solver: string
              features: struct<n_demos: int64, n_tests: int64, same_dims_all: bool, in_dims: list<item: int64>, out_dims: li (... 113 chars omitted)
                child 0, n_demos: int64
                child 1, n_tests: int64
                child 2, same_dims_all: bool
                child 3, in_dims: list<item: int64>
                    child 0, item: int64
                child 4, out_dims: list<item: int64>
                    child 0, item: int64
                child 5, in_palette: list<item: int64>
                    child 0, item: int64
                child 6, out_palette: list<item: int64>
                    child 0, item: int64
                child 7, palette_delta: list<item: int64>
                    child 0, item: int64
              actions: list<item: struct<op: string, split: string, detail: string, solver: string, result: string, ms: int (... 47 chars omitted)
                child 0, item: struct<op: string, split: string, detail: string, solver: string, result: string, ms: int64, answer: (... 35 chars omitted)
                    child 0, op: string
                    child 1, split: string
                    child 2, detail: string
                    child 3, solver: string
                    child 4, result: string
                    child 5, ms: int64
                    child 6, answer: string
                    child 7, solved: bool
                    child 8, valid: bool
              to
              {'task_id': Value('string'), 'split': Value('string'), 'solved': Value('bool'), 'committed_solver': Value('string'), 'actions': 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.

arc-agi-2-solver-traces

Chain-of-dynamic-action traces and mixture-of-solvers coverage supervision for ARC-AGI-2, generated by datasets_build/build_action_traces.py in arc-agi-2-kaggle.

Files

  • actions/dynamic_action_traces.jsonl — one record per task: an ordered action trace observe -> fit -> apply -> check -> commit over a 10-solver council (7 rule solvers in arc_rules.SOLVERS + symmetry_fill, hole_recolor, path_propagation primitives). Each fit is demo-verified (rule must reproduce ALL train pairs); check scores exact|mismatch|unscored where ground truth exists.
  • router/mixture_coverage.jsonl — long-format rows {task_id, solver, fitted, demos_verified, applied, valid_grid, test_exact}: supervision for a mixture-of-solvers router (which solver to trust per task family).
  • stats.json — build stats.

Source tasks

split source n solved-by-council
arc2_train ARC-AGI-2 official training (arcprize/ARC-AGI-2, consolidated; canonical) 1000 20
arc2_eval ARC-AGI-2 official evaluation 120 3
synthetic hf_synthv1_build/curriculum_v1_verified.jsonl (program-verified) 3106 3106

The low solve rate on real tasks is the point: deterministic solvers cover only a small fraction of ARC — the coverage rows quantify exactly where the council is blind, which is where the Qwen3.8 LLM legs in the NVARC pipeline must do the work.

Corpus note: data/arc_dataset (the Kaggle-format files) is verified identical to the official ARC-AGI-2 corpus — all 1120 task IDs shared, 1118 byte-identical train pairs (d8e07eb2, faa9f03d differ slightly; the official versions are canonical). The builder emits each real task exactly once under arc2_* splits — no double counting.

Intended use

  • SFT/routing supervision for a mixture-of-puzzle-solvers head on Qwen3.8-27B-FP8.

  • Dynamic-action transcripts for training a chain-of-tool agent (observe→fit→verify→apply→commit).

  • agents/mixture_of_solving_agents.jsonl — two-round mixture-of-solving-agents transcript: each solver runs as a named persona (Geometer, Symmetrist, Painter, ...); round 1 = claim solve|candidate|no_fit|apply_error with evidence, round 2 = rebuttal citing global reliability P(exact|verified), judge commits the first exact-on-held-out solver.

  • router/solver_reliability.json — per-solver fitted/exact counts and conditional exact rates, global and per split (routing priors).

  • router/task_features.jsonl — per-task routing features (shape deltas, palettes, demo/test arity, symmetry proxies) + which solver committed.

  • actions/hard_cases.jsonl — tough traces: false_confidence (demo-verified but wrong on held-out), contested (multiple exact solvers), unsolved_with_candidate. 552 contested + 32 false-confidence cases.
Downloads last month
91
Total file size:
27.8 MB