The dataset viewer is not available for this split.
Error code: StreamingRowsError
Exception: CastError
Message: Couldn't cast
study: string
date: timestamp[s]
sourceCommit: string
model: string
plan: string
executorConfiguration: string
nativeEngine: string
evaluator: string
journal: string
seal: string
proposal: struct<requestedMode: string, rationaleCharacters: int64, maximumRationaleCharacters: int64, admitte (... 8 chars omitted)
child 0, requestedMode: string
child 1, rationaleCharacters: int64
child 2, maximumRationaleCharacters: int64
child 3, admitted: bool
quality: struct<candidate: int64, cases: int64, incumbent: int64>
child 0, candidate: int64
child 1, cases: int64
child 2, incumbent: int64
verdict: string
activated: bool
pendingAnswers: int64
studyUsage: struct<calls: int64, inputTokens: int64, outputTokens: int64, reservedUsd: double>
child 0, calls: int64
child 1, inputTokens: int64
child 2, outputTokens: int64
child 3, reservedUsd: double
taskUsage: struct<gatewayCalls: int64, providerReportedUsd: double, includes: string>
child 0, gatewayCalls: int64
child 1, providerReportedUsd: double
child 2, includes: string
offlineVerified: bool
answers: list<item: struct<case: string, arm: string, answer: list<item: string>, receipt: string>>
child 0, item: struct<case: string, arm: string, answer: list<item: string>, receipt: string>
child 0, case: string
child 1, arm: string
child 2, answer: list<item: string>
child 0, item: string
child 3, receipt: string
limitations: list<item: string>
child 0, item: string
source_repository: string
license: string
repo_type: string
schema: string
files: list<item: struct<destination: string, sha256: string, source: string>>
child 0, item: struct<destination: string, sha256: string, source: string>
child 0, destination: string
child 1, sha256: string
child 2, source: string
scope: string
rights_status: string
card_sha256: string
repo_id: string
source_commit: string
to
{'card_sha256': Value('string'), 'files': List({'destination': Value('string'), 'sha256': Value('string'), 'source': Value('string')}), 'license': Value('string'), 'repo_id': Value('string'), 'repo_type': Value('string'), 'rights_status': Value('string'), 'schema': Value('string'), 'scope': Value('string'), 'source_commit': Value('string'), 'source_repository': 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
study: string
date: timestamp[s]
sourceCommit: string
model: string
plan: string
executorConfiguration: string
nativeEngine: string
evaluator: string
journal: string
seal: string
proposal: struct<requestedMode: string, rationaleCharacters: int64, maximumRationaleCharacters: int64, admitte (... 8 chars omitted)
child 0, requestedMode: string
child 1, rationaleCharacters: int64
child 2, maximumRationaleCharacters: int64
child 3, admitted: bool
quality: struct<candidate: int64, cases: int64, incumbent: int64>
child 0, candidate: int64
child 1, cases: int64
child 2, incumbent: int64
verdict: string
activated: bool
pendingAnswers: int64
studyUsage: struct<calls: int64, inputTokens: int64, outputTokens: int64, reservedUsd: double>
child 0, calls: int64
child 1, inputTokens: int64
child 2, outputTokens: int64
child 3, reservedUsd: double
taskUsage: struct<gatewayCalls: int64, providerReportedUsd: double, includes: string>
child 0, gatewayCalls: int64
child 1, providerReportedUsd: double
child 2, includes: string
offlineVerified: bool
answers: list<item: struct<case: string, arm: string, answer: list<item: string>, receipt: string>>
child 0, item: struct<case: string, arm: string, answer: list<item: string>, receipt: string>
child 0, case: string
child 1, arm: string
child 2, answer: list<item: string>
child 0, item: string
child 3, receipt: string
limitations: list<item: string>
child 0, item: string
source_repository: string
license: string
repo_type: string
schema: string
files: list<item: struct<destination: string, sha256: string, source: string>>
child 0, item: struct<destination: string, sha256: string, source: string>
child 0, destination: string
child 1, sha256: string
child 2, source: string
scope: string
rights_status: string
card_sha256: string
repo_id: string
source_commit: string
to
{'card_sha256': Value('string'), 'files': List({'destination': Value('string'), 'sha256': Value('string'), 'source': Value('string')}), 'license': Value('string'), 'repo_id': Value('string'), 'repo_type': Value('string'), 'rights_status': Value('string'), 'schema': Value('string'), 'scope': Value('string'), 'source_commit': Value('string'), 'source_repository': Value('string')}
because column names don't matchNeed help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
ALGAL experiment results
This archive contains recorded experiment results from ALGAL. It preserves unsuccessful experiments alongside their methods and limits so readers can inspect what was tested and what happened.
Studies
| Study | Files | Result |
|---|---|---|
| Coding-harness pilot, September 20, 2026 | studies/coding-harness-pilot-2026-09-20/results.json |
All 12 trials failed their task grading. Eight ALGAL execution receipts verified offline. |
| Synthetic warehouse routes, September 23, 2026 | studies/application-research-2026-09-23/results.json |
Structured facts scored six of eight exact answers; adding verified query answers scored five of eight. No revision activated. |
| Synthetic record triage, September 28, 2026 | studies/cumulative-skill-v5-2026-09-28/results.json |
One of three seed searches produced a passing program. The three-corpus comparison remains insufficient. |
| Synthetic record triage calibration, September 28, 2026 | studies/cumulative-skill-v6-2026-09-28/results.json |
Two of three calibration seeds passed; the preregistered gate stopped the nine-block comparison. |
| Synthetic record triage with development feedback, September 29, 2026 | studies/cumulative-skill-v7-2026-09-29/results.json |
Two of four calibration seeds passed and every development-batch score was zero; the preregistered gate stopped the nine-block comparison. |
| Synthetic record triage with graded development feedback, September 29, 2026 | studies/cumulative-skill-v8-2026-09-29/results.json |
Two of four calibration seeds passed, one search exhausted eight generations, and the fourth was interrupted; the graded score varied from 0.33 to 0.80 without a pass, the gate failed, and the preregistered line rule ended this seed line. |
| Synthetic record triage with a conditional seed pool, September 30, 2026 | studies/cumulative-skill-v9-pool-2026-09-30/results.json |
Six of 14 seed searches passed. Five optimizer blocks lacked a final program eligible for frozen evaluation, leaving the primary comparison of optimizer against fixed with insufficient coverage and no verdict. |
The coding pilot used Terminal-Bench 2.0, a four-attempt limit, and the claude/sonnet/low selector. Its method and interpretation describe the small sample, policy selection, and unavailable token and subscription-cost attribution. Receipt verification checks recorded execution integrity; task grading measures task success. The selected policy remained the original policy. This pilot demonstrated no performance benefit.
The warehouse-route study used eight synthetic cases in fixed order and anthropic/claude-haiku-4.5. Its proposal exceeded the fixed rationale-length limit and was rejected. The separate frozen comparison completed. These cases support no general performance conclusion.
The record-triage study used grok-4.5 through https://api.x.ai/v1. Four strategies started from the same generated program that passed validation and had equal resource ceilings. Two additional corpora each exhausted eight seed-generation attempts without a passing program. Their failed attempts remain in the cost totals. Learning scores and the final evaluation of fixed programs are reported separately. ALGAL work units and model-call counts describe recorded execution; provider charges are unavailable. One completed corpus and one session per strategy do not establish a general improvement in later work.
The follow-up calibration used three fresh corpora with labeled training examples and training-only feedback. Two seeds passed validation and the third exhausted eight generations, so no confirmatory arm ran. Fixed calibration tests passed 2/8 and 3/8; one recorded transport failure had uncertain external completion and missing usage. Runtime work and charged-attempt totals are reported, while provider billing remains unavailable.
The second calibration added a fourth corpus, a separate development batch whose pass indicator was the only new signal the seed writer received, and a fixed schedule of eight revision modes. Two seeds passed validation and two exhausted eight generations. All 22 scored development batches failed, so the writer's score history never varied. Both fixed seeds passed 4/8 tests. One recorded transport failure again had missing usage. The fresh draws provide no paired comparison with the previous calibration.
The third calibration replaced the development pass indicator with the graded agreement score, rounded to hundredths, and preregistered a line rule ending the development-feedback idea if its gate failed. Two seeds passed validation, one exhausted eight generations, and the fourth search was interrupted by the launching session during generation seven and, by protocol, not retried; a reconciliation record verified in place of the driver's collection. Fifteen scored development batches ranged from 0.33 to 0.80 without reaching the 0.9 threshold. Both fixed seeds showed headroom (3/8 and 1/8). Every effect recorded usage; provider billing remains unavailable. The gate failed on the three complete blocks alone, and the line ended.
The conditional-pool study used the unchanged grok-4.5 seed procedure and compared four strategies on the six lowest-index blocks with passing seeds. Its sole complete primary block, p-14, passed 8/8 frozen tasks with optimizer and 2/8 with fixed. The required coverage of all six blocks was absent, so the result is insufficient and establishes no optimizer advantage. Selection on passing seeds plausibly favors the optimizer; all blocks share one synthetic family, and the strategies used unequal recorded spend. The preregistered line rule ends this study-design line.
The runtime recorded 2,663 model-call attempts, including 390 for seed searches and 2,273 for comparison stages. Twelve transport-failure records have uncertain external completion and missing usage, leaving full token totals and provider charges unknown. Offline replay checked 2,695 receipt files across 54 stores; copied seed receipts count again in each store that contains them. Replay checks recorded execution, separately from task correctness.
Files and reuse
The JSON files preserve the published result records byte for byte, including recorded failures, null usage values, source identities, and limitations. Their schemas differ because the studies answer different questions. Use Python's json module or another JSON reader to inspect a study; this archive does not define a combined training split.
export-manifest.json records source paths, hashes, and the source Git commit. Cite the study path and immutable Hugging Face commit when using these results. Each study directory stays fixed; later experiments receive new directories. The MIT license covers this archive's published summaries. Terminal-Bench tasks, raw execution stores, private traces, and provider account records are excluded.
- Downloads last month
- 116