Dataset Preview
Duplicate
The full dataset viewer is not available (click to read why). Only showing a preview of the rows.
The dataset generation failed because of a cast error
Error code:   DatasetGenerationCastError
Exception:    DatasetGenerationCastError
Message:      An error occurred while generating the dataset

All the data files must have the same columns, but at some point there are 11 new columns ({'sample_id', 'correct_raw', 'correct_norm', 'scores_ll_4cand', 'task', 'gold', 'model', 'pred_raw', 'truncated', 'passage_or_group', 'pred_norm'}) and 6 missing columns ({'n', 'ci_hi', 'acc_pct', 'subject', 'ci_lo', 'acc'}).

This happened while the csv dataset builder was generating data using

hf://datasets/umeiko/Haidass-Translate-143M-eval/base_capability_eval/predictions_all.csv (at revision c84c7cfdf4626f9347a348e09cb1a57ca013ebeb), ['hf://datasets/umeiko/Haidass-Translate-143M-eval@c84c7cfdf4626f9347a348e09cb1a57ca013ebeb/base_capability_eval/cmmlu_subjects_cagliostro_v3.csv', 'hf://datasets/umeiko/Haidass-Translate-143M-eval@c84c7cfdf4626f9347a348e09cb1a57ca013ebeb/base_capability_eval/cmmlu_subjects_gptx25_135m.csv', 'hf://datasets/umeiko/Haidass-Translate-143M-eval@c84c7cfdf4626f9347a348e09cb1a57ca013ebeb/base_capability_eval/cmmlu_subjects_haidass143m.csv', 'hf://datasets/umeiko/Haidass-Translate-143M-eval@c84c7cfdf4626f9347a348e09cb1a57ca013ebeb/base_capability_eval/cmmlu_subjects_mobilellm_r1_140m.csv', 'hf://datasets/umeiko/Haidass-Translate-143M-eval@c84c7cfdf4626f9347a348e09cb1a57ca013ebeb/base_capability_eval/cmmlu_subjects_qwen25_05b.csv', 'hf://datasets/umeiko/Haidass-Translate-143M-eval@c84c7cfdf4626f9347a348e09cb1a57ca013ebeb/base_capability_eval/cmmlu_subjects_qwen3_06b.csv', 'hf://datasets/umeiko/Haidass-Translate-143M-eval@c84c7cfdf4626f9347a348e09cb1a57ca013ebeb/base_capability_eval/cmmlu_subjects_smollm2_135m.csv', 'hf://datasets/umeiko/Haidass-Translate-143M-eval@c84c7cfdf4626f9347a348e09cb1a57ca013ebeb/base_capability_eval/predictions_all.csv', 'hf://datasets/umeiko/Haidass-Translate-143M-eval@c84c7cfdf4626f9347a348e09cb1a57ca013ebeb/base_capability_eval/slm_summary.csv', 'hf://datasets/umeiko/Haidass-Translate-143M-eval@c84c7cfdf4626f9347a348e09cb1a57ca013ebeb/base_capability_eval/summary.csv', 'hf://datasets/umeiko/Haidass-Translate-143M-eval@c84c7cfdf4626f9347a348e09cb1a57ca013ebeb/base_capability_eval/summary_accnorm.csv', 'hf://datasets/umeiko/Haidass-Translate-143M-eval@c84c7cfdf4626f9347a348e09cb1a57ca013ebeb/base_capability_eval/summary_ci.csv']

Please either edit the data files to have matching columns, or separate them into different configurations (see docs at https://hf.co/docs/hub/datasets-manual-configuration#multiple-configurations)
Traceback:    Traceback (most recent call last):
                File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1848, in _prepare_split_single
                  writer.write_table(table)
                  ~~~~~~~~~~~~~~~~~~^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/arrow_writer.py", line 784, in write_table
                  self._write_table(pa_table, writer_batch_size=writer_batch_size)
                  ~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/arrow_writer.py", line 795, in _write_table
                  pa_table = table_cast(pa_table, self._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
              model: string
              task: string
              sample_id: string
              passage_or_group: string
              gold: string
              scores_ll_4cand: string
              pred_norm: string
              correct_norm: int64
              pred_raw: string
              correct_raw: int64
              truncated: int64
              -- schema metadata --
              pandas: '{"index_columns": [{"kind": "range", "name": null, "start": 0, "' + 1576
              to
              {'subject': Value('string'), 'n': Value('int64'), 'acc': Value('float64'), 'acc_pct': Value('float64'), 'ci_lo': Value('float64'), 'ci_hi': Value('float64')}
              because column names don't match
              
              During handling of the above exception, another exception occurred:
              
              Traceback (most recent call last):
                File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 1369, in compute_config_parquet_and_info_response
                  parquet_operations, partial, estimated_dataset_info = stream_convert_to_parquet(
                                                                        ~~~~~~~~~~~~~~~~~~~~~~~~~^
                      builder, max_dataset_size_bytes=max_dataset_size_bytes
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                  )
                  ^
                File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 948, in stream_convert_to_parquet
                  builder._prepare_split(split_generator=splits_generators[split], file_format="parquet")
                  ~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1694, in _prepare_split
                  for job_id, done, content in self._prepare_split_single(
                                               ~~~~~~~~~~~~~~~~~~~~~~~~~~^
                      gen_kwargs=gen_kwargs, job_id=job_id, **_prepare_split_args
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                  ):
                  ^
                File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1850, in _prepare_split_single
                  raise DatasetGenerationCastError.from_cast_error(
                  ...<4 lines>...
                  )
              datasets.exceptions.DatasetGenerationCastError: An error occurred while generating the dataset
              
              All the data files must have the same columns, but at some point there are 11 new columns ({'sample_id', 'correct_raw', 'correct_norm', 'scores_ll_4cand', 'task', 'gold', 'model', 'pred_raw', 'truncated', 'passage_or_group', 'pred_norm'}) and 6 missing columns ({'n', 'ci_hi', 'acc_pct', 'subject', 'ci_lo', 'acc'}).
              
              This happened while the csv dataset builder was generating data using
              
              hf://datasets/umeiko/Haidass-Translate-143M-eval/base_capability_eval/predictions_all.csv (at revision c84c7cfdf4626f9347a348e09cb1a57ca013ebeb), ['hf://datasets/umeiko/Haidass-Translate-143M-eval@c84c7cfdf4626f9347a348e09cb1a57ca013ebeb/base_capability_eval/cmmlu_subjects_cagliostro_v3.csv', 'hf://datasets/umeiko/Haidass-Translate-143M-eval@c84c7cfdf4626f9347a348e09cb1a57ca013ebeb/base_capability_eval/cmmlu_subjects_gptx25_135m.csv', 'hf://datasets/umeiko/Haidass-Translate-143M-eval@c84c7cfdf4626f9347a348e09cb1a57ca013ebeb/base_capability_eval/cmmlu_subjects_haidass143m.csv', 'hf://datasets/umeiko/Haidass-Translate-143M-eval@c84c7cfdf4626f9347a348e09cb1a57ca013ebeb/base_capability_eval/cmmlu_subjects_mobilellm_r1_140m.csv', 'hf://datasets/umeiko/Haidass-Translate-143M-eval@c84c7cfdf4626f9347a348e09cb1a57ca013ebeb/base_capability_eval/cmmlu_subjects_qwen25_05b.csv', 'hf://datasets/umeiko/Haidass-Translate-143M-eval@c84c7cfdf4626f9347a348e09cb1a57ca013ebeb/base_capability_eval/cmmlu_subjects_qwen3_06b.csv', 'hf://datasets/umeiko/Haidass-Translate-143M-eval@c84c7cfdf4626f9347a348e09cb1a57ca013ebeb/base_capability_eval/cmmlu_subjects_smollm2_135m.csv', 'hf://datasets/umeiko/Haidass-Translate-143M-eval@c84c7cfdf4626f9347a348e09cb1a57ca013ebeb/base_capability_eval/predictions_all.csv', 'hf://datasets/umeiko/Haidass-Translate-143M-eval@c84c7cfdf4626f9347a348e09cb1a57ca013ebeb/base_capability_eval/slm_summary.csv', 'hf://datasets/umeiko/Haidass-Translate-143M-eval@c84c7cfdf4626f9347a348e09cb1a57ca013ebeb/base_capability_eval/summary.csv', 'hf://datasets/umeiko/Haidass-Translate-143M-eval@c84c7cfdf4626f9347a348e09cb1a57ca013ebeb/base_capability_eval/summary_accnorm.csv', 'hf://datasets/umeiko/Haidass-Translate-143M-eval@c84c7cfdf4626f9347a348e09cb1a57ca013ebeb/base_capability_eval/summary_ci.csv']
              
              Please either edit the data files to have matching columns, or separate them into different configurations (see docs at https://hf.co/docs/hub/datasets-manual-configuration#multiple-configurations)

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.

subject
string
n
int64
acc
float64
acc_pct
float64
ci_lo
float64
ci_hi
float64
agronomy
169
0.242604
24.26
0.1775
0.3077
anatomy
148
0.256757
25.68
0.1892
0.3378
ancient_chinese
164
0.256098
25.61
0.1951
0.3232
arts
160
0.25625
25.62
0.1938
0.3312
astronomy
165
0.248485
24.85
0.1818
0.3152
business_ethics
209
0.248804
24.88
0.1914
0.311
chinese_civil_service_exam
160
0.25625
25.62
0.1938
0.3187
chinese_driving_rule
131
0.251908
25.19
0.1756
0.3282
chinese_food_culture
136
0.25
25
0.1838
0.3309
chinese_foreign_policy
107
0.252336
25.23
0.1682
0.3364
chinese_history
323
0.250774
25.08
0.2043
0.3003
chinese_literature
204
0.254902
25.49
0.1961
0.3137
chinese_teacher_qualification
179
0.251397
25.14
0.1899
0.3184
clinical_knowledge
237
0.253165
25.32
0.1983
0.308
college_actuarial_science
106
0.245283
24.53
0.1792
0.3302
college_education
107
0.317757
31.78
0.2336
0.4019
college_engineering_hydrology
106
0.301887
30.19
0.217
0.3868
college_law
108
0.212963
21.3
0.1481
0.2963
college_mathematics
105
0.219048
21.9
0.1429
0.2952
college_medical_statistics
106
0.254717
25.47
0.1792
0.3396
college_medicine
273
0.241758
24.18
0.1941
0.293
computer_science
204
0.254902
25.49
0.1961
0.3186
computer_security
171
0.25731
25.73
0.193
0.3216
conceptual_physics
147
0.251701
25.17
0.1837
0.3265
construction_project_management
139
0.244604
24.46
0.1799
0.3165
economics
159
0.245283
24.53
0.1761
0.3145
education
163
0.251534
25.15
0.184
0.319
electrical_engineering
172
0.25
25
0.186
0.314
elementary_chinese
252
0.281746
28.17
0.2262
0.3413
elementary_commonsense
198
0.242424
24.24
0.1869
0.303
elementary_information_and_technology
238
0.268908
26.89
0.2101
0.3277
elementary_mathematics
230
0.278261
27.83
0.2217
0.3348
ethnology
135
0.251852
25.19
0.1778
0.3259
food_science
143
0.251748
25.17
0.1818
0.3217
genetics
176
0.244318
24.43
0.1818
0.3125
global_facts
149
0.248322
24.83
0.1812
0.3221
high_school_biology
169
0.248521
24.85
0.1834
0.3136
high_school_chemistry
132
0.257576
25.76
0.1818
0.3333
high_school_geography
118
0.254237
25.42
0.178
0.339
high_school_mathematics
164
0.25
25
0.1829
0.3171
high_school_physics
110
0.254545
25.45
0.1727
0.3455
high_school_politics
143
0.251748
25.17
0.1818
0.3287
human_sexuality
126
0.253968
25.4
0.1825
0.3333
international_law
185
0.248649
24.86
0.1892
0.3135
journalism
172
0.25
25
0.186
0.314
jurisprudence
411
0.250608
25.06
0.2117
0.292
legal_and_moral_basis
214
0.252336
25.23
0.1963
0.3084
logical
123
0.252033
25.2
0.1789
0.3333
machine_learning
122
0.254098
25.41
0.1721
0.3279
management
210
0.247619
24.76
0.1952
0.3048
marketing
180
0.25
25
0.1944
0.3167
marxist_theory
189
0.248677
24.87
0.1905
0.3122
modern_chinese
116
0.25
25
0.1724
0.3362
nutrition
145
0.255172
25.52
0.1862
0.3241
philosophy
105
0.247619
24.76
0.1619
0.3333
professional_accounting
175
0.251429
25.14
0.1943
0.32
professional_law
211
0.251185
25.12
0.1896
0.3081
professional_medicine
376
0.25
25
0.2048
0.2979
professional_psychology
232
0.25
25
0.1983
0.3017
public_relations
174
0.252874
25.29
0.1954
0.3161
security_study
135
0.251852
25.19
0.1778
0.3259
sociology
226
0.252212
25.22
0.1991
0.3097
sports_science
165
0.248485
24.85
0.1818
0.3212
traditional_chinese_medicine
185
0.248649
24.86
0.1838
0.3189
virology
169
0.248521
24.85
0.1834
0.3136
world_history
161
0.242236
24.22
0.1801
0.3043
world_religions
160
0.25625
25.62
0.1938
0.3187
agronomy
169
0.242604
24.26
0.1775
0.3077
anatomy
148
0.256757
25.68
0.1892
0.3378
ancient_chinese
164
0.256098
25.61
0.1951
0.3232
arts
160
0.25625
25.62
0.1938
0.3312
astronomy
165
0.248485
24.85
0.1818
0.3152
business_ethics
209
0.248804
24.88
0.1914
0.311
chinese_civil_service_exam
160
0.25625
25.62
0.1938
0.3187
chinese_driving_rule
131
0.251908
25.19
0.1756
0.3282
chinese_food_culture
136
0.25
25
0.1838
0.3309
chinese_foreign_policy
107
0.252336
25.23
0.1682
0.3364
chinese_history
323
0.250774
25.08
0.2043
0.3003
chinese_literature
204
0.254902
25.49
0.1961
0.3137
chinese_teacher_qualification
179
0.251397
25.14
0.1899
0.3184
clinical_knowledge
237
0.253165
25.32
0.1983
0.308
college_actuarial_science
106
0.245283
24.53
0.1792
0.3302
college_education
107
0.317757
31.78
0.2336
0.4019
college_engineering_hydrology
106
0.301887
30.19
0.217
0.3868
college_law
108
0.212963
21.3
0.1481
0.2963
college_mathematics
105
0.219048
21.9
0.1429
0.2952
college_medical_statistics
106
0.254717
25.47
0.1792
0.3396
college_medicine
273
0.241758
24.18
0.1941
0.293
computer_science
204
0.254902
25.49
0.1961
0.3186
computer_security
171
0.251462
25.15
0.1871
0.3158
conceptual_physics
147
0.251701
25.17
0.1837
0.3265
construction_project_management
139
0.244604
24.46
0.1799
0.3165
economics
159
0.245283
24.53
0.1761
0.3145
education
163
0.251534
25.15
0.184
0.319
electrical_engineering
172
0.25
25
0.186
0.314
elementary_chinese
252
0.281746
28.17
0.2262
0.3413
elementary_commonsense
198
0.242424
24.24
0.1869
0.303
elementary_information_and_technology
238
0.273109
27.31
0.2143
0.3319
elementary_mathematics
230
0.278261
27.83
0.2217
0.3348
ethnology
135
0.251852
25.19
0.1778
0.3259
End of preview.

Haidass-Translate-143M Evaluation Assets

Haidass Translate

English | 中文

Evaluation assets for DALabCommunity/Haidass-Translate-143M and 19 comparison models on FLORES-200 dev and FLORES+ devtest (Chinese–English, both directions): per-sentence predictions, score tables for both benchmarks, decontamination audits, and the scripts needed to recompute every reported number.

Evaluation setup

  • Benchmarks: FLORES-200 dev (997 sentences per direction) and FLORES+ devtest (1,012 sentences per direction, zero overlap with dev); language pair eng_Latn ↔ zho_Hans
  • Decoding: greedy (do_sample=False), max_new_tokens=512, bf16
  • Metrics: sacreBLEU corpus BLEU (tokenize=zh for Chinese targets, tokenize=13a for English targets) and chrF++ (word_order=2)
  • Prompt rule: each model is evaluated with prompts byte-identical to its training or official template — chat template for LLMs, forced-BOS for NLLB / M2M / OPUS-MT

Repository structure

Path Contents
ZEROSHOT_RESULTS.md Score tables for both benchmarks (20 models), metric definitions, reading notes, ablation analysis and sample spot checks
error_analysis/LOW_SCORE_CASES_zh.md Human-readable low-score cases and a weakness summary for the current public 143M weights on both benchmarks (Chinese)
error_analysis/sentence_scores.csv Reproducible per-sentence scores, heuristic symptom tags, and src/ref/hyp for the public model's default greedy decoding
audit_report.json n-gram decontamination audit, 8M training data vs FLORES-200 dev (top-50 overlapping samples included)
audit_floresplus_devtest.json Same audit vs FLORES+ devtest
audit_floresplus_devtest_stepfun.json Same audit restricted to the 157,593 general-domain (STEP_FUN) samples used in the mix variant, vs FLORES+ devtest
predictions/ Per-sentence predictions: one pred_<model>_flores_dev.jsonl / pred_<model>_floresplus_devtest.jsonl per model per benchmark
scripts/eval_translate.py LLM evaluation script (chat template; --eval-dir selects the benchmark)
scripts/eval_translate_seq2seq.py Seq2seq evaluation script (NLLB / M2M / OPUS-MT)
scripts/compute_metrics.py Metric recomputation (BLEU / chrF++)
scripts/ngram_audit.py Decontamination audit script

Prediction file format

One JSON object per line:

{"id": 0, "direction": "en-zh", "src": "...", "ref": "...", "hyp": "..."}

direction is en-zh or zh-en; src is the source sentence, ref the reference translation, hyp the model output.

Recomputing the metrics

pip install sacrebleu
python scripts/compute_metrics.py predictions/pred_haidass-8M-trans_floresplus_devtest.jsonl
# == predictions/pred_haidass-8M-trans_floresplus_devtest.jsonl
#   en-zh: n=1012 BLEU=25.06 chrF++=17.72
#   zh-en: n=1012 BLEU=13.68 chrF++=40.02

Summary of findings

  • All 20 model rows were re-evaluated on FLORES+ devtest, which has zero overlap with dev. Both tables and all per-sentence predictions are published here for inspection. The released model's scores move within +1.2 / -1.0 BLEU between the two benchmarks (dev 23.89 / 14.65 to devtest 25.06 / 13.68).
  • Drafter-8M control (same data and recipe at 8M scale, but trained from random initialization): en→zh 12.04 / zh→en 5.47, versus 23.89 / 14.65 when starting from the pretrained base. Without the base, scaling the same parallel SFT data from 1M to 8M adds only +1.2 / +0.3 BLEU — base-model pretraining contributes language knowledge that additional SFT parallel data does not replace.
  • With the pretrained base, en→zh BLEU rises monotonically with data scale: 16.65 → 20.10 → 23.64 → 23.89 (1M / 4M / 8M-mix / 8M translation-only), while zh→en plateaus around 14.4. The 8M-mix-v1 collapse (zh→en 7.34) was traced to uncleaned general-domain data and missing inter-document attention isolation during packing; both were fixed in v2. Full analysis: ZEROSHOT_RESULTS.md.

Decontamination

To check whether the reported scores are inflated by test-set leakage, every training sample was cut into consecutive fragments (8 words for English, 10 characters for Chinese), and any sample sharing at least one fragment with any test sentence was counted as a hit. Of the 15,830,983 training samples, 1,147 (0.0072%) hit FLORES-200 dev and 1,788 (0.0113%) hit FLORES+ devtest; the STEP_FUN-only subset shows 6 hits (0.0038%) against FLORES+ devtest. Manual inspection of the most-overlapping cases shows common-phrase-level overlap rather than full-sentence leakage. Each audit report includes its 50 most-overlapping samples for independent review.

Downloads last month
114