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_answer_only_high_school_computer_science__ec2c76a5__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 532.0118095837913, \"nll_base\": 6.276665687561035, \"eval_ref\": \"mmlusr_answer_only_high_school_computer_science...
{ "index": "mmlusr_answer_only_high_school_computer_science__ec2c76a5__s000__tb131072__v001", "uid": "mmlusr_answer_only_high_school_computer_science__ec2c76a5__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_answer_only_high_school_computer_science__ec2c76a5__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 532.0118095837913, \"nll_base\": 6.276665687561035, \"eval_ref\": \"mmlusr_answer_only_high_school_computer_science__...
{ "index": "mmlusr_answer_only_high_school_computer_science__ec2c76a5__s000__tb65536__v000", "uid": "mmlusr_answer_only_high_school_computer_science__ec2c76a5__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_answer_only_high_school_european_history__5c00b799__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 884.1999523831076, \"nll_base\": 6.7846832275390625, \"eval_ref\": \"mmlusr_answer_only_high_school_european_histor...
{ "index": "mmlusr_answer_only_high_school_european_history__5c00b799__s000__tb131072__v000", "uid": "mmlusr_answer_only_high_school_european_history__5c00b799__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_answer_only_high_school_european_history__5c00b799__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 884.1999523831076, \"nll_base\": 6.7846832275390625, \"eval_ref\": \"mmlusr_answer_only_high_school_european_histor...
{ "index": "mmlusr_answer_only_high_school_european_history__5c00b799__s000__tb262144__v001", "uid": "mmlusr_answer_only_high_school_european_history__5c00b799__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_answer_only_high_school_geography__fb460eb7__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 948.1739375632781, \"nll_base\": 6.8545379638671875, \"eval_ref\": \"mmlusr_answer_only_high_school_geography__fb460eb7__s...
{ "index": "mmlusr_answer_only_high_school_geography__fb460eb7__s000__tb131072__v000", "uid": "mmlusr_answer_only_high_school_geography__fb460eb7__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_answer_only_high_school_government_and_politics__df16c26b__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 531.4456369235141, \"nll_base\": 6.275600910186768, \"eval_ref\": \"mmlusr_answer_only_high_school_governmen...
{ "index": "mmlusr_answer_only_high_school_government_and_politics__df16c26b__s000__tb262144__v000", "uid": "mmlusr_answer_only_high_school_government_and_politics__df16c26b__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_answer_only_high_school_government_and_politics__df16c26b__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 531.4456369235141, \"nll_base\": 6.275600910186768, \"eval_ref\": \"mmlusr_answer_only_high_school_government_...
{ "index": "mmlusr_answer_only_high_school_government_and_politics__df16c26b__s000__tb65536__v001", "uid": "mmlusr_answer_only_high_school_government_and_politics__df16c26b__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_answer_only_high_school_macroeconomics__63334df8__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 487.3995858403235, \"nll_base\": 6.18908429145813, \"eval_ref\": \"mmlusr_answer_only_high_school_macroeconomics__633...
{ "index": "mmlusr_answer_only_high_school_macroeconomics__63334df8__s000__tb131072__v000", "uid": "mmlusr_answer_only_high_school_macroeconomics__63334df8__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_answer_only_high_school_macroeconomics__63334df8__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 487.3995858403235, \"nll_base\": 6.18908429145813, \"eval_ref\": \"mmlusr_answer_only_high_school_macroeconomics__633...
{ "index": "mmlusr_answer_only_high_school_macroeconomics__63334df8__s000__tb262144__v001", "uid": "mmlusr_answer_only_high_school_macroeconomics__63334df8__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_answer_only_high_school_microeconomics__21039ab3__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 569.5548382157864, \"nll_base\": 6.344855070114136, \"eval_ref\": \"mmlusr_answer_only_high_school_microeconomics__21...
{ "index": "mmlusr_answer_only_high_school_microeconomics__21039ab3__s000__tb131072__v001", "uid": "mmlusr_answer_only_high_school_microeconomics__21039ab3__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_answer_only_high_school_microeconomics__21039ab3__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 569.5548382157864, \"nll_base\": 6.344855070114136, \"eval_ref\": \"mmlusr_answer_only_high_school_microeconomics__2103...
{ "index": "mmlusr_answer_only_high_school_microeconomics__21039ab3__s000__tb65536__v000", "uid": "mmlusr_answer_only_high_school_microeconomics__21039ab3__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_answer_only_high_school_physics__67d6f152__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 416.4441454058216, \"nll_base\": 6.031752347946167, \"eval_ref\": \"mmlusr_answer_only_high_school_physics__67d6f152__s000__...
{ "index": "mmlusr_answer_only_high_school_physics__67d6f152__s000__tb131072__v000", "uid": "mmlusr_answer_only_high_school_physics__67d6f152__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_answer_only_high_school_psychology__c42e7f4b__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 641.7887165930335, \"nll_base\": 6.464259147644043, \"eval_ref\": \"mmlusr_answer_only_high_school_psychology__c42e7f4b__...
{ "index": "mmlusr_answer_only_high_school_psychology__c42e7f4b__s000__tb262144__v000", "uid": "mmlusr_answer_only_high_school_psychology__c42e7f4b__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_answer_only_high_school_psychology__c42e7f4b__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 641.7887165930335, \"nll_base\": 6.464259147644043, \"eval_ref\": \"mmlusr_answer_only_high_school_psychology__c42e7f4b__s0...
{ "index": "mmlusr_answer_only_high_school_psychology__c42e7f4b__s000__tb65536__v001", "uid": "mmlusr_answer_only_high_school_psychology__c42e7f4b__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_answer_only_high_school_statistics__a1d2e986__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 622.8054198431913, \"nll_base\": 6.434234142303467, \"eval_ref\": \"mmlusr_answer_only_high_school_statistics__a1d2e986__...
{ "index": "mmlusr_answer_only_high_school_statistics__a1d2e986__s000__tb131072__v000", "uid": "mmlusr_answer_only_high_school_statistics__a1d2e986__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_answer_only_high_school_us_history__de1d06ea__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 501.01578691117584, \"nll_base\": 6.21663761138916, \"eval_ref\": \"mmlusr_answer_only_high_school_us_history__de1d06ea__...
{ "index": "mmlusr_answer_only_high_school_us_history__de1d06ea__s000__tb131072__v000", "uid": "mmlusr_answer_only_high_school_us_history__de1d06ea__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_answer_only_high_school_us_history__de1d06ea__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 501.01578691117584, \"nll_base\": 6.21663761138916, \"eval_ref\": \"mmlusr_answer_only_high_school_us_history__de1d06ea__...
{ "index": "mmlusr_answer_only_high_school_us_history__de1d06ea__s000__tb262144__v001", "uid": "mmlusr_answer_only_high_school_us_history__de1d06ea__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_answer_only_high_school_world_history__88849057__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 669.504358945459, \"nll_base\": 6.506537675857544, \"eval_ref\": \"mmlusr_answer_only_high_school_world_history__88849...
{ "index": "mmlusr_answer_only_high_school_world_history__88849057__s000__tb131072__v001", "uid": "mmlusr_answer_only_high_school_world_history__88849057__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_answer_only_high_school_world_history__88849057__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 669.504358945459, \"nll_base\": 6.506537675857544, \"eval_ref\": \"mmlusr_answer_only_high_school_world_history__8884905...
{ "index": "mmlusr_answer_only_high_school_world_history__88849057__s000__tb65536__v000", "uid": "mmlusr_answer_only_high_school_world_history__88849057__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_answer_only_human_aging__e2a52482__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 969.8604590635451, \"nll_base\": 6.87715220451355, \"eval_ref\": \"mmlusr_answer_only_human_aging__e2a52482__s000__tb131072__v001\",...
{ "index": "mmlusr_answer_only_human_aging__e2a52482__s000__tb131072__v001", "uid": "mmlusr_answer_only_human_aging__e2a52482__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_answer_only_human_aging__e2a52482__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 969.8604590635451, \"nll_base\": 6.87715220451355, \"eval_ref\": \"mmlusr_answer_only_human_aging__e2a52482__s000__tb65536__v000\", \"...
{ "index": "mmlusr_answer_only_human_aging__e2a52482__s000__tb65536__v000", "uid": "mmlusr_answer_only_human_aging__e2a52482__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_answer_only_human_sexuality__a91ecbc3__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 703.6832294025023, \"nll_base\": 6.556328296661377, \"eval_ref\": \"mmlusr_answer_only_human_sexuality__a91ecbc3__s000__tb262144...
{ "index": "mmlusr_answer_only_human_sexuality__a91ecbc3__s000__tb262144__v000", "uid": "mmlusr_answer_only_human_sexuality__a91ecbc3__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_answer_only_international_law__14511359__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 304.40737733171744, \"nll_base\": 5.718366861343384, \"eval_ref\": \"mmlusr_answer_only_international_law__14511359__s000__tb1...
{ "index": "mmlusr_answer_only_international_law__14511359__s000__tb131072__v000", "uid": "mmlusr_answer_only_international_law__14511359__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_answer_only_management__1cbc8113__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1535.3432148321317, \"nll_base\": 7.3365092277526855, \"eval_ref\": \"mmlusr_answer_only_management__1cbc8113__s000__tb65536__v000\", \...
{ "index": "mmlusr_answer_only_management__1cbc8113__s000__tb65536__v000", "uid": "mmlusr_answer_only_management__1cbc8113__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_answer_only_marketing__ef1707e1__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 588.1279592122006, \"nll_base\": 6.376944541931152, \"eval_ref\": \"mmlusr_answer_only_marketing__ef1707e1__s000__tb131072__v000\", \"...
{ "index": "mmlusr_answer_only_marketing__ef1707e1__s000__tb131072__v000", "uid": "mmlusr_answer_only_marketing__ef1707e1__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_answer_only_marketing__ef1707e1__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 588.1279592122006, \"nll_base\": 6.376944541931152, \"eval_ref\": \"mmlusr_answer_only_marketing__ef1707e1__s000__tb262144__v001\", \"...
{ "index": "mmlusr_answer_only_marketing__ef1707e1__s000__tb262144__v001", "uid": "mmlusr_answer_only_marketing__ef1707e1__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_answer_only_moral_disputes__64386c01__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 714.9036985083412, \"nll_base\": 6.572147846221924, \"eval_ref\": \"mmlusr_answer_only_moral_disputes__64386c01__s000__tb131072__...
{ "index": "mmlusr_answer_only_moral_disputes__64386c01__s000__tb131072__v000", "uid": "mmlusr_answer_only_moral_disputes__64386c01__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_answer_only_moral_scenarios__e7bbc002__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 892.6145764875478, \"nll_base\": 6.79415488243103, \"eval_ref\": \"mmlusr_answer_only_moral_scenarios__e7bbc002__s000__tb131072_...
{ "index": "mmlusr_answer_only_moral_scenarios__e7bbc002__s000__tb131072__v000", "uid": "mmlusr_answer_only_moral_scenarios__e7bbc002__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_answer_only_moral_scenarios__e7bbc002__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 892.6145764875478, \"nll_base\": 6.79415488243103, \"eval_ref\": \"mmlusr_answer_only_moral_scenarios__e7bbc002__s000__tb262144_...
{ "index": "mmlusr_answer_only_moral_scenarios__e7bbc002__s000__tb262144__v001", "uid": "mmlusr_answer_only_moral_scenarios__e7bbc002__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_answer_only_nutrition__855a2593__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 711.1494540537736, \"nll_base\": 6.566882610321045, \"eval_ref\": \"mmlusr_answer_only_nutrition__855a2593__s000__tb65536__v000\", \"ora...
{ "index": "mmlusr_answer_only_nutrition__855a2593__s000__tb65536__v000", "uid": "mmlusr_answer_only_nutrition__855a2593__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_answer_only_prehistory__cc9b92b2__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1010.4255134087839, \"nll_base\": 6.918126821517944, \"eval_ref\": \"mmlusr_answer_only_prehistory__cc9b92b2__s000__tb262144__v000\",...
{ "index": "mmlusr_answer_only_prehistory__cc9b92b2__s000__tb262144__v000", "uid": "mmlusr_answer_only_prehistory__cc9b92b2__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_answer_only_prehistory__cc9b92b2__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1010.4255134087839, \"nll_base\": 6.918126821517944, \"eval_ref\": \"mmlusr_answer_only_prehistory__cc9b92b2__s000__tb65536__v001\", \"...
{ "index": "mmlusr_answer_only_prehistory__cc9b92b2__s000__tb65536__v001", "uid": "mmlusr_answer_only_prehistory__cc9b92b2__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_answer_only_professional_accounting__a209d2c6__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 583.0842404969619, \"nll_base\": 6.368331670761108, \"eval_ref\": \"mmlusr_answer_only_professional_accounting__a209d2c6...
{ "index": "mmlusr_answer_only_professional_accounting__a209d2c6__s000__tb131072__v000", "uid": "mmlusr_answer_only_professional_accounting__a209d2c6__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_answer_only_professional_law__f6351f07__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 219.68270548298952, \"nll_base\": 5.392184257507324, \"eval_ref\": \"mmlusr_answer_only_professional_law__f6351f07__s000__tb65536...
{ "index": "mmlusr_answer_only_professional_law__f6351f07__s000__tb65536__v000", "uid": "mmlusr_answer_only_professional_law__f6351f07__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_answer_only_professional_medicine__0bd73972__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 530.6679750469561, \"nll_base\": 6.274136543273926, \"eval_ref\": \"mmlusr_answer_only_professional_medicine__0bd73972__s0...
{ "index": "mmlusr_answer_only_professional_medicine__0bd73972__s000__tb131072__v000", "uid": "mmlusr_answer_only_professional_medicine__0bd73972__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_answer_only_professional_psychology__15426d59__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 561.2203725695284, \"nll_base\": 6.330113649368286, \"eval_ref\": \"mmlusr_answer_only_professional_psychology__15426d59...
{ "index": "mmlusr_answer_only_professional_psychology__15426d59__s000__tb262144__v000", "uid": "mmlusr_answer_only_professional_psychology__15426d59__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_answer_only_public_relations__458567c3__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 844.5316112289036, \"nll_base\": 6.738782167434692, \"eval_ref\": \"mmlusr_answer_only_public_relations__458567c3__s000__tb65536_...
{ "index": "mmlusr_answer_only_public_relations__458567c3__s000__tb65536__v001", "uid": "mmlusr_answer_only_public_relations__458567c3__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_answer_only_sociology__4270fe17__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 565.3934109791046, \"nll_base\": 6.33752179145813, \"eval_ref\": \"mmlusr_answer_only_sociology__4270fe17__s000__tb131072__v000\", \"o...
{ "index": "mmlusr_answer_only_sociology__4270fe17__s000__tb131072__v000", "uid": "mmlusr_answer_only_sociology__4270fe17__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_answer_only_sociology__4270fe17__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 565.3934109791046, \"nll_base\": 6.33752179145813, \"eval_ref\": \"mmlusr_answer_only_sociology__4270fe17__s000__tb262144__v001\", \"o...
{ "index": "mmlusr_answer_only_sociology__4270fe17__s000__tb262144__v001", "uid": "mmlusr_answer_only_sociology__4270fe17__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_answer_only_virology__5642a4ca__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 798.9511299610333, \"nll_base\": 6.683299779891968, \"eval_ref\": \"mmlusr_answer_only_virology__5642a4ca__s000__tb131072__v000\", \"or...
{ "index": "mmlusr_answer_only_virology__5642a4ca__s000__tb131072__v000", "uid": "mmlusr_answer_only_virology__5642a4ca__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_answer_only_world_religions__4a0fb7a2__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1544.212976225995, \"nll_base\": 7.342269659042358, \"eval_ref\": \"mmlusr_answer_only_world_religions__4a0fb7a2__s000__tb65536__v...
{ "index": "mmlusr_answer_only_world_religions__4a0fb7a2__s000__tb65536__v000", "uid": "mmlusr_answer_only_world_religions__4a0fb7a2__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_anatomy__fb236192__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 615.6868981519764, \"nll_base\": 6.422738552093506, \"eval_ref\": \"mmlusr_question_and_answer_anatomy__fb236192__s000__tb131072...
{ "index": "mmlusr_question_and_answer_anatomy__fb236192__s000__tb131072__v001", "uid": "mmlusr_question_and_answer_anatomy__fb236192__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_anatomy__fb236192__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 615.6868981519764, \"nll_base\": 6.422738552093506, \"eval_ref\": \"mmlusr_question_and_answer_anatomy__fb236192__s000__tb65536__v...
{ "index": "mmlusr_question_and_answer_anatomy__fb236192__s000__tb65536__v000", "uid": "mmlusr_question_and_answer_anatomy__fb236192__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_astronomy__8601fc29__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 683.479814089214, \"nll_base\": 6.5271971225738525, \"eval_ref\": \"mmlusr_question_and_answer_astronomy__8601fc29__s000__tb13...
{ "index": "mmlusr_question_and_answer_astronomy__8601fc29__s000__tb131072__v000", "uid": "mmlusr_question_and_answer_astronomy__8601fc29__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_astronomy__8601fc29__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 683.479814089214, \"nll_base\": 6.5271971225738525, \"eval_ref\": \"mmlusr_question_and_answer_astronomy__8601fc29__s000__tb26...
{ "index": "mmlusr_question_and_answer_astronomy__8601fc29__s000__tb262144__v001", "uid": "mmlusr_question_and_answer_astronomy__8601fc29__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_business_ethics__6378f8a8__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 727.2921756552205, \"nll_base\": 6.589328289031982, \"eval_ref\": \"mmlusr_question_and_answer_business_ethics__6378f8a8...
{ "index": "mmlusr_question_and_answer_business_ethics__6378f8a8__s000__tb131072__v000", "uid": "mmlusr_question_and_answer_business_ethics__6378f8a8__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_business_ethics__6378f8a8__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 727.2921756552205, \"nll_base\": 6.589328289031982, \"eval_ref\": \"mmlusr_question_and_answer_business_ethics__6378f8a8...
{ "index": "mmlusr_question_and_answer_business_ethics__6378f8a8__s000__tb262144__v001", "uid": "mmlusr_question_and_answer_business_ethics__6378f8a8__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_clinical_knowledge__5035a911__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 585.0132422716205, \"nll_base\": 6.371634483337402, \"eval_ref\": \"mmlusr_question_and_answer_clinical_knowledge__50...
{ "index": "mmlusr_question_and_answer_clinical_knowledge__5035a911__s000__tb131072__v001", "uid": "mmlusr_question_and_answer_clinical_knowledge__5035a911__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_and_answer_clinical_knowledge__5035a911__s000__tb262144__v002\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 585.0132422716205, \"nll_base\": 6.371634483337402, \"eval_ref\": \"mmlusr_question_and_answer_clinical_knowledge__50...
{ "index": "mmlusr_question_and_answer_clinical_knowledge__5035a911__s000__tb262144__v002", "uid": "mmlusr_question_and_answer_clinical_knowledge__5035a911__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_and_answer_clinical_knowledge__5035a911__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 585.0132422716205, \"nll_base\": 6.371634483337402, \"eval_ref\": \"mmlusr_question_and_answer_clinical_knowledge__5035...
{ "index": "mmlusr_question_and_answer_clinical_knowledge__5035a911__s000__tb65536__v000", "uid": "mmlusr_question_and_answer_clinical_knowledge__5035a911__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_and_answer_college_biology__814853b1__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 415.7489159211427, \"nll_base\": 6.030081510543823, \"eval_ref\": \"mmlusr_question_and_answer_college_biology__814853b1...
{ "index": "mmlusr_question_and_answer_college_biology__814853b1__s000__tb131072__v000", "uid": "mmlusr_question_and_answer_college_biology__814853b1__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_college_biology__814853b1__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 415.7489159211427, \"nll_base\": 6.030081510543823, \"eval_ref\": \"mmlusr_question_and_answer_college_biology__814853b1...
{ "index": "mmlusr_question_and_answer_college_biology__814853b1__s000__tb262144__v001", "uid": "mmlusr_question_and_answer_college_biology__814853b1__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_college_chemistry__59134c7d__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 926.4738778684133, \"nll_base\": 6.831385850906372, \"eval_ref\": \"mmlusr_question_and_answer_college_chemistry__5913...
{ "index": "mmlusr_question_and_answer_college_chemistry__59134c7d__s000__tb131072__v000", "uid": "mmlusr_question_and_answer_college_chemistry__59134c7d__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_college_mathematics__5e1846f8__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 884.4930264209023, \"nll_base\": 6.785014629364014, \"eval_ref\": \"mmlusr_question_and_answer_college_mathematics__...
{ "index": "mmlusr_question_and_answer_college_mathematics__5e1846f8__s000__tb131072__v001", "uid": "mmlusr_question_and_answer_college_mathematics__5e1846f8__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_college_mathematics__5e1846f8__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 884.4930264209023, \"nll_base\": 6.785014629364014, \"eval_ref\": \"mmlusr_question_and_answer_college_mathematics__5e...
{ "index": "mmlusr_question_and_answer_college_mathematics__5e1846f8__s000__tb65536__v000", "uid": "mmlusr_question_and_answer_college_mathematics__5e1846f8__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_college_medicine__cc22281e__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 635.2954195897322, \"nll_base\": 6.454090118408203, \"eval_ref\": \"mmlusr_question_and_answer_college_medicine__cc2228...
{ "index": "mmlusr_question_and_answer_college_medicine__cc22281e__s000__tb131072__v000", "uid": "mmlusr_question_and_answer_college_medicine__cc22281e__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_computer_security__02bf2dab__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 814.0754646074606, \"nll_base\": 6.702053070068359, \"eval_ref\": \"mmlusr_question_and_answer_computer_security__02bf2d...
{ "index": "mmlusr_question_and_answer_computer_security__02bf2dab__s000__tb65536__v000", "uid": "mmlusr_question_and_answer_computer_security__02bf2dab__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_conceptual_physics__14518a99__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 707.8917208355109, \"nll_base\": 6.562291145324707, \"eval_ref\": \"mmlusr_question_and_answer_conceptual_physics__14...
{ "index": "mmlusr_question_and_answer_conceptual_physics__14518a99__s000__tb262144__v000", "uid": "mmlusr_question_and_answer_conceptual_physics__14518a99__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_conceptual_physics__14518a99__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 707.8917208355109, \"nll_base\": 6.562291145324707, \"eval_ref\": \"mmlusr_question_and_answer_conceptual_physics__1451...
{ "index": "mmlusr_question_and_answer_conceptual_physics__14518a99__s000__tb65536__v001", "uid": "mmlusr_question_and_answer_conceptual_physics__14518a99__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_econometrics__789f7403__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 547.6161884553956, \"nll_base\": 6.305574655532837, \"eval_ref\": \"mmlusr_question_and_answer_econometrics__789f7403__s000...
{ "index": "mmlusr_question_and_answer_econometrics__789f7403__s000__tb131072__v000", "uid": "mmlusr_question_and_answer_econometrics__789f7403__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_elementary_mathematics__7bc66837__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1345.7197210548045, \"nll_base\": 7.204684257507324, \"eval_ref\": \"mmlusr_question_and_answer_elementary_mathem...
{ "index": "mmlusr_question_and_answer_elementary_mathematics__7bc66837__s000__tb131072__v000", "uid": "mmlusr_question_and_answer_elementary_mathematics__7bc66837__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_elementary_mathematics__7bc66837__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1345.7197210548045, \"nll_base\": 7.204684257507324, \"eval_ref\": \"mmlusr_question_and_answer_elementary_mathem...
{ "index": "mmlusr_question_and_answer_elementary_mathematics__7bc66837__s000__tb262144__v001", "uid": "mmlusr_question_and_answer_elementary_mathematics__7bc66837__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_formal_logic__014d9743__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 682.0494336123818, \"nll_base\": 6.525102138519287, \"eval_ref\": \"mmlusr_question_and_answer_formal_logic__014d9743__s000...
{ "index": "mmlusr_question_and_answer_formal_logic__014d9743__s000__tb262144__v000", "uid": "mmlusr_question_and_answer_formal_logic__014d9743__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_formal_logic__014d9743__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 682.0494336123818, \"nll_base\": 6.525102138519287, \"eval_ref\": \"mmlusr_question_and_answer_formal_logic__014d9743__s000__...
{ "index": "mmlusr_question_and_answer_formal_logic__014d9743__s000__tb65536__v001", "uid": "mmlusr_question_and_answer_formal_logic__014d9743__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_global_facts__3a2a5917__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1166.7013223308936, \"nll_base\": 7.061935663223267, \"eval_ref\": \"mmlusr_question_and_answer_global_facts__3a2a5917__s00...
{ "index": "mmlusr_question_and_answer_global_facts__3a2a5917__s000__tb131072__v000", "uid": "mmlusr_question_and_answer_global_facts__3a2a5917__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_global_facts__3a2a5917__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1166.7013223308936, \"nll_base\": 7.061935663223267, \"eval_ref\": \"mmlusr_question_and_answer_global_facts__3a2a5917__s00...
{ "index": "mmlusr_question_and_answer_global_facts__3a2a5917__s000__tb262144__v001", "uid": "mmlusr_question_and_answer_global_facts__3a2a5917__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_high_school_biology__03998f5b__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 688.3094899911539, \"nll_base\": 6.534238576889038, \"eval_ref\": \"mmlusr_question_and_answer_high_school_biology__...
{ "index": "mmlusr_question_and_answer_high_school_biology__03998f5b__s000__tb131072__v000", "uid": "mmlusr_question_and_answer_high_school_biology__03998f5b__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_high_school_biology__03998f5b__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 688.3094899911539, \"nll_base\": 6.534238576889038, \"eval_ref\": \"mmlusr_question_and_answer_high_school_biology__...
{ "index": "mmlusr_question_and_answer_high_school_biology__03998f5b__s000__tb262144__v001", "uid": "mmlusr_question_and_answer_high_school_biology__03998f5b__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_high_school_chemistry__9df7d057__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 429.57633422015675, \"nll_base\": 6.062799453735352, \"eval_ref\": \"mmlusr_question_and_answer_high_school_chemis...
{ "index": "mmlusr_question_and_answer_high_school_chemistry__9df7d057__s000__tb131072__v000", "uid": "mmlusr_question_and_answer_high_school_chemistry__9df7d057__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_high_school_european_history__31166648__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 757.1659159809493, \"nll_base\": 6.629582405090332, \"eval_ref\": \"mmlusr_question_and_answer_high_school_...
{ "index": "mmlusr_question_and_answer_high_school_european_history__31166648__s000__tb131072__v000", "uid": "mmlusr_question_and_answer_high_school_european_history__31166648__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_high_school_european_history__31166648__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 757.1659159809493, \"nll_base\": 6.629582405090332, \"eval_ref\": \"mmlusr_question_and_answer_high_school_...
{ "index": "mmlusr_question_and_answer_high_school_european_history__31166648__s000__tb262144__v001", "uid": "mmlusr_question_and_answer_high_school_european_history__31166648__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_high_school_geography__d6a6c8c1__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 558.5248013422382, \"nll_base\": 6.325299024581909, \"eval_ref\": \"mmlusr_question_and_answer_high_school_geograp...
{ "index": "mmlusr_question_and_answer_high_school_geography__d6a6c8c1__s000__tb131072__v001", "uid": "mmlusr_question_and_answer_high_school_geography__d6a6c8c1__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_high_school_geography__d6a6c8c1__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 558.5248013422382, \"nll_base\": 6.325299024581909, \"eval_ref\": \"mmlusr_question_and_answer_high_school_geography...
{ "index": "mmlusr_question_and_answer_high_school_geography__d6a6c8c1__s000__tb65536__v000", "uid": "mmlusr_question_and_answer_high_school_geography__d6a6c8c1__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_high_school_mathematics__1bd042d3__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1913.3845000792296, \"nll_base\": 7.556628942489624, \"eval_ref\": \"mmlusr_question_and_answer_high_school_math...
{ "index": "mmlusr_question_and_answer_high_school_mathematics__1bd042d3__s000__tb262144__v000", "uid": "mmlusr_question_and_answer_high_school_mathematics__1bd042d3__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_high_school_psychology__8cd257a8__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 890.0219948772061, \"nll_base\": 6.791246175765991, \"eval_ref\": \"mmlusr_question_and_answer_high_school_psycho...
{ "index": "mmlusr_question_and_answer_high_school_psychology__8cd257a8__s000__tb131072__v000", "uid": "mmlusr_question_and_answer_high_school_psychology__8cd257a8__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_high_school_statistics__bb2c2b87__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 902.472402790939, \"nll_base\": 6.805138111114502, \"eval_ref\": \"mmlusr_question_and_answer_high_school_statist...
{ "index": "mmlusr_question_and_answer_high_school_statistics__bb2c2b87__s000__tb131072__v001", "uid": "mmlusr_question_and_answer_high_school_statistics__bb2c2b87__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_high_school_statistics__bb2c2b87__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 902.472402790939, \"nll_base\": 6.805138111114502, \"eval_ref\": \"mmlusr_question_and_answer_high_school_statistic...
{ "index": "mmlusr_question_and_answer_high_school_statistics__bb2c2b87__s000__tb65536__v000", "uid": "mmlusr_question_and_answer_high_school_statistics__bb2c2b87__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_high_school_us_history__8086a56f__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 549.9984954088703, \"nll_base\": 6.309915542602539, \"eval_ref\": \"mmlusr_question_and_answer_high_school_us_his...
{ "index": "mmlusr_question_and_answer_high_school_us_history__8086a56f__s000__tb262144__v000", "uid": "mmlusr_question_and_answer_high_school_us_history__8086a56f__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_high_school_us_history__8086a56f__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 549.9984954088703, \"nll_base\": 6.309915542602539, \"eval_ref\": \"mmlusr_question_and_answer_high_school_us_histo...
{ "index": "mmlusr_question_and_answer_high_school_us_history__8086a56f__s000__tb65536__v001", "uid": "mmlusr_question_and_answer_high_school_us_history__8086a56f__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_human_aging__114fff8b__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 808.8678526480803, \"nll_base\": 6.695635557174683, \"eval_ref\": \"mmlusr_question_and_answer_human_aging__114fff8b__s000__...
{ "index": "mmlusr_question_and_answer_human_aging__114fff8b__s000__tb131072__v001", "uid": "mmlusr_question_and_answer_human_aging__114fff8b__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_human_aging__114fff8b__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 808.8678526480803, \"nll_base\": 6.695635557174683, \"eval_ref\": \"mmlusr_question_and_answer_human_aging__114fff8b__s000__tb...
{ "index": "mmlusr_question_and_answer_human_aging__114fff8b__s000__tb65536__v000", "uid": "mmlusr_question_and_answer_human_aging__114fff8b__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_international_law__fde83772__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 347.8517424183393, \"nll_base\": 5.851776361465454, \"eval_ref\": \"mmlusr_question_and_answer_international_law__fde8...
{ "index": "mmlusr_question_and_answer_international_law__fde83772__s000__tb131072__v000", "uid": "mmlusr_question_and_answer_international_law__fde83772__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_jurisprudence__95a462de__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 478.65688704645385, \"nll_base\": 6.1709840297698975, \"eval_ref\": \"mmlusr_question_and_answer_jurisprudence__95a462de__...
{ "index": "mmlusr_question_and_answer_jurisprudence__95a462de__s000__tb131072__v000", "uid": "mmlusr_question_and_answer_jurisprudence__95a462de__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_jurisprudence__95a462de__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 478.65688704645385, \"nll_base\": 6.1709840297698975, \"eval_ref\": \"mmlusr_question_and_answer_jurisprudence__95a462de__...
{ "index": "mmlusr_question_and_answer_jurisprudence__95a462de__s000__tb262144__v001", "uid": "mmlusr_question_and_answer_jurisprudence__95a462de__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_machine_learning__b600aa2b__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 586.8025470979431, \"nll_base\": 6.374688386917114, \"eval_ref\": \"mmlusr_question_and_answer_machine_learning__b600aa2b...
{ "index": "mmlusr_question_and_answer_machine_learning__b600aa2b__s000__tb65536__v000", "uid": "mmlusr_question_and_answer_machine_learning__b600aa2b__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_management__2f696a12__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1662.4180857773918, \"nll_base\": 7.4160284996032715, \"eval_ref\": \"mmlusr_question_and_answer_management__2f696a12__s000__...
{ "index": "mmlusr_question_and_answer_management__2f696a12__s000__tb262144__v000", "uid": "mmlusr_question_and_answer_management__2f696a12__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_management__2f696a12__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1662.4180857773918, \"nll_base\": 7.4160284996032715, \"eval_ref\": \"mmlusr_question_and_answer_management__2f696a12__s000__tb...
{ "index": "mmlusr_question_and_answer_management__2f696a12__s000__tb65536__v001", "uid": "mmlusr_question_and_answer_management__2f696a12__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_medical_genetics__1a6d8596__s000__tb131072__v002\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 900.7295204628006, \"nll_base\": 6.8032050132751465, \"eval_ref\": \"mmlusr_question_and_answer_medical_genetics__1a6d8...
{ "index": "mmlusr_question_and_answer_medical_genetics__1a6d8596__s000__tb131072__v002", "uid": "mmlusr_question_and_answer_medical_genetics__1a6d8596__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\": \"mmlusr_question_and_answer_medical_genetics__1a6d8596__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 900.7295204628006, \"nll_base\": 6.8032050132751465, \"eval_ref\": \"mmlusr_question_and_answer_medical_genetics__1a6d8...
{ "index": "mmlusr_question_and_answer_medical_genetics__1a6d8596__s000__tb262144__v000", "uid": "mmlusr_question_and_answer_medical_genetics__1a6d8596__s000__tb262144__v000", "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_and_answer_medical_genetics__1a6d8596__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 900.7295204628006, \"nll_base\": 6.8032050132751465, \"eval_ref\": \"mmlusr_question_and_answer_medical_genetics__1a6d859...
{ "index": "mmlusr_question_and_answer_medical_genetics__1a6d8596__s000__tb65536__v001", "uid": "mmlusr_question_and_answer_medical_genetics__1a6d8596__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\": \"mmlusr_question_and_answer_medical_genetics__1a6d8596__s001__tb262144__v003\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 749.6034007406769, \"nll_base\": 6.619544267654419, \"eval_ref\": \"mmlusr_question_and_answer_medical_genetics__1a6d85...
{ "index": "mmlusr_question_and_answer_medical_genetics__1a6d8596__s001__tb262144__v003", "uid": "mmlusr_question_and_answer_medical_genetics__1a6d8596__s001__tb262144__v003", "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_and_answer_miscellaneous__16608824__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1143.6948161124074, \"nll_base\": 7.042019367218018, \"eval_ref\": \"mmlusr_question_and_answer_miscellaneous__16608824__s00...
{ "index": "mmlusr_question_and_answer_miscellaneous__16608824__s000__tb65536__v000", "uid": "mmlusr_question_and_answer_miscellaneous__16608824__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_moral_scenarios__1087724d__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 723.7834218721125, \"nll_base\": 6.584492206573486, \"eval_ref\": \"mmlusr_question_and_answer_moral_scenarios__1087724d...
{ "index": "mmlusr_question_and_answer_moral_scenarios__1087724d__s000__tb131072__v001", "uid": "mmlusr_question_and_answer_moral_scenarios__1087724d__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_moral_scenarios__1087724d__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 723.7834218721125, \"nll_base\": 6.584492206573486, \"eval_ref\": \"mmlusr_question_and_answer_moral_scenarios__1087724d__...
{ "index": "mmlusr_question_and_answer_moral_scenarios__1087724d__s000__tb65536__v000", "uid": "mmlusr_question_and_answer_moral_scenarios__1087724d__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_philosophy__c5d8bdba__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 768.7393963073376, \"nll_base\": 6.644752025604248, \"eval_ref\": \"mmlusr_question_and_answer_philosophy__c5d8bdba__s000__tb...
{ "index": "mmlusr_question_and_answer_philosophy__c5d8bdba__s000__tb131072__v001", "uid": "mmlusr_question_and_answer_philosophy__c5d8bdba__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_philosophy__c5d8bdba__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 768.7393963073376, \"nll_base\": 6.644752025604248, \"eval_ref\": \"mmlusr_question_and_answer_philosophy__c5d8bdba__s000__tb65...
{ "index": "mmlusr_question_and_answer_philosophy__c5d8bdba__s000__tb65536__v000", "uid": "mmlusr_question_and_answer_philosophy__c5d8bdba__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_prehistory__4bec01f7__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 728.3960096610557, \"nll_base\": 6.5908448696136475, \"eval_ref\": \"mmlusr_question_and_answer_prehistory__4bec01f7__s000__t...
{ "index": "mmlusr_question_and_answer_prehistory__4bec01f7__s000__tb131072__v000", "uid": "mmlusr_question_and_answer_prehistory__4bec01f7__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_prehistory__4bec01f7__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 728.3960096610557, \"nll_base\": 6.5908448696136475, \"eval_ref\": \"mmlusr_question_and_answer_prehistory__4bec01f7__s000__t...
{ "index": "mmlusr_question_and_answer_prehistory__4bec01f7__s000__tb262144__v001", "uid": "mmlusr_question_and_answer_prehistory__4bec01f7__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_professional_accounting__9b4b2b34__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 576.6782849301267, \"nll_base\": 6.3572845458984375, \"eval_ref\": \"mmlusr_question_and_answer_professional_acc...
{ "index": "mmlusr_question_and_answer_professional_accounting__9b4b2b34__s000__tb262144__v000", "uid": "mmlusr_question_and_answer_professional_accounting__9b4b2b34__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_law__4ae8be34__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 414.559270718215, \"nll_base\": 6.027215957641602, \"eval_ref\": \"mmlusr_question_and_answer_professional_law__4ae8be34_...
{ "index": "mmlusr_question_and_answer_professional_law__4ae8be34__s000__tb65536__v000", "uid": "mmlusr_question_and_answer_professional_law__4ae8be34__s000__tb65536__v000", "axis": "knowledge", "budget": 2048, "token_budget": 65536, "variant": "filter" }