data_source
stringclasses
1 value
prompt
listlengths
2
2
ability
stringclasses
1 value
reward_model
dict
extra_info
dict
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_and_answer_professional_medicine__c4c16f2f__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 840.0115179912483, \"nll_base\": 6.733415603637695, \"eval_ref\": \"mmlusr_question_and_answer_professional_medici...
{ "index": "mmlusr_question_and_answer_professional_medicine__c4c16f2f__s000__tb262144__v000", "uid": "mmlusr_question_and_answer_professional_medicine__c4c16f2f__s000__tb262144__v000", "axis": "knowledge", "budget": 2048, "token_budget": 262144, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_and_answer_professional_psychology__bbc1cebb__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 435.90175998281, \"nll_base\": 6.077416896820068, \"eval_ref\": \"mmlusr_question_and_answer_professional_psycho...
{ "index": "mmlusr_question_and_answer_professional_psychology__bbc1cebb__s000__tb262144__v000", "uid": "mmlusr_question_and_answer_professional_psychology__bbc1cebb__s000__tb262144__v000", "axis": "knowledge", "budget": 2048, "token_budget": 262144, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_and_answer_professional_psychology__bbc1cebb__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 435.90175998281, \"nll_base\": 6.077416896820068, \"eval_ref\": \"mmlusr_question_and_answer_professional_psycholo...
{ "index": "mmlusr_question_and_answer_professional_psychology__bbc1cebb__s000__tb65536__v001", "uid": "mmlusr_question_and_answer_professional_psychology__bbc1cebb__s000__tb65536__v001", "axis": "knowledge", "budget": 2048, "token_budget": 65536, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_and_answer_public_relations__0d5f6563__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 889.0750359190179, \"nll_base\": 6.790181636810303, \"eval_ref\": \"mmlusr_question_and_answer_public_relations__0d5f6563...
{ "index": "mmlusr_question_and_answer_public_relations__0d5f6563__s000__tb65536__v000", "uid": "mmlusr_question_and_answer_public_relations__0d5f6563__s000__tb65536__v000", "axis": "knowledge", "budget": 2048, "token_budget": 65536, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_and_answer_security_studies__2606c39d__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 698.4468746120743, \"nll_base\": 6.548859119415283, \"eval_ref\": \"mmlusr_question_and_answer_security_studies__2606c3...
{ "index": "mmlusr_question_and_answer_security_studies__2606c39d__s000__tb131072__v001", "uid": "mmlusr_question_and_answer_security_studies__2606c39d__s000__tb131072__v001", "axis": "knowledge", "budget": 2048, "token_budget": 131072, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_and_answer_security_studies__2606c39d__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 698.4468746120743, \"nll_base\": 6.548859119415283, \"eval_ref\": \"mmlusr_question_and_answer_security_studies__2606c39d...
{ "index": "mmlusr_question_and_answer_security_studies__2606c39d__s000__tb65536__v000", "uid": "mmlusr_question_and_answer_security_studies__2606c39d__s000__tb65536__v000", "axis": "knowledge", "budget": 2048, "token_budget": 65536, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_and_answer_sociology__39c35e58__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 664.5945912167971, \"nll_base\": 6.4991772174835205, \"eval_ref\": \"mmlusr_question_and_answer_sociology__39c35e58__s000__tb1...
{ "index": "mmlusr_question_and_answer_sociology__39c35e58__s000__tb131072__v000", "uid": "mmlusr_question_and_answer_sociology__39c35e58__s000__tb131072__v000", "axis": "knowledge", "budget": 2048, "token_budget": 131072, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_and_answer_sociology__39c35e58__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 664.5945912167971, \"nll_base\": 6.4991772174835205, \"eval_ref\": \"mmlusr_question_and_answer_sociology__39c35e58__s000__tb2...
{ "index": "mmlusr_question_and_answer_sociology__39c35e58__s000__tb262144__v001", "uid": "mmlusr_question_and_answer_sociology__39c35e58__s000__tb262144__v001", "axis": "knowledge", "budget": 2048, "token_budget": 262144, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_and_answer_virology__88a83282__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 732.3217865511282, \"nll_base\": 6.596220016479492, \"eval_ref\": \"mmlusr_question_and_answer_virology__88a83282__s000__tb1310...
{ "index": "mmlusr_question_and_answer_virology__88a83282__s000__tb131072__v001", "uid": "mmlusr_question_and_answer_virology__88a83282__s000__tb131072__v001", "axis": "knowledge", "budget": 2048, "token_budget": 131072, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_and_answer_virology__88a83282__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 732.3217865511282, \"nll_base\": 6.596220016479492, \"eval_ref\": \"mmlusr_question_and_answer_virology__88a83282__s000__tb65536_...
{ "index": "mmlusr_question_and_answer_virology__88a83282__s000__tb65536__v000", "uid": "mmlusr_question_and_answer_virology__88a83282__s000__tb65536__v000", "axis": "knowledge", "budget": 2048, "token_budget": 65536, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_and_answer_world_religions__7c72f1ed__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1601.9601612749032, \"nll_base\": 7.37898325920105, \"eval_ref\": \"mmlusr_question_and_answer_world_religions__7c72f1ed...
{ "index": "mmlusr_question_and_answer_world_religions__7c72f1ed__s000__tb131072__v000", "uid": "mmlusr_question_and_answer_world_religions__7c72f1ed__s000__tb131072__v000", "axis": "knowledge", "budget": 2048, "token_budget": 131072, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_and_answer_world_religions__7c72f1ed__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1601.9601612749032, \"nll_base\": 7.37898325920105, \"eval_ref\": \"mmlusr_question_and_answer_world_religions__7c72f1ed...
{ "index": "mmlusr_question_and_answer_world_religions__7c72f1ed__s000__tb262144__v001", "uid": "mmlusr_question_and_answer_world_religions__7c72f1ed__s000__tb262144__v001", "axis": "knowledge", "budget": 2048, "token_budget": 262144, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_anatomy__16a98853__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 979.2618730035028, \"nll_base\": 6.886799097061157, \"eval_ref\": \"mmlusr_question_only_anatomy__16a98853__s000__tb131072__v000\", \"...
{ "index": "mmlusr_question_only_anatomy__16a98853__s000__tb131072__v000", "uid": "mmlusr_question_only_anatomy__16a98853__s000__tb131072__v000", "axis": "knowledge", "budget": 2048, "token_budget": 131072, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_astronomy__1f55b766__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1920.4986908070414, \"nll_base\": 7.560340166091919, \"eval_ref\": \"mmlusr_question_only_astronomy__1f55b766__s000__tb262144__v000\...
{ "index": "mmlusr_question_only_astronomy__1f55b766__s000__tb262144__v000", "uid": "mmlusr_question_only_astronomy__1f55b766__s000__tb262144__v000", "axis": "knowledge", "budget": 2048, "token_budget": 262144, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_astronomy__1f55b766__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1920.4986908070414, \"nll_base\": 7.560340166091919, \"eval_ref\": \"mmlusr_question_only_astronomy__1f55b766__s000__tb65536__v001\", ...
{ "index": "mmlusr_question_only_astronomy__1f55b766__s000__tb65536__v001", "uid": "mmlusr_question_only_astronomy__1f55b766__s000__tb65536__v001", "axis": "knowledge", "budget": 2048, "token_budget": 65536, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_business_ethics__3be6b044__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1134.8940814529562, \"nll_base\": 7.034294605255127, \"eval_ref\": \"mmlusr_question_only_business_ethics__3be6b044__s000__tb1...
{ "index": "mmlusr_question_only_business_ethics__3be6b044__s000__tb131072__v001", "uid": "mmlusr_question_only_business_ethics__3be6b044__s000__tb131072__v001", "axis": "knowledge", "budget": 2048, "token_budget": 131072, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_business_ethics__3be6b044__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1134.8940814529562, \"nll_base\": 7.034294605255127, \"eval_ref\": \"mmlusr_question_only_business_ethics__3be6b044__s000__tb655...
{ "index": "mmlusr_question_only_business_ethics__3be6b044__s000__tb65536__v000", "uid": "mmlusr_question_only_business_ethics__3be6b044__s000__tb65536__v000", "axis": "knowledge", "budget": 2048, "token_budget": 65536, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_clinical_knowledge__30330417__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1233.5482259500518, \"nll_base\": 7.117650032043457, \"eval_ref\": \"mmlusr_question_only_clinical_knowledge__30330417__s00...
{ "index": "mmlusr_question_only_clinical_knowledge__30330417__s000__tb131072__v001", "uid": "mmlusr_question_only_clinical_knowledge__30330417__s000__tb131072__v001", "axis": "domain", "budget": 2048, "token_budget": 131072, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_clinical_knowledge__30330417__s000__tb262144__v002\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1233.5482259500518, \"nll_base\": 7.117650032043457, \"eval_ref\": \"mmlusr_question_only_clinical_knowledge__30330417__s00...
{ "index": "mmlusr_question_only_clinical_knowledge__30330417__s000__tb262144__v002", "uid": "mmlusr_question_only_clinical_knowledge__30330417__s000__tb262144__v002", "axis": "domain", "budget": 2048, "token_budget": 262144, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_clinical_knowledge__30330417__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1233.5482259500518, \"nll_base\": 7.117650032043457, \"eval_ref\": \"mmlusr_question_only_clinical_knowledge__30330417__s000_...
{ "index": "mmlusr_question_only_clinical_knowledge__30330417__s000__tb65536__v000", "uid": "mmlusr_question_only_clinical_knowledge__30330417__s000__tb65536__v000", "axis": "domain", "budget": 2048, "token_budget": 65536, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_clinical_knowledge__30330417__s001__tb65536__v003\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 957.1690818109338, \"nll_base\": 6.863980054855347, \"eval_ref\": \"mmlusr_question_only_clinical_knowledge__30330417__s001__...
{ "index": "mmlusr_question_only_clinical_knowledge__30330417__s001__tb65536__v003", "uid": "mmlusr_question_only_clinical_knowledge__30330417__s001__tb65536__v003", "axis": "domain", "budget": 2048, "token_budget": 65536, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_college_biology__9502d37f__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 967.8168988888376, \"nll_base\": 6.875042915344238, \"eval_ref\": \"mmlusr_question_only_college_biology__9502d37f__s000__tb13...
{ "index": "mmlusr_question_only_college_biology__9502d37f__s000__tb131072__v001", "uid": "mmlusr_question_only_college_biology__9502d37f__s000__tb131072__v001", "axis": "knowledge", "budget": 2048, "token_budget": 131072, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_college_biology__9502d37f__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 967.8168988888376, \"nll_base\": 6.875042915344238, \"eval_ref\": \"mmlusr_question_only_college_biology__9502d37f__s000__tb6553...
{ "index": "mmlusr_question_only_college_biology__9502d37f__s000__tb65536__v000", "uid": "mmlusr_question_only_college_biology__9502d37f__s000__tb65536__v000", "axis": "knowledge", "budget": 2048, "token_budget": 65536, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_college_chemistry__2dd29337__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2029.7081961794227, \"nll_base\": 7.615647315979004, \"eval_ref\": \"mmlusr_question_only_college_chemistry__2dd29337__s000_...
{ "index": "mmlusr_question_only_college_chemistry__2dd29337__s000__tb131072__v001", "uid": "mmlusr_question_only_college_chemistry__2dd29337__s000__tb131072__v001", "axis": "knowledge", "budget": 2048, "token_budget": 131072, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_college_chemistry__2dd29337__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2029.7081961794227, \"nll_base\": 7.615647315979004, \"eval_ref\": \"mmlusr_question_only_college_chemistry__2dd29337__s000__t...
{ "index": "mmlusr_question_only_college_chemistry__2dd29337__s000__tb65536__v000", "uid": "mmlusr_question_only_college_chemistry__2dd29337__s000__tb65536__v000", "axis": "knowledge", "budget": 2048, "token_budget": 65536, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_college_computer_science__4422d2c8__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1300.0539061183852, \"nll_base\": 7.170161008834839, \"eval_ref\": \"mmlusr_question_only_college_computer_science__4...
{ "index": "mmlusr_question_only_college_computer_science__4422d2c8__s000__tb262144__v000", "uid": "mmlusr_question_only_college_computer_science__4422d2c8__s000__tb262144__v000", "axis": "knowledge", "budget": 2048, "token_budget": 262144, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_college_computer_science__4422d2c8__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1300.0539061183852, \"nll_base\": 7.170161008834839, \"eval_ref\": \"mmlusr_question_only_college_computer_science__442...
{ "index": "mmlusr_question_only_college_computer_science__4422d2c8__s000__tb65536__v001", "uid": "mmlusr_question_only_college_computer_science__4422d2c8__s000__tb65536__v001", "axis": "knowledge", "budget": 2048, "token_budget": 65536, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_college_medicine__2175e43e__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1229.1841552332385, \"nll_base\": 7.114105939865112, \"eval_ref\": \"mmlusr_question_only_college_medicine__2175e43e__s000__t...
{ "index": "mmlusr_question_only_college_medicine__2175e43e__s000__tb262144__v000", "uid": "mmlusr_question_only_college_medicine__2175e43e__s000__tb262144__v000", "axis": "knowledge", "budget": 2048, "token_budget": 262144, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_computer_security__4f5d9083__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 770.2380839396656, \"nll_base\": 6.646699666976929, \"eval_ref\": \"mmlusr_question_only_computer_security__4f5d9083__s000__...
{ "index": "mmlusr_question_only_computer_security__4f5d9083__s000__tb262144__v000", "uid": "mmlusr_question_only_computer_security__4f5d9083__s000__tb262144__v000", "axis": "knowledge", "budget": 2048, "token_budget": 262144, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_computer_security__4f5d9083__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 770.2380839396656, \"nll_base\": 6.646699666976929, \"eval_ref\": \"mmlusr_question_only_computer_security__4f5d9083__s000__tb...
{ "index": "mmlusr_question_only_computer_security__4f5d9083__s000__tb65536__v001", "uid": "mmlusr_question_only_computer_security__4f5d9083__s000__tb65536__v001", "axis": "knowledge", "budget": 2048, "token_budget": 65536, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_electrical_engineering__8d9168e9__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 744.4105499910507, \"nll_base\": 6.612592697143555, \"eval_ref\": \"mmlusr_question_only_electrical_engineering__8d9168...
{ "index": "mmlusr_question_only_electrical_engineering__8d9168e9__s000__tb131072__v001", "uid": "mmlusr_question_only_electrical_engineering__8d9168e9__s000__tb131072__v001", "axis": "knowledge", "budget": 2048, "token_budget": 131072, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_electrical_engineering__8d9168e9__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 744.4105499910507, \"nll_base\": 6.612592697143555, \"eval_ref\": \"mmlusr_question_only_electrical_engineering__8d9168e9...
{ "index": "mmlusr_question_only_electrical_engineering__8d9168e9__s000__tb65536__v000", "uid": "mmlusr_question_only_electrical_engineering__8d9168e9__s000__tb65536__v000", "axis": "knowledge", "budget": 2048, "token_budget": 65536, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_global_facts__81515419__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1068.5238074642414, \"nll_base\": 6.974033355712891, \"eval_ref\": \"mmlusr_question_only_global_facts__81515419__s000__tb131072_...
{ "index": "mmlusr_question_only_global_facts__81515419__s000__tb131072__v000", "uid": "mmlusr_question_only_global_facts__81515419__s000__tb131072__v000", "axis": "knowledge", "budget": 2048, "token_budget": 131072, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_high_school_chemistry__3a556ae9__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 741.8784807860184, \"nll_base\": 6.609185457229614, \"eval_ref\": \"mmlusr_question_only_high_school_chemistry__3a556ae9__...
{ "index": "mmlusr_question_only_high_school_chemistry__3a556ae9__s000__tb65536__v000", "uid": "mmlusr_question_only_high_school_chemistry__3a556ae9__s000__tb65536__v000", "axis": "knowledge", "budget": 2048, "token_budget": 65536, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_high_school_computer_science__791274d7__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1028.9035685747217, \"nll_base\": 6.936249017715454, \"eval_ref\": \"mmlusr_question_only_high_school_computer_sc...
{ "index": "mmlusr_question_only_high_school_computer_science__791274d7__s000__tb262144__v000", "uid": "mmlusr_question_only_high_school_computer_science__791274d7__s000__tb262144__v000", "axis": "knowledge", "budget": 2048, "token_budget": 262144, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_high_school_computer_science__791274d7__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1028.9035685747217, \"nll_base\": 6.936249017715454, \"eval_ref\": \"mmlusr_question_only_high_school_computer_scie...
{ "index": "mmlusr_question_only_high_school_computer_science__791274d7__s000__tb65536__v001", "uid": "mmlusr_question_only_high_school_computer_science__791274d7__s000__tb65536__v001", "axis": "knowledge", "budget": 2048, "token_budget": 65536, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_high_school_european_history__e1d4a9fa__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 840.7464456237745, \"nll_base\": 6.73429012298584, \"eval_ref\": \"mmlusr_question_only_high_school_european_hist...
{ "index": "mmlusr_question_only_high_school_european_history__e1d4a9fa__s000__tb262144__v000", "uid": "mmlusr_question_only_high_school_european_history__e1d4a9fa__s000__tb262144__v000", "axis": "knowledge", "budget": 2048, "token_budget": 262144, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_high_school_government_and_politics__0abe57d3__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 731.7503720799822, \"nll_base\": 6.595439434051514, \"eval_ref\": \"mmlusr_question_only_high_school_gover...
{ "index": "mmlusr_question_only_high_school_government_and_politics__0abe57d3__s000__tb131072__v000", "uid": "mmlusr_question_only_high_school_government_and_politics__0abe57d3__s000__tb131072__v000", "axis": "knowledge", "budget": 2048, "token_budget": 131072, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_high_school_government_and_politics__0abe57d3__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 731.7503720799822, \"nll_base\": 6.595439434051514, \"eval_ref\": \"mmlusr_question_only_high_school_gover...
{ "index": "mmlusr_question_only_high_school_government_and_politics__0abe57d3__s000__tb262144__v001", "uid": "mmlusr_question_only_high_school_government_and_politics__0abe57d3__s000__tb262144__v001", "axis": "knowledge", "budget": 2048, "token_budget": 262144, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_high_school_macroeconomics__a6f0b420__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 663.9371835816962, \"nll_base\": 6.49818754196167, \"eval_ref\": \"mmlusr_question_only_high_school_macroeconomics_...
{ "index": "mmlusr_question_only_high_school_macroeconomics__a6f0b420__s000__tb262144__v000", "uid": "mmlusr_question_only_high_school_macroeconomics__a6f0b420__s000__tb262144__v000", "axis": "knowledge", "budget": 2048, "token_budget": 262144, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_high_school_macroeconomics__a6f0b420__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 663.9371835816962, \"nll_base\": 6.49818754196167, \"eval_ref\": \"mmlusr_question_only_high_school_macroeconomics__a...
{ "index": "mmlusr_question_only_high_school_macroeconomics__a6f0b420__s000__tb65536__v001", "uid": "mmlusr_question_only_high_school_macroeconomics__a6f0b420__s000__tb65536__v001", "axis": "knowledge", "budget": 2048, "token_budget": 65536, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_high_school_mathematics__3acd5ad5__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1739.358288424737, \"nll_base\": 7.461271524429321, \"eval_ref\": \"mmlusr_question_only_high_school_mathematics__3acd...
{ "index": "mmlusr_question_only_high_school_mathematics__3acd5ad5__s000__tb131072__v000", "uid": "mmlusr_question_only_high_school_mathematics__3acd5ad5__s000__tb131072__v000", "axis": "knowledge", "budget": 2048, "token_budget": 131072, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_high_school_physics__e2ae8183__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1285.289276593697, \"nll_base\": 7.15873908996582, \"eval_ref\": \"mmlusr_question_only_high_school_physics__e2ae8183__s00...
{ "index": "mmlusr_question_only_high_school_physics__e2ae8183__s000__tb131072__v000", "uid": "mmlusr_question_only_high_school_physics__e2ae8183__s000__tb131072__v000", "axis": "knowledge", "budget": 2048, "token_budget": 131072, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_high_school_physics__e2ae8183__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1285.289276593697, \"nll_base\": 7.15873908996582, \"eval_ref\": \"mmlusr_question_only_high_school_physics__e2ae8183__s00...
{ "index": "mmlusr_question_only_high_school_physics__e2ae8183__s000__tb262144__v001", "uid": "mmlusr_question_only_high_school_physics__e2ae8183__s000__tb262144__v001", "axis": "knowledge", "budget": 2048, "token_budget": 262144, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_high_school_psychology__7e3cfd48__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1377.6883827524666, \"nll_base\": 7.2281622886657715, \"eval_ref\": \"mmlusr_question_only_high_school_psychology__7e3c...
{ "index": "mmlusr_question_only_high_school_psychology__7e3cfd48__s000__tb262144__v000", "uid": "mmlusr_question_only_high_school_psychology__7e3cfd48__s000__tb262144__v000", "axis": "knowledge", "budget": 2048, "token_budget": 262144, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_high_school_statistics__c33e08ce__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 991.2925569024414, \"nll_base\": 6.899009704589844, \"eval_ref\": \"mmlusr_question_only_high_school_statistics__c33e08ce...
{ "index": "mmlusr_question_only_high_school_statistics__c33e08ce__s000__tb65536__v000", "uid": "mmlusr_question_only_high_school_statistics__c33e08ce__s000__tb65536__v000", "axis": "knowledge", "budget": 2048, "token_budget": 65536, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_human_aging__3e35e990__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 995.239277979423, \"nll_base\": 6.90298318862915, \"eval_ref\": \"mmlusr_question_only_human_aging__3e35e990__s000__tb131072__v000...
{ "index": "mmlusr_question_only_human_aging__3e35e990__s000__tb131072__v000", "uid": "mmlusr_question_only_human_aging__3e35e990__s000__tb131072__v000", "axis": "knowledge", "budget": 2048, "token_budget": 131072, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_human_sexuality__b76e9d47__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1304.3504854408188, \"nll_base\": 7.173460483551025, \"eval_ref\": \"mmlusr_question_only_human_sexuality__b76e9d47__s000__tb1...
{ "index": "mmlusr_question_only_human_sexuality__b76e9d47__s000__tb131072__v000", "uid": "mmlusr_question_only_human_sexuality__b76e9d47__s000__tb131072__v000", "axis": "knowledge", "budget": 2048, "token_budget": 131072, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_human_sexuality__b76e9d47__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1304.3504854408188, \"nll_base\": 7.173460483551025, \"eval_ref\": \"mmlusr_question_only_human_sexuality__b76e9d47__s000__tb2...
{ "index": "mmlusr_question_only_human_sexuality__b76e9d47__s000__tb262144__v001", "uid": "mmlusr_question_only_human_sexuality__b76e9d47__s000__tb262144__v001", "axis": "knowledge", "budget": 2048, "token_budget": 262144, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_international_law__5926aa41__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 538.6280759699356, \"nll_base\": 6.28902530670166, \"eval_ref\": \"mmlusr_question_only_international_law__5926aa41__s000__t...
{ "index": "mmlusr_question_only_international_law__5926aa41__s000__tb131072__v001", "uid": "mmlusr_question_only_international_law__5926aa41__s000__tb131072__v001", "axis": "knowledge", "budget": 2048, "token_budget": 131072, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_international_law__5926aa41__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 538.6280759699356, \"nll_base\": 6.28902530670166, \"eval_ref\": \"mmlusr_question_only_international_law__5926aa41__s000__tb6...
{ "index": "mmlusr_question_only_international_law__5926aa41__s000__tb65536__v000", "uid": "mmlusr_question_only_international_law__5926aa41__s000__tb65536__v000", "axis": "knowledge", "budget": 2048, "token_budget": 65536, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_jurisprudence__752bbc38__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1577.946255512085, \"nll_base\": 7.363879442214966, \"eval_ref\": \"mmlusr_question_only_jurisprudence__752bbc38__s000__tb131072...
{ "index": "mmlusr_question_only_jurisprudence__752bbc38__s000__tb131072__v000", "uid": "mmlusr_question_only_jurisprudence__752bbc38__s000__tb131072__v000", "axis": "knowledge", "budget": 2048, "token_budget": 131072, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_logical_fallacies__39427e3d__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1216.5935642742975, \"nll_base\": 7.10381007194519, \"eval_ref\": \"mmlusr_question_only_logical_fallacies__39427e3d__s000__...
{ "index": "mmlusr_question_only_logical_fallacies__39427e3d__s000__tb131072__v000", "uid": "mmlusr_question_only_logical_fallacies__39427e3d__s000__tb131072__v000", "axis": "knowledge", "budget": 2048, "token_budget": 131072, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_logical_fallacies__39427e3d__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1216.5935642742975, \"nll_base\": 7.10381007194519, \"eval_ref\": \"mmlusr_question_only_logical_fallacies__39427e3d__s000__...
{ "index": "mmlusr_question_only_logical_fallacies__39427e3d__s000__tb262144__v001", "uid": "mmlusr_question_only_logical_fallacies__39427e3d__s000__tb262144__v001", "axis": "knowledge", "budget": 2048, "token_budget": 262144, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_machine_learning__be90dc88__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 895.81060774761, \"nll_base\": 6.797729015350342, \"eval_ref\": \"mmlusr_question_only_machine_learning__be90dc88__s000__tb13...
{ "index": "mmlusr_question_only_machine_learning__be90dc88__s000__tb131072__v000", "uid": "mmlusr_question_only_machine_learning__be90dc88__s000__tb131072__v000", "axis": "knowledge", "budget": 2048, "token_budget": 131072, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_management__ab30f009__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2376.357782900255, \"nll_base\": 7.773324251174927, \"eval_ref\": \"mmlusr_question_only_management__ab30f009__s000__tb131072__v001...
{ "index": "mmlusr_question_only_management__ab30f009__s000__tb131072__v001", "uid": "mmlusr_question_only_management__ab30f009__s000__tb131072__v001", "axis": "knowledge", "budget": 2048, "token_budget": 131072, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_management__ab30f009__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2376.357782900255, \"nll_base\": 7.773324251174927, \"eval_ref\": \"mmlusr_question_only_management__ab30f009__s000__tb65536__v000\",...
{ "index": "mmlusr_question_only_management__ab30f009__s000__tb65536__v000", "uid": "mmlusr_question_only_management__ab30f009__s000__tb65536__v000", "axis": "knowledge", "budget": 2048, "token_budget": 65536, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_marketing__9ecbebf9__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 997.6943892563969, \"nll_base\": 6.905447006225586, \"eval_ref\": \"mmlusr_question_only_marketing__9ecbebf9__s000__tb131072__v001\"...
{ "index": "mmlusr_question_only_marketing__9ecbebf9__s000__tb131072__v001", "uid": "mmlusr_question_only_marketing__9ecbebf9__s000__tb131072__v001", "axis": "knowledge", "budget": 2048, "token_budget": 131072, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_marketing__9ecbebf9__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 997.6943892563969, \"nll_base\": 6.905447006225586, \"eval_ref\": \"mmlusr_question_only_marketing__9ecbebf9__s000__tb65536__v000\", \...
{ "index": "mmlusr_question_only_marketing__9ecbebf9__s000__tb65536__v000", "uid": "mmlusr_question_only_marketing__9ecbebf9__s000__tb65536__v000", "axis": "knowledge", "budget": 2048, "token_budget": 65536, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_medical_genetics__7490dcf4__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 928.1056919038201, \"nll_base\": 6.833145618438721, \"eval_ref\": \"mmlusr_question_only_medical_genetics__7490dcf4__s000__tb...
{ "index": "mmlusr_question_only_medical_genetics__7490dcf4__s000__tb131072__v001", "uid": "mmlusr_question_only_medical_genetics__7490dcf4__s000__tb131072__v001", "axis": "domain", "budget": 2048, "token_budget": 131072, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_medical_genetics__7490dcf4__s000__tb262144__v002\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 928.1056919038201, \"nll_base\": 6.833145618438721, \"eval_ref\": \"mmlusr_question_only_medical_genetics__7490dcf4__s000__tb...
{ "index": "mmlusr_question_only_medical_genetics__7490dcf4__s000__tb262144__v002", "uid": "mmlusr_question_only_medical_genetics__7490dcf4__s000__tb262144__v002", "axis": "domain", "budget": 2048, "token_budget": 262144, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_medical_genetics__7490dcf4__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 928.1056919038201, \"nll_base\": 6.833145618438721, \"eval_ref\": \"mmlusr_question_only_medical_genetics__7490dcf4__s000__tb65...
{ "index": "mmlusr_question_only_medical_genetics__7490dcf4__s000__tb65536__v000", "uid": "mmlusr_question_only_medical_genetics__7490dcf4__s000__tb65536__v000", "axis": "domain", "budget": 2048, "token_budget": 65536, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_medical_genetics__7490dcf4__s001__tb65536__v003\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1036.5845335612828, \"nll_base\": 6.943686485290527, \"eval_ref\": \"mmlusr_question_only_medical_genetics__7490dcf4__s001__tb6...
{ "index": "mmlusr_question_only_medical_genetics__7490dcf4__s001__tb65536__v003", "uid": "mmlusr_question_only_medical_genetics__7490dcf4__s001__tb65536__v003", "axis": "domain", "budget": 2048, "token_budget": 65536, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_miscellaneous__11a91f57__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1748.2483607056838, \"nll_base\": 7.46636962890625, \"eval_ref\": \"mmlusr_question_only_miscellaneous__11a91f57__s000__tb131072...
{ "index": "mmlusr_question_only_miscellaneous__11a91f57__s000__tb131072__v000", "uid": "mmlusr_question_only_miscellaneous__11a91f57__s000__tb131072__v000", "axis": "knowledge", "budget": 2048, "token_budget": 131072, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_miscellaneous__11a91f57__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1748.2483607056838, \"nll_base\": 7.46636962890625, \"eval_ref\": \"mmlusr_question_only_miscellaneous__11a91f57__s000__tb262144...
{ "index": "mmlusr_question_only_miscellaneous__11a91f57__s000__tb262144__v001", "uid": "mmlusr_question_only_miscellaneous__11a91f57__s000__tb262144__v001", "axis": "knowledge", "budget": 2048, "token_budget": 262144, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_moral_disputes__2646a26a__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 954.8895035004849, \"nll_base\": 6.861595630645752, \"eval_ref\": \"mmlusr_question_only_moral_disputes__2646a26a__s000__tb1310...
{ "index": "mmlusr_question_only_moral_disputes__2646a26a__s000__tb131072__v000", "uid": "mmlusr_question_only_moral_disputes__2646a26a__s000__tb131072__v000", "axis": "knowledge", "budget": 2048, "token_budget": 131072, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_moral_disputes__2646a26a__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 954.8895035004849, \"nll_base\": 6.861595630645752, \"eval_ref\": \"mmlusr_question_only_moral_disputes__2646a26a__s000__tb2621...
{ "index": "mmlusr_question_only_moral_disputes__2646a26a__s000__tb262144__v001", "uid": "mmlusr_question_only_moral_disputes__2646a26a__s000__tb262144__v001", "axis": "knowledge", "budget": 2048, "token_budget": 262144, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_moral_scenarios__c707f7e1__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2334.275132201741, \"nll_base\": 7.7554566860198975, \"eval_ref\": \"mmlusr_question_only_moral_scenarios__c707f7e1__s000__tb1...
{ "index": "mmlusr_question_only_moral_scenarios__c707f7e1__s000__tb131072__v000", "uid": "mmlusr_question_only_moral_scenarios__c707f7e1__s000__tb131072__v000", "axis": "knowledge", "budget": 2048, "token_budget": 131072, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_moral_scenarios__c707f7e1__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2334.275132201741, \"nll_base\": 7.7554566860198975, \"eval_ref\": \"mmlusr_question_only_moral_scenarios__c707f7e1__s000__tb2...
{ "index": "mmlusr_question_only_moral_scenarios__c707f7e1__s000__tb262144__v001", "uid": "mmlusr_question_only_moral_scenarios__c707f7e1__s000__tb262144__v001", "axis": "knowledge", "budget": 2048, "token_budget": 262144, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_nutrition__c98869bb__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1120.305374178168, \"nll_base\": 7.021356582641602, \"eval_ref\": \"mmlusr_question_only_nutrition__c98869bb__s000__tb262144__v000\"...
{ "index": "mmlusr_question_only_nutrition__c98869bb__s000__tb262144__v000", "uid": "mmlusr_question_only_nutrition__c98869bb__s000__tb262144__v000", "axis": "knowledge", "budget": 2048, "token_budget": 262144, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_nutrition__c98869bb__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1120.305374178168, \"nll_base\": 7.021356582641602, \"eval_ref\": \"mmlusr_question_only_nutrition__c98869bb__s000__tb65536__v001\", \...
{ "index": "mmlusr_question_only_nutrition__c98869bb__s000__tb65536__v001", "uid": "mmlusr_question_only_nutrition__c98869bb__s000__tb65536__v001", "axis": "knowledge", "budget": 2048, "token_budget": 65536, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_philosophy__88a2bac7__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1463.8214448610156, \"nll_base\": 7.288805723190308, \"eval_ref\": \"mmlusr_question_only_philosophy__88a2bac7__s000__tb131072__v00...
{ "index": "mmlusr_question_only_philosophy__88a2bac7__s000__tb131072__v001", "uid": "mmlusr_question_only_philosophy__88a2bac7__s000__tb131072__v001", "axis": "knowledge", "budget": 2048, "token_budget": 131072, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_philosophy__88a2bac7__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1463.8214448610156, \"nll_base\": 7.288805723190308, \"eval_ref\": \"mmlusr_question_only_philosophy__88a2bac7__s000__tb65536__v000\"...
{ "index": "mmlusr_question_only_philosophy__88a2bac7__s000__tb65536__v000", "uid": "mmlusr_question_only_philosophy__88a2bac7__s000__tb65536__v000", "axis": "knowledge", "budget": 2048, "token_budget": 65536, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_prehistory__70a08f2f__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 707.3666892724424, \"nll_base\": 6.561549186706543, \"eval_ref\": \"mmlusr_question_only_prehistory__70a08f2f__s000__tb65536__v000\",...
{ "index": "mmlusr_question_only_prehistory__70a08f2f__s000__tb65536__v000", "uid": "mmlusr_question_only_prehistory__70a08f2f__s000__tb65536__v000", "axis": "knowledge", "budget": 2048, "token_budget": 65536, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_professional_accounting__6b531f4c__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 962.1023483838236, \"nll_base\": 6.869120836257935, \"eval_ref\": \"mmlusr_question_only_professional_accounting__6b53...
{ "index": "mmlusr_question_only_professional_accounting__6b531f4c__s000__tb131072__v001", "uid": "mmlusr_question_only_professional_accounting__6b531f4c__s000__tb131072__v001", "axis": "knowledge", "budget": 2048, "token_budget": 131072, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_professional_accounting__6b531f4c__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 962.1023483838236, \"nll_base\": 6.869120836257935, \"eval_ref\": \"mmlusr_question_only_professional_accounting__6b531f...
{ "index": "mmlusr_question_only_professional_accounting__6b531f4c__s000__tb65536__v000", "uid": "mmlusr_question_only_professional_accounting__6b531f4c__s000__tb65536__v000", "axis": "knowledge", "budget": 2048, "token_budget": 65536, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_professional_medicine__dfd580c6__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1357.9769249947472, \"nll_base\": 7.213751316070557, \"eval_ref\": \"mmlusr_question_only_professional_medicine__dfd580c...
{ "index": "mmlusr_question_only_professional_medicine__dfd580c6__s000__tb131072__v000", "uid": "mmlusr_question_only_professional_medicine__dfd580c6__s000__tb131072__v000", "axis": "knowledge", "budget": 2048, "token_budget": 131072, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_professional_psychology__eab791b2__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 765.8676167736543, \"nll_base\": 6.641009330749512, \"eval_ref\": \"mmlusr_question_only_professional_psychology__eab7...
{ "index": "mmlusr_question_only_professional_psychology__eab791b2__s000__tb131072__v000", "uid": "mmlusr_question_only_professional_psychology__eab791b2__s000__tb131072__v000", "axis": "knowledge", "budget": 2048, "token_budget": 131072, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_public_relations__887cb989__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1730.0171528791766, \"nll_base\": 7.455886602401733, \"eval_ref\": \"mmlusr_question_only_public_relations__887cb989__s000__t...
{ "index": "mmlusr_question_only_public_relations__887cb989__s000__tb131072__v000", "uid": "mmlusr_question_only_public_relations__887cb989__s000__tb131072__v000", "axis": "knowledge", "budget": 2048, "token_budget": 131072, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_public_relations__887cb989__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1730.0171528791766, \"nll_base\": 7.455886602401733, \"eval_ref\": \"mmlusr_question_only_public_relations__887cb989__s000__t...
{ "index": "mmlusr_question_only_public_relations__887cb989__s000__tb262144__v001", "uid": "mmlusr_question_only_public_relations__887cb989__s000__tb262144__v001", "axis": "knowledge", "budget": 2048, "token_budget": 262144, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_security_studies__8e4e99c0__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1131.7243441833396, \"nll_base\": 7.0314977169036865, \"eval_ref\": \"mmlusr_question_only_security_studies__8e4e99c0__s000__...
{ "index": "mmlusr_question_only_security_studies__8e4e99c0__s000__tb131072__v000", "uid": "mmlusr_question_only_security_studies__8e4e99c0__s000__tb131072__v000", "axis": "knowledge", "budget": 2048, "token_budget": 131072, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_security_studies__8e4e99c0__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1131.7243441833396, \"nll_base\": 7.0314977169036865, \"eval_ref\": \"mmlusr_question_only_security_studies__8e4e99c0__s000__...
{ "index": "mmlusr_question_only_security_studies__8e4e99c0__s000__tb262144__v001", "uid": "mmlusr_question_only_security_studies__8e4e99c0__s000__tb262144__v001", "axis": "knowledge", "budget": 2048, "token_budget": 262144, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_virology__4c8fa92b__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1123.1324664513925, \"nll_base\": 7.023876905441284, \"eval_ref\": \"mmlusr_question_only_virology__4c8fa92b__s000__tb131072__v001\",...
{ "index": "mmlusr_question_only_virology__4c8fa92b__s000__tb131072__v001", "uid": "mmlusr_question_only_virology__4c8fa92b__s000__tb131072__v001", "axis": "knowledge", "budget": 2048, "token_budget": 131072, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_virology__4c8fa92b__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1123.1324664513925, \"nll_base\": 7.023876905441284, \"eval_ref\": \"mmlusr_question_only_virology__4c8fa92b__s000__tb65536__v000\", \"...
{ "index": "mmlusr_question_only_virology__4c8fa92b__s000__tb65536__v000", "uid": "mmlusr_question_only_virology__4c8fa92b__s000__tb65536__v000", "axis": "knowledge", "budget": 2048, "token_budget": 65536, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmlusr_question_only_world_religions__75de890c__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1660.158834203031, \"nll_base\": 7.414668560028076, \"eval_ref\": \"mmlusr_question_only_world_religions__75de890c__s000__tb26...
{ "index": "mmlusr_question_only_world_religions__75de890c__s000__tb262144__v000", "uid": "mmlusr_question_only_world_religions__75de890c__s000__tb262144__v000", "axis": "knowledge", "budget": 2048, "token_budget": 262144, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmmlu_ar_xy_clinical_knowledge__16bb946c__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 516.9446309383651, \"nll_base\": 6.247935771942139, \"eval_ref\": \"mmmlu_ar_xy_clinical_knowledge__16bb946c__s000__tb65536__v001\", \...
{ "index": "mmmlu_ar_xy_clinical_knowledge__16bb946c__s000__tb65536__v001", "uid": "mmmlu_ar_xy_clinical_knowledge__16bb946c__s000__tb65536__v001", "axis": "domain", "budget": 2048, "token_budget": 65536, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmmlu_ar_xy_medical_genetics__9ba14068__s000__tb65536__v002\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 749.7226160444661, \"nll_base\": 6.61970329284668, \"eval_ref\": \"mmmlu_ar_xy_medical_genetics__9ba14068__s000__tb65536__v002\", \"orac...
{ "index": "mmmlu_ar_xy_medical_genetics__9ba14068__s000__tb65536__v002", "uid": "mmmlu_ar_xy_medical_genetics__9ba14068__s000__tb65536__v002", "axis": "domain", "budget": 2048, "token_budget": 65536, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmmlu_ar_xy_nutrition__f5803a3a__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 452.9223730604147, \"nll_base\": 6.115720748901367, \"eval_ref\": \"mmmlu_ar_xy_nutrition__f5803a3a__s000__tb65536__v000\", \"oracle_delta\": 0...
{ "index": "mmmlu_ar_xy_nutrition__f5803a3a__s000__tb65536__v000", "uid": "mmmlu_ar_xy_nutrition__f5803a3a__s000__tb65536__v000", "axis": "multilingual", "budget": 2048, "token_budget": 65536, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmmlu_bn_bd_clinical_knowledge__52b42830__s000__tb131072__v002\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 613.2245687293237, \"nll_base\": 6.418731212615967, \"eval_ref\": \"mmmlu_bn_bd_clinical_knowledge__52b42830__s000__tb131072__v002\"...
{ "index": "mmmlu_bn_bd_clinical_knowledge__52b42830__s000__tb131072__v002", "uid": "mmmlu_bn_bd_clinical_knowledge__52b42830__s000__tb131072__v002", "axis": "domain", "budget": 2048, "token_budget": 131072, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmmlu_bn_bd_clinical_knowledge__52b42830__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 613.2245687293237, \"nll_base\": 6.418731212615967, \"eval_ref\": \"mmmlu_bn_bd_clinical_knowledge__52b42830__s000__tb65536__v001\", \...
{ "index": "mmmlu_bn_bd_clinical_knowledge__52b42830__s000__tb65536__v001", "uid": "mmmlu_bn_bd_clinical_knowledge__52b42830__s000__tb65536__v001", "axis": "domain", "budget": 2048, "token_budget": 65536, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmmlu_bn_bd_college_physics__a856c191__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1307.0220599578229, \"nll_base\": 7.175506591796875, \"eval_ref\": \"mmmlu_bn_bd_college_physics__a856c191__s000__tb65536__v000\", \"orac...
{ "index": "mmmlu_bn_bd_college_physics__a856c191__s000__tb65536__v000", "uid": "mmmlu_bn_bd_college_physics__a856c191__s000__tb65536__v000", "axis": "multilingual", "budget": 2048, "token_budget": 65536, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmmlu_bn_bd_electrical_engineering__3727e6a5__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 620.2288482714811, \"nll_base\": 6.430088520050049, \"eval_ref\": \"mmmlu_bn_bd_electrical_engineering__3727e6a5__s000__tb65536__v...
{ "index": "mmmlu_bn_bd_electrical_engineering__3727e6a5__s000__tb65536__v000", "uid": "mmmlu_bn_bd_electrical_engineering__3727e6a5__s000__tb65536__v000", "axis": "multilingual", "budget": 2048, "token_budget": 65536, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmmlu_bn_bd_elementary_mathematics__4f3a3116__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1119.7795745242813, \"nll_base\": 7.020887136459351, \"eval_ref\": \"mmmlu_bn_bd_elementary_mathematics__4f3a3116__s000__tb65536__...
{ "index": "mmmlu_bn_bd_elementary_mathematics__4f3a3116__s000__tb65536__v000", "uid": "mmmlu_bn_bd_elementary_mathematics__4f3a3116__s000__tb65536__v000", "axis": "multilingual", "budget": 2048, "token_budget": 65536, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmmlu_bn_bd_formal_logic__34b43964__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 427.72450528128354, \"nll_base\": 6.058479309082031, \"eval_ref\": \"mmmlu_bn_bd_formal_logic__34b43964__s000__tb65536__v000\", \"oracle_del...
{ "index": "mmmlu_bn_bd_formal_logic__34b43964__s000__tb65536__v000", "uid": "mmmlu_bn_bd_formal_logic__34b43964__s000__tb65536__v000", "axis": "multilingual", "budget": 2048, "token_budget": 65536, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmmlu_bn_bd_high_school_government_and_politics__fe476c43__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 581.211898699799, \"nll_base\": 6.365115404129028, \"eval_ref\": \"mmmlu_bn_bd_high_school_government_and_politics_...
{ "index": "mmmlu_bn_bd_high_school_government_and_politics__fe476c43__s000__tb131072__v000", "uid": "mmmlu_bn_bd_high_school_government_and_politics__fe476c43__s000__tb131072__v000", "axis": "multilingual", "budget": 2048, "token_budget": 131072, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmmlu_bn_bd_human_sexuality__15ab5900__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 813.8000964654364, \"nll_base\": 6.701714754104614, \"eval_ref\": \"mmmlu_bn_bd_human_sexuality__15ab5900__s000__tb131072__v000\", \"or...
{ "index": "mmmlu_bn_bd_human_sexuality__15ab5900__s000__tb131072__v000", "uid": "mmmlu_bn_bd_human_sexuality__15ab5900__s000__tb131072__v000", "axis": "multilingual", "budget": 2048, "token_budget": 131072, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmmlu_bn_bd_medical_genetics__9a5f0cd1__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 848.2664687914222, \"nll_base\": 6.743194818496704, \"eval_ref\": \"mmmlu_bn_bd_medical_genetics__9a5f0cd1__s000__tb131072__v000\", \"...
{ "index": "mmmlu_bn_bd_medical_genetics__9a5f0cd1__s000__tb131072__v000", "uid": "mmmlu_bn_bd_medical_genetics__9a5f0cd1__s000__tb131072__v000", "axis": "domain", "budget": 2048, "token_budget": 131072, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmmlu_bn_bd_medical_genetics__9a5f0cd1__s001__tb131072__v003\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 791.9324887453324, \"nll_base\": 6.674476146697998, \"eval_ref\": \"mmmlu_bn_bd_medical_genetics__9a5f0cd1__s001__tb131072__v003\", \"...
{ "index": "mmmlu_bn_bd_medical_genetics__9a5f0cd1__s001__tb131072__v003", "uid": "mmmlu_bn_bd_medical_genetics__9a5f0cd1__s001__tb131072__v003", "axis": "domain", "budget": 2048, "token_budget": 131072, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmmlu_bn_bd_professional_accounting__3a0d8a9c__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 728.7998883947176, \"nll_base\": 6.591399192810059, \"eval_ref\": \"mmmlu_bn_bd_professional_accounting__3a0d8a9c__s000__tb65536_...
{ "index": "mmmlu_bn_bd_professional_accounting__3a0d8a9c__s000__tb65536__v000", "uid": "mmmlu_bn_bd_professional_accounting__3a0d8a9c__s000__tb65536__v000", "axis": "multilingual", "budget": 2048, "token_budget": 65536, "variant": "filter" }
datacuration
[ { "role": "system", "content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e...
datacuration_selection
{ "style": "datacuration", "ground_truth": "{\"task_id\": \"mmmlu_de_de_abstract_algebra__cd2d98fc__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 898.8110513631686, \"nll_base\": 6.801072835922241, \"eval_ref\": \"mmmlu_de_de_abstract_algebra__cd2d98fc__s000__tb131072__v000\", \"...
{ "index": "mmmlu_de_de_abstract_algebra__cd2d98fc__s000__tb131072__v000", "uid": "mmmlu_de_de_abstract_algebra__cd2d98fc__s000__tb131072__v000", "axis": "multilingual", "budget": 2048, "token_budget": 131072, "variant": "filter" }