Hugging Face's logo

Datasets:
m6rtine
/
comm-gym

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:    TypeError
Message:      Couldn't cast array of type
struct<coupon_id: string, code: string, amount_cents: int64, kind: string, commodities: list<item: string>, threshold_cents: int64>
to
{'coupon_id': Value('string'), 'code': Value('string'), 'amount_cents': Value('int64'), 'kind': Value('string')}
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 2315, in cast_table_to_schema
                  cast_array_to_feature(
                  ~~~~~~~~~~~~~~~~~~~~~^
                      table[name] if name in table_column_names else pa.array([None] * len(table), type=schema.field(name).type),
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                      feature,
                      ^^^^^^^^
                  )
                  ^
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 1861, in wrapper
                  return pa.chunked_array([func(chunk, *args, **kwargs) for chunk in array.chunks])
                                           ~~~~^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2071, in cast_array_to_feature
                  _c(children[name] if name in array_fields else null_array, subfeature)
                  ~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 1863, in wrapper
                  return func(array, *args, **kwargs)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2121, in cast_array_to_feature
                  casted_array_values = _c(array.values, feature.feature)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 1863, in wrapper
                  return func(array, *args, **kwargs)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2161, in cast_array_to_feature
                  raise TypeError(f"Couldn't cast array of type\n{_short_str(array.type)}\nto\n{_short_str(feature)}")
              TypeError: Couldn't cast array of type
              struct<coupon_id: string, code: string, amount_cents: int64, kind: string, commodities: list<item: string>, threshold_cents: int64>
              to
              {'coupon_id': Value('string'), 'code': Value('string'), 'amount_cents': Value('int64'), 'kind': Value('string')}

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.

comm-gym: offer & cartopt — instances and final rollouts

Evaluation artifacts from the comm-gym project: synthetic grocery-communication tasks derived from the Dunnhumby Complete Journey dataset. JSONL throughout; each line is one self-contained JSON object.

Contents

file rows size (MB)
offer/instances/dev.jsonl 10 0.07
offer/instances/eval_unseen.jsonl 50 0.44
offer/instances/eval_unseen.refs.jsonl 150 0.05
offer/rollouts/openai__gpt-6-luna/eval_unseen_L1.jsonl 400 1.33
offer/rollouts/openai__gpt-6-luna/eval_unseen_L1.judged.jsonl 353 0.16
offer/rollouts/openai__gpt-6-luna/eval_unseen_L2.jsonl 400 1.43
offer/rollouts/openai__gpt-6-luna/eval_unseen_L2.judged.jsonl 342 0.15
offer/rollouts/openai__gpt-6-luna/eval_unseen_L3.jsonl 400 1.61
offer/rollouts/openai__gpt-6-luna/eval_unseen_L3.judged.jsonl 307 0.14
offer/rollouts/qwen__qwen3.8-27b/eval_unseen_L1.jsonl 400 1.46
offer/rollouts/qwen__qwen3.8-27b/eval_unseen_L1.judged.jsonl 289 0.13
offer/rollouts/qwen__qwen3.8-27b/eval_unseen_L2.jsonl 400 1.55
offer/rollouts/qwen__qwen3.8-27b/eval_unseen_L2.judged.jsonl 293 0.13
offer/rollouts/qwen__qwen3.8-27b/eval_unseen_L3.jsonl 400 1.76
offer/rollouts/qwen__qwen3.8-27b/eval_unseen_L3.judged.jsonl 254 0.11
cartopt/instances/dev_L1.jsonl 30 0.10
cartopt/instances/dev_L2.jsonl 30 0.16
cartopt/instances/dev_L3.jsonl 30 0.19
cartopt/instances/dev_L4.jsonl 30 0.29
cartopt/instances/eval_unseen_L1.jsonl 100 0.34
cartopt/instances/eval_unseen_L2.jsonl 100 0.53
cartopt/instances/eval_unseen_L3.jsonl 100 0.64
cartopt/instances/eval_unseen_L4.jsonl 100 0.97
cartopt/rollouts/openai__gpt-6-luna/eval_unseen_L1.jsonl 800 2.83
cartopt/rollouts/openai__gpt-6-luna/eval_unseen_L2.jsonl 800 3.45
cartopt/rollouts/openai__gpt-6-luna/eval_unseen_L3.jsonl 800 3.81
cartopt/rollouts/openai__gpt-6-luna/eval_unseen_L4.jsonl 800 4.85
cartopt/rollouts/qwen__qwen3.8-27b/eval_unseen_L1.jsonl 80 0.36
cartopt/rollouts/qwen__qwen3.8-27b/eval_unseen_L2.jsonl 80 0.46

Tasks

offer

Write a grocery push notification grounded in a household's live coupons. eval_unseen = 50 late-window anchor instances; levels L1-L3 vary the number of relevant coupons. Scoring: policy gate (groundedness), then pairwise LLM judge on gate passers. eval_unseen.refs.jsonl holds judge-side reference messages (written by the judge model family). dev = 10 development instances.

cartopt

Cart optimization: given a menu of products, coupons, and constraints, assemble the cheapest feasible cart. Levels L1-L4; graded programmatically against solver ground truth (info.c_star_cents), no judge. Each instance file carries solver metadata in info and the full environment in instance. dev splits are 30 instances, eval_unseen are 100 per level.

Rollout records

Fields: instance_id, level, stratum, model, sampling_params, rollout_idx, prompt (message list), completion, parsed, parse_reason, components, reward, tokens, cost_usd, latency_s, seed, git_sha, timestamp. Offer rollouts have *.judged.jsonl sidecars (judge_win, order_scores, judge_model, completion_sha256, tokens, cost_usd).

Provenance

  • Sampling: temperature 1.0 everywhere; offer max_tokens 4096, cartopt 8192; reasoning effort low, except luna's offer run (reasoning off, historical).
  • k = 8 rollouts per instance; split eval_unseen.
  • Runs 2026-10-23 (cartopt luna) through 2026-10-25 (both qwen sweeps); each record embeds its own git_sha and timestamp.
  • Judge (offer only): anthropic/claude-haiku-5.5 via OpenRouter.

Caveats

  • cartopt qwen3.8-27b covers L1/L2 only (10 instances x k=8 per level): the API-key budget was exhausted mid-sweep and L3/L4 completions truncated at max_tokens 8192 (parse:no_answer_block), so those rollouts were discarded rather than published. The L3/L4 absence is a budget artifact, not a capability result.
  • Offer judged sidecars cover only messages that passed the policy gate; scores computed over passers are a selection effect.
  • Offer judge and reference-message author are the same model family (Haiku-5.5); a self-preference caveat applies.
  • Dev splits are included for reference; headline numbers use eval_unseen.

License: research/evaluation use only. Generated 2026-10-25 by scripts/publish_hf.py.
Downloads last month
53
Total file size:
29.5 MB