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
id: string
bench: string
sample: int64
gold: string
finish_reason: string
n_tokens: int64
text: string
per_problem: list<item: struct<id: string, bench: string, gold: string, preds: list<item: string>, hits: list<ite (... 50 chars omitted)
child 0, item: struct<id: string, bench: string, gold: string, preds: list<item: string>, hits: list<item: int64>, (... 38 chars omitted)
child 0, id: string
child 1, bench: string
child 2, gold: string
child 3, preds: list<item: string>
child 0, item: string
child 4, hits: list<item: int64>
child 0, item: int64
child 5, truncated: int64
child 6, mean_tokens: double
summary: struct<label: string, model: string, adapter: null, n_samples: int64, temperature: double, max_new_t (... 224 chars omitted)
child 0, label: string
child 1, model: string
child 2, adapter: null
child 3, n_samples: int64
child 4, temperature: double
child 5, max_new_tokens: int64
child 6, max_new_tokens_requested: int64
child 7, model_context: int64
child 8, generations_file: string
child 9, benchmarks: struct<train: struct<accuracy: double, n_problems: int64, truncated_generations: int64, mean_output_ (... 16 chars omitted)
child 0, train: struct<accuracy: double, n_problems: int64, truncated_generations: int64, mean_output_tokens: double (... 1 chars omitted)
child 0, accuracy: double
child 1, n_problems: int64
child 2, truncated_generations: int64
child 3, mean_output_tokens: double
to
{'summary': {'label': Value('string'), 'model': Value('string'), 'adapter': Value('null'), 'n_samples': Value('int64'), 'temperature': Value('float64'), 'max_new_tokens': Value('int64'), 'max_new_tokens_requested': Value('int64'), 'model_context': Value('int64'), 'generations_file': Value('string'), 'benchmarks': {'train': {'accuracy': Value('float64'), 'n_problems': Value('int64'), 'truncated_generations': Value('int64'), 'mean_output_tokens': Value('float64')}}}, 'per_problem': List({'id': Value('string'), 'bench': Value('string'), 'gold': Value('string'), 'preds': List(Value('string')), 'hits': List(Value('int64')), 'truncated': Value('int64'), 'mean_tokens': Value('float64')})}
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
id: string
bench: string
sample: int64
gold: string
finish_reason: string
n_tokens: int64
text: string
per_problem: list<item: struct<id: string, bench: string, gold: string, preds: list<item: string>, hits: list<ite (... 50 chars omitted)
child 0, item: struct<id: string, bench: string, gold: string, preds: list<item: string>, hits: list<item: int64>, (... 38 chars omitted)
child 0, id: string
child 1, bench: string
child 2, gold: string
child 3, preds: list<item: string>
child 0, item: string
child 4, hits: list<item: int64>
child 0, item: int64
child 5, truncated: int64
child 6, mean_tokens: double
summary: struct<label: string, model: string, adapter: null, n_samples: int64, temperature: double, max_new_t (... 224 chars omitted)
child 0, label: string
child 1, model: string
child 2, adapter: null
child 3, n_samples: int64
child 4, temperature: double
child 5, max_new_tokens: int64
child 6, max_new_tokens_requested: int64
child 7, model_context: int64
child 8, generations_file: string
child 9, benchmarks: struct<train: struct<accuracy: double, n_problems: int64, truncated_generations: int64, mean_output_ (... 16 chars omitted)
child 0, train: struct<accuracy: double, n_problems: int64, truncated_generations: int64, mean_output_tokens: double (... 1 chars omitted)
child 0, accuracy: double
child 1, n_problems: int64
child 2, truncated_generations: int64
child 3, mean_output_tokens: double
to
{'summary': {'label': Value('string'), 'model': Value('string'), 'adapter': Value('null'), 'n_samples': Value('int64'), 'temperature': Value('float64'), 'max_new_tokens': Value('int64'), 'max_new_tokens_requested': Value('int64'), 'model_context': Value('int64'), 'generations_file': Value('string'), 'benchmarks': {'train': {'accuracy': Value('float64'), 'n_problems': Value('int64'), 'truncated_generations': Value('int64'), 'mean_output_tokens': Value('float64')}}}, 'per_problem': List({'id': Value('string'), 'bench': Value('string'), 'gold': Value('string'), 'preds': List(Value('string')), 'hits': List(Value('int64')), 'truncated': Value('int64'), 'mean_tokens': Value('float64')})}
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.
DetectDistill — reproducibility archive (2026-11-03)
Project-level archive of the DetectDistill research project (teacher attribution / distillation detection for LLM students).
It complements the experiment repositories of francescortu and records, for every historical experiment, where each input,
output and checkpoint lives, with sha256 manifests and pinned revisions.
Layout
inventory/— the reproducibility inventory:inventory.md/inventory.json(per experiment: runs incl. failed and incomplete ones, inputs with revisions, artifact locations, missing items, third-party constraints), the local file manifests (manifests/local/*.jsonl: every local file with size and sha256, 33,004 files / 9.39 TB), the remote manifests of every DetectDistill HF repo at a pinned revision (manifests/remote/), the archive plan (plan.jsonl.gz: status of every local file), environment locks (environment/: pip freezes of the six virtual environments, system info), upload records (uploads.jsonl) and restoration checks (restore_checks.jsonl).data/<experiment>/...— scratch data that had no public copy (evaluations, generations, traces, training data, pools, configs), mirrored at its original relative path.records/— run records (runs.md, specs, READMEs, notes), results and reports (records/results/<experiment>.tar.gz,records/report/<experiment>.tar.gz, each with a.members.jsonlmanifest), the environment specification (pyproject.toml,uv.lock).logs/<experiment>.tar.gz— experiment logs (Slurm / HTCondor job logs), with member manifests.MANIFESTS/<name>.jsonl— per upload batch: every file, its size, sha256, local source and the revision it was verified at.
What is NOT here (and where it is)
- Code (
src/,cluster_scripts/) is kept in the private git repository (owner decision); this archive holds records and data only. - Checkpoints already public:
francescortu/DistillDetect-*,DistillDetect-traj-*,DistillDetect-ctrl-*,detectdistill-v2-lora-students,detectdistill-v5-lora-students,Llama-3.1-8B-detectdistill-v3-v4-lora(byte-identity with the local copies is recorded ininventory/plan.jsonl.gz, statusarchived). - Regenerable: merged bf16 LoRA exports (base at its pinned revision + public adapter; bit-identical re-merge verified, see
inventory/restore_checks.jsonl), Hugging Face cache copies, text-only exports of pinned snapshots. - Not archived by owner decision (no reported result depends on them; path, size and sha256 recorded): optimizer states, the intermediate FSDP checkpoints of exp_controlled_distill_2026-08-26, the weights of 13 unpublished trajectory runs (their evaluations are here), exp_students_v2 weights (never evaluated), engineering sanity adapters.
- Private by owner decision: the Qwen3.5 teacher traces of exp_controlled_distill_2026-08-26 and the v2 prompt file.
- Third-party data not redistributed: Rawat et al.'s DistillDetect data (github.com/RajatRawat-creator/DistillDetect, commit b7cfbae, MIT; contains vendor API outputs) and the reference papers.
- Experiments run on the MPI-IS cluster (Aug 2026) whose generated data never reached Orfeo or the Hub are listed as missing.
Licenses and terms
Mixed. Teacher outputs come from open-weight models (gpt-oss-120b, GLM-5.x, Gemma-4, Qwen3, MiMo-V2.5, Granite, OLMo-3, SmolLM3, Hunyuan) and, in run_v39, Claude Haiku rewrites; questions come from public benchmarks (SuperGPQA, MATH, Codeforces, AIME, GPQA-Diamond, dolly-15k) under their own licenses; student outputs inherit the base-model licenses (Llama 3.x, Qwen, Mistral, Falcon, Gemma). Use each part under its original terms. Evaluation questions are included for reproducibility only.
- Downloads last month
- 272