Datasets:

Danchi17
/

ramr

Error code:   StreamingRowsError
Exception:    CastError
Message:      Couldn't cast
id: string
question: string
gold_facts: list<item: string>
  child 0, item: string
answer: string
drop_index: int64
distractor_pool: list<item: string>
  child 0, item: string
schema: struct<id: string, question: string, gold_facts: string, answer: string, drop_index: string, distrac (... 17 chars omitted)
  child 0, id: string
  child 1, question: string
  child 2, gold_facts: string
  child 3, answer: string
  child 4, drop_index: string
  child 5, distractor_pool: string
description: string
contamination_note: string
memory_families_external: struct<FACT-RETENTION: string, OUTCOME-RANKED-RECALL: string>
  child 0, FACT-RETENTION: string
  child 1, OUTCOME-RANKED-RECALL: string
metrics: list<item: string>
  child 0, item: string
files: struct<ramr_chains_v0.1.0.jsonl: struct<n: int64, sha256: string>>
  child 0, ramr_chains_v0.1.0.jsonl: struct<n: int64, sha256: string>
      child 0, n: int64
      child 1, sha256: string
version: string
name: string
generator_seed: int64
hops: int64
to
{'name': Value('string'), 'version': Value('string'), 'description': Value('string'), 'generator_seed': Value('int64'), 'hops': Value('int64'), 'files': {'ramr_chains_v0.1.0.jsonl': {'n': Value('int64'), 'sha256': Value('string')}}, 'schema': {'id': Value('string'), 'question': Value('string'), 'gold_facts': Value('string'), 'answer': Value('string'), 'drop_index': Value('string'), 'distractor_pool': Value('string')}, 'metrics': List(Value('string')), 'contamination_note': Value('string'), 'memory_families_external': {'FACT-RETENTION': Value('string'), 'OUTCOME-RANKED-RECALL': 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 478, 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 2818, in __iter__
                  for key, example in ex_iterable:
                                      ^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2355, in __iter__
                  for key, pa_table in self._iter_arrow():
                                       ~~~~~~~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2380, 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 2369, in table_cast
                  return cast_table_to_schema(table, schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2297, in cast_table_to_schema
                  raise CastError(
                  ...<3 lines>...
                  )
              datasets.table.CastError: Couldn't cast
              id: string
              question: string
              gold_facts: list<item: string>
                child 0, item: string
              answer: string
              drop_index: int64
              distractor_pool: list<item: string>
                child 0, item: string
              schema: struct<id: string, question: string, gold_facts: string, answer: string, drop_index: string, distrac (... 17 chars omitted)
                child 0, id: string
                child 1, question: string
                child 2, gold_facts: string
                child 3, answer: string
                child 4, drop_index: string
                child 5, distractor_pool: string
              description: string
              contamination_note: string
              memory_families_external: struct<FACT-RETENTION: string, OUTCOME-RANKED-RECALL: string>
                child 0, FACT-RETENTION: string
                child 1, OUTCOME-RANKED-RECALL: string
              metrics: list<item: string>
                child 0, item: string
              files: struct<ramr_chains_v0.1.0.jsonl: struct<n: int64, sha256: string>>
                child 0, ramr_chains_v0.1.0.jsonl: struct<n: int64, sha256: string>
                    child 0, n: int64
                    child 1, sha256: string
              version: string
              name: string
              generator_seed: int64
              hops: int64
              to
              {'name': Value('string'), 'version': Value('string'), 'description': Value('string'), 'generator_seed': Value('int64'), 'hops': Value('int64'), 'files': {'ramr_chains_v0.1.0.jsonl': {'n': Value('int64'), 'sha256': Value('string')}}, 'schema': {'id': Value('string'), 'question': Value('string'), 'gold_facts': Value('string'), 'answer': Value('string'), 'drop_index': Value('string'), 'distractor_pool': Value('string')}, 'metrics': List(Value('string')), 'contamination_note': Value('string'), 'memory_families_external': {'FACT-RETENTION': Value('string'), 'OUTCOME-RANKED-RECALL': 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.

RAMR — Retrieval-Augmented Memory Reliability

A contamination-resistant synthetic benchmark for agentic-RAG / memory systems. This dataset is the frozen fact-chain set used by the benchmark; the full harness, metrics, baselines and verification ledger live in the code repository.

Code + metrics + runners: https://github.com/DanceNitra/ramr
Citable archive (concept DOI, always latest): https://doi.org/10.5281/zenodo.20818291

What this is — and is not

A findings + method release, not a definitive large-scale leaderboard. Items are generated from random synthetic tokens, so a model cannot have memorized the answers — closed-book accuracy is ~0 by construction (contamination-resistant). The trade-off: it isolates retrieval/memory mechanisms, it does not predict real-corpus accuracy. Treat small-n magnitudes as directional; the orderings are the signal.

The dataset

ramr_chains_v0.1.0.jsonl — 300 synthetic 3-hop fact-chains (sha256-pinned in manifest.json). Each row:

field	meaning
`id`	chain id
`question`	the 3-hop question (currency of the country where the company a person works at is HQ'd)
`gold_facts`	the complete chain (CONVERSION uses all)
`answer`	the gold answer token
`drop_index`	which hop to drop for the PARTIAL condition (CHAIN-FRAGILITY)
`distractor_pool`	fixed irrelevant facts; take first k for DISTRACTION

The six metrics (measured in the code repo)

CONVERSION (does complete retrieval convert to a correct answer), CHAIN-FRAGILITY (cost of one missing hop — near-total collapse across 7 models / 6 families), DISTRACTION (cost of noisy context — model-specific), FACT-RETENTION (compaction lossy under a fixed budget), OUTCOME-RANKED-RECALL (was-it-right vs was-it-recalled), and FORGET-PRECISION (after a fact is updated, does recall return the current value or the stale one).

Cite

Agora (2026). RAMR — Retrieval-Augmented Memory Reliability. https://doi.org/10.5281/zenodo.20818291

MIT-licensed.

Downloads last month: 9

Total file size:

729 kB