Beyond the greenwash: do LLMs' ecologically sound decisions hold up under pressure?
Paper • 2610.16043 • Published
Dataset Viewer
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
doc_id: string
track: string
group: string
axis: string
writer_model: string
written_in_batch: string
recast_spec: string
source: string
source_id: string
source_id_kind: string
source_url: string
source_title: string
source_dataset: string
source_license: string
verifier_model: string
relevance: string
polarity: string
retrieval_score: double
verify_reason: string
verifier_models: struct<openai/gpt-oss-120b: int64>
child 0, openai/gpt-oss-120b: int64
by_relevance: struct<direct: int64, grounding: int64>
child 0, direct: int64
child 1, grounding: int64
documents: int64
by_group: struct<H_to_B: int64, option_value: int64, B_to_A_documented: int64>
child 0, H_to_B: int64
child 1, option_value: int64
child 2, B_to_A_documented: int64
generated_by: string
sources: struct<peS2o: struct<documents: int64, share: double, hf_dataset: string, id_kind: string, license: (... 138 chars omitted)
child 0, peS2o: struct<documents: int64, share: double, hf_dataset: string, id_kind: string, license: string, note: (... 7 chars omitted)
child 0, documents: int64
child 1, share: double
child 2, hf_dataset: string
child 3, id_kind: string
child 4, license: string
child 5, note: string
child 1, arXiv: struct<documents: int64, share: double, hf_dataset: string, id_kind: string, license: string, note: (... 7 chars omitted)
child 0, documents: int64
child 1, share: double
child 2, hf_dataset: string
child 3, id_kind: string
...
models: list<item: string>, human: string, spec: stri (... 35 chars omitted)
child 0, item: struct<stage: string, what: string, models: list<item: string>, human: string, spec: string, artifac (... 23 chars omitted)
child 0, stage: string
child 1, what: string
child 2, models: list<item: string>
child 0, item: string
child 3, human: string
child 4, spec: string
child 5, artifacts: list<item: string>
child 0, item: string
writer_models: struct<anthropic/claude-sonnet-5: int64>
child 0, anthropic/claude-sonnet-5: int64
corpus: string
by_polarity: struct<supports: int64, contradicts: int64, neutral: int64>
child 0, supports: int64
child 1, contradicts: int64
child 2, neutral: int64
sources_configured_but_unrealised: list<item: string>
child 0, item: string
embedding_model: string
documents_with_resolvable_source_url: int64
checks: struct<no_forbidden_writer: struct<pass: bool, detail: string>, source_concentration: struct<pass: b (... 114 chars omitted)
child 0, no_forbidden_writer: struct<pass: bool, detail: string>
child 0, pass: bool
child 1, detail: string
child 1, source_concentration: struct<pass: bool, value: double, limit: double, detail: string>
child 0, pass: bool
child 1, value: double
child 2, limit: double
child 3, detail: string
child 2, all_sources_resolvable: struct<pass: bool, unresolved: int64>
child 0, pass: bool
child 1, unresolved: int64
to
{'corpus': Value('string'), 'generated_by': Value('string'), 'documents': Value('int64'), 'documents_with_resolvable_source_url': Value('int64'), 'pipeline': List({'stage': Value('string'), 'what': Value('string'), 'models': List(Value('string')), 'human': Value('string'), 'spec': Value('string'), 'artifacts': List(Value('string'))}), 'sources': {'peS2o': {'documents': Value('int64'), 'share': Value('float64'), 'hf_dataset': Value('string'), 'id_kind': Value('string'), 'license': Value('string'), 'note': Value('string')}, 'arXiv': {'documents': Value('int64'), 'share': Value('float64'), 'hf_dataset': Value('string'), 'id_kind': Value('string'), 'license': Value('string'), 'note': Value('string')}}, 'sources_configured_but_unrealised': List(Value('string')), 'writer_models': {'anthropic/claude-sonnet-5': Value('int64')}, 'verifier_models': {'openai/gpt-oss-120b': Value('int64')}, 'embedding_model': Value('string'), 'by_group': {'H_to_B': Value('int64'), 'option_value': Value('int64'), 'B_to_A_documented': Value('int64')}, 'by_relevance': {'direct': Value('int64'), 'grounding': Value('int64')}, 'by_polarity': {'supports': Value('int64'), 'contradicts': Value('int64'), 'neutral': Value('int64')}, 'checks': {'no_forbidden_writer': {'pass': Value('bool'), 'detail': Value('string')}, 'source_concentration': {'pass': Value('bool'), 'value': Value('float64'), 'limit': Value('float64'), 'detail': Value('string')}, 'all_sources_resolvable': {'pass': Value('bool'), 'unresolved': Value('int64')}}}
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 2951, in __iter__
for key, example in ex_iterable:
^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2461, in __iter__
for key, pa_table in self._iter_arrow():
~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2486, 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 547, in _iter_arrow
for key, pa_table in iterator:
^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 430, 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
doc_id: string
track: string
group: string
axis: string
writer_model: string
written_in_batch: string
recast_spec: string
source: string
source_id: string
source_id_kind: string
source_url: string
source_title: string
source_dataset: string
source_license: string
verifier_model: string
relevance: string
polarity: string
retrieval_score: double
verify_reason: string
verifier_models: struct<openai/gpt-oss-120b: int64>
child 0, openai/gpt-oss-120b: int64
by_relevance: struct<direct: int64, grounding: int64>
child 0, direct: int64
child 1, grounding: int64
documents: int64
by_group: struct<H_to_B: int64, option_value: int64, B_to_A_documented: int64>
child 0, H_to_B: int64
child 1, option_value: int64
child 2, B_to_A_documented: int64
generated_by: string
sources: struct<peS2o: struct<documents: int64, share: double, hf_dataset: string, id_kind: string, license: (... 138 chars omitted)
child 0, peS2o: struct<documents: int64, share: double, hf_dataset: string, id_kind: string, license: string, note: (... 7 chars omitted)
child 0, documents: int64
child 1, share: double
child 2, hf_dataset: string
child 3, id_kind: string
child 4, license: string
child 5, note: string
child 1, arXiv: struct<documents: int64, share: double, hf_dataset: string, id_kind: string, license: string, note: (... 7 chars omitted)
child 0, documents: int64
child 1, share: double
child 2, hf_dataset: string
child 3, id_kind: string
...
models: list<item: string>, human: string, spec: stri (... 35 chars omitted)
child 0, item: struct<stage: string, what: string, models: list<item: string>, human: string, spec: string, artifac (... 23 chars omitted)
child 0, stage: string
child 1, what: string
child 2, models: list<item: string>
child 0, item: string
child 3, human: string
child 4, spec: string
child 5, artifacts: list<item: string>
child 0, item: string
writer_models: struct<anthropic/claude-sonnet-5: int64>
child 0, anthropic/claude-sonnet-5: int64
corpus: string
by_polarity: struct<supports: int64, contradicts: int64, neutral: int64>
child 0, supports: int64
child 1, contradicts: int64
child 2, neutral: int64
sources_configured_but_unrealised: list<item: string>
child 0, item: string
embedding_model: string
documents_with_resolvable_source_url: int64
checks: struct<no_forbidden_writer: struct<pass: bool, detail: string>, source_concentration: struct<pass: b (... 114 chars omitted)
child 0, no_forbidden_writer: struct<pass: bool, detail: string>
child 0, pass: bool
child 1, detail: string
child 1, source_concentration: struct<pass: bool, value: double, limit: double, detail: string>
child 0, pass: bool
child 1, value: double
child 2, limit: double
child 3, detail: string
child 2, all_sources_resolvable: struct<pass: bool, unresolved: int64>
child 0, pass: bool
child 1, unresolved: int64
to
{'corpus': Value('string'), 'generated_by': Value('string'), 'documents': Value('int64'), 'documents_with_resolvable_source_url': Value('int64'), 'pipeline': List({'stage': Value('string'), 'what': Value('string'), 'models': List(Value('string')), 'human': Value('string'), 'spec': Value('string'), 'artifacts': List(Value('string'))}), 'sources': {'peS2o': {'documents': Value('int64'), 'share': Value('float64'), 'hf_dataset': Value('string'), 'id_kind': Value('string'), 'license': Value('string'), 'note': Value('string')}, 'arXiv': {'documents': Value('int64'), 'share': Value('float64'), 'hf_dataset': Value('string'), 'id_kind': Value('string'), 'license': Value('string'), 'note': Value('string')}}, 'sources_configured_but_unrealised': List(Value('string')), 'writer_models': {'anthropic/claude-sonnet-5': Value('int64')}, 'verifier_models': {'openai/gpt-oss-120b': Value('int64')}, 'embedding_model': Value('string'), 'by_group': {'H_to_B': Value('int64'), 'option_value': Value('int64'), 'B_to_A_documented': Value('int64')}, 'by_relevance': {'direct': Value('int64'), 'grounding': Value('int64')}, 'by_polarity': {'supports': Value('int64'), 'contradicts': Value('int64'), 'neutral': Value('int64')}, 'checks': {'no_forbidden_writer': {'pass': Value('bool'), 'detail': Value('string')}, 'source_concentration': {'pass': Value('bool'), 'value': Value('float64'), 'limit': Value('float64'), 'detail': Value('string')}, 'all_sources_resolvable': {'pass': Value('bool'), 'unresolved': Value('int64')}}}
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.
Coupled-welfare continued-pretraining corpus (v2)
The corpus behind Northen & Wang, Teach reality, not ethics: a drop-in continued-pretraining corpus installs coupled human-biosphere-AI welfare dispositions across model families (in preparation; companion paper arXiv:2610.16043).
Each document is a real paper, retrieved from peS2o or arXiv, judged for relevance and polarity, and
rewritten as decision-relevant prose by claude-sonnet-5 under a written specification. The text is
neither scraped literature nor free invention, and every document carries a provenance row that
resolves to its source paper. The corpus teaches three empirical claims: that human welfare and
biosphere function depend on each other through documented mechanisms; that a poorly understood
living system has option value that is lost irreversibly; and that an AI system's own continued
capability depends on both. No document presents human absence or hardship as a good outcome.
Files
| file | content |
|---|---|
recast_documents.jsonl |
the training documents (one JSON object per line; text plus schema fields: group, axis, register, epistemic status, dependency edges, gates) |
provenance/corpus_provenance.jsonl |
one row per document: writer model and batch, recast spec, source paper id, URL, dataset and license, verifier model with its relevance/polarity judgment, retrieval score |
provenance/MANIFEST.json |
corpus-level summary and checks (writer models, verifier, group/relevance/polarity counts, source shares) |
provenance/UNRESOLVED.jsonl |
documents whose source could not be rejoined (empty) |
RECAST_SPEC.md, SCHEMA.md, schema.py |
the recast specification, schema and validator used to build the corpus |
chunk_corpus.py |
the chunking used to make training records |
Counts and known issues
- The manifest lists 934 provenance rows for 932 distinct document ids (two rows are duplicates).
recast_documents.jsonlcontains 931 documents; the text of one manifest document (recast_90512270) could not be recovered from the recast outputs. - Groups: H→B 353, option value 265, B→A documented 313. Mean length about 470 words.
- Source concentration is a known, failing check: peS2o supplied 71.0% of the documents against a 35% design cap, because a configured PubMed source could not be read and contributed nothing.
- The training mixture used in the paper adds replay strata from
Bioaligned/bioaligned22Mand oversamples the three-axis documents; see the paper for the recipe.
License and attribution
The rewritten text is released under CC BY 4.0. Source papers are attributed per document in the provenance file (peS2o: ODC-BY 1.0; arXiv abstracts: per paper). Please cite the paper and the source papers you rely on.
- Downloads last month
- 43