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\": \"bigbench_checkmate_in_one_multiple_choice__3a66c19e__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 50.740352455160796, \"nll_base\": 3.9267215004450158, \"eval_ref\": \"bigbench_checkmate_in_one_multiple_choice__3a66c19e__...
{ "index": "bigbench_checkmate_in_one_multiple_choice__3a66c19e__s000__tb65536__v001", "uid": "bigbench_checkmate_in_one_multiple_choice__3a66c19e__s000__tb65536__v001", "axis": "reasoning", "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\": \"bigbench_cifar10_classification_generate_until__c8bda41d__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1069.6333792131322, \"nll_base\": 6.975071232493331, \"eval_ref\": \"bigbench_cifar10_classification_generate_until_...
{ "index": "bigbench_cifar10_classification_generate_until__c8bda41d__s000__tb131072__v001", "uid": "bigbench_cifar10_classification_generate_until__c8bda41d__s000__tb131072__v001", "axis": "reasoning", "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\": \"bigbench_cifar10_classification_generate_until__c8bda41d__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1069.6333792131322, \"nll_base\": 6.975071232493331, \"eval_ref\": \"bigbench_cifar10_classification_generate_until__c...
{ "index": "bigbench_cifar10_classification_generate_until__c8bda41d__s000__tb65536__v000", "uid": "bigbench_cifar10_classification_generate_until__c8bda41d__s000__tb65536__v000", "axis": "reasoning", "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\": \"bigbench_cifar10_classification_multiple_choice__b0ce235d__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1303.0669178462579, \"nll_base\": 7.172475932556906, \"eval_ref\": \"bigbench_cifar10_classification_multiple_choic...
{ "index": "bigbench_cifar10_classification_multiple_choice__b0ce235d__s000__tb131072__v000", "uid": "bigbench_cifar10_classification_multiple_choice__b0ce235d__s000__tb131072__v000", "axis": "reasoning", "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\": \"bigbench_cifar10_classification_multiple_choice__b0ce235d__s000__tb65536__v002\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1303.0669178462579, \"nll_base\": 7.172475932556906, \"eval_ref\": \"bigbench_cifar10_classification_multiple_choice_...
{ "index": "bigbench_cifar10_classification_multiple_choice__b0ce235d__s000__tb65536__v002", "uid": "bigbench_cifar10_classification_multiple_choice__b0ce235d__s000__tb65536__v002", "axis": "reasoning", "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\": \"bigbench_code_line_description_multiple_choice__4d302ca8__s000__tb131072__v002\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4.111120336633118, \"nll_base\": 1.4136955793513808, \"eval_ref\": \"bigbench_code_line_description_multiple_choice_...
{ "index": "bigbench_code_line_description_multiple_choice__4d302ca8__s000__tb131072__v002", "uid": "bigbench_code_line_description_multiple_choice__4d302ca8__s000__tb131072__v002", "axis": "code", "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\": \"bigbench_code_line_description_multiple_choice__4d302ca8__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4.111120336633118, \"nll_base\": 1.4136955793513808, \"eval_ref\": \"bigbench_code_line_description_multiple_choice__4...
{ "index": "bigbench_code_line_description_multiple_choice__4d302ca8__s000__tb65536__v001", "uid": "bigbench_code_line_description_multiple_choice__4d302ca8__s000__tb65536__v001", "axis": "code", "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\": \"bigbench_code_line_description_multiple_choice__4d302ca8__s001__tb65536__v004\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4.767528361850376, \"nll_base\": 1.5618280074732476, \"eval_ref\": \"bigbench_code_line_description_multiple_choice__4...
{ "index": "bigbench_code_line_description_multiple_choice__4d302ca8__s001__tb65536__v004", "uid": "bigbench_code_line_description_multiple_choice__4d302ca8__s001__tb65536__v004", "axis": "code", "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\": \"bigbench_code_line_description_multiple_choice__4d302ca8__s002__tb65536__v007\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 5.729455346407893, \"nll_base\": 1.7456204732259115, \"eval_ref\": \"bigbench_code_line_description_multiple_choice__4...
{ "index": "bigbench_code_line_description_multiple_choice__4d302ca8__s002__tb65536__v007", "uid": "bigbench_code_line_description_multiple_choice__4d302ca8__s002__tb65536__v007", "axis": "code", "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\": \"bigbench_code_line_description_multiple_choice__4d302ca8__s003__tb131072__v011\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4.111120336633118, \"nll_base\": 1.4136955793513808, \"eval_ref\": \"bigbench_code_line_description_multiple_choice_...
{ "index": "bigbench_code_line_description_multiple_choice__4d302ca8__s003__tb131072__v011", "uid": "bigbench_code_line_description_multiple_choice__4d302ca8__s003__tb131072__v011", "axis": "code", "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\": \"bigbench_code_line_description_multiple_choice__4d302ca8__s003__tb65536__v010\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4.111120336633118, \"nll_base\": 1.4136955793513808, \"eval_ref\": \"bigbench_code_line_description_multiple_choice__4...
{ "index": "bigbench_code_line_description_multiple_choice__4d302ca8__s003__tb65536__v010", "uid": "bigbench_code_line_description_multiple_choice__4d302ca8__s003__tb65536__v010", "axis": "code", "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\": \"bigbench_code_line_description_multiple_choice__4d302ca8__s004__tb65536__v013\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4.767528361850376, \"nll_base\": 1.5618280074732476, \"eval_ref\": \"bigbench_code_line_description_multiple_choice__4...
{ "index": "bigbench_code_line_description_multiple_choice__4d302ca8__s004__tb65536__v013", "uid": "bigbench_code_line_description_multiple_choice__4d302ca8__s004__tb65536__v013", "axis": "code", "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\": \"bigbench_code_line_description_multiple_choice__4d302ca8__s005__tb65536__v016\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 5.729455346407893, \"nll_base\": 1.7456204732259115, \"eval_ref\": \"bigbench_code_line_description_multiple_choice__4...
{ "index": "bigbench_code_line_description_multiple_choice__4d302ca8__s005__tb65536__v016", "uid": "bigbench_code_line_description_multiple_choice__4d302ca8__s005__tb65536__v016", "axis": "code", "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\": \"bigbench_code_line_description_multiple_choice__4d302ca8__s006__tb131072__v020\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4.111120336633118, \"nll_base\": 1.4136955793513808, \"eval_ref\": \"bigbench_code_line_description_multiple_choice_...
{ "index": "bigbench_code_line_description_multiple_choice__4d302ca8__s006__tb131072__v020", "uid": "bigbench_code_line_description_multiple_choice__4d302ca8__s006__tb131072__v020", "axis": "code", "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\": \"bigbench_code_line_description_multiple_choice__4d302ca8__s006__tb65536__v019\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4.111120336633118, \"nll_base\": 1.4136955793513808, \"eval_ref\": \"bigbench_code_line_description_multiple_choice__4...
{ "index": "bigbench_code_line_description_multiple_choice__4d302ca8__s006__tb65536__v019", "uid": "bigbench_code_line_description_multiple_choice__4d302ca8__s006__tb65536__v019", "axis": "code", "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\": \"bigbench_code_line_description_multiple_choice__4d302ca8__s007__tb65536__v022\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4.767528361850376, \"nll_base\": 1.5618280074732476, \"eval_ref\": \"bigbench_code_line_description_multiple_choice__4...
{ "index": "bigbench_code_line_description_multiple_choice__4d302ca8__s007__tb65536__v022", "uid": "bigbench_code_line_description_multiple_choice__4d302ca8__s007__tb65536__v022", "axis": "code", "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\": \"bigbench_code_line_description_multiple_choice__4d302ca8__s008__tb65536__v025\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 5.729455346407893, \"nll_base\": 1.7456204732259115, \"eval_ref\": \"bigbench_code_line_description_multiple_choice__4...
{ "index": "bigbench_code_line_description_multiple_choice__4d302ca8__s008__tb65536__v025", "uid": "bigbench_code_line_description_multiple_choice__4d302ca8__s008__tb65536__v025", "axis": "code", "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\": \"bigbench_code_line_description_multiple_choice__4d302ca8__s009__tb131072__v029\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4.111120336633118, \"nll_base\": 1.4136955793513808, \"eval_ref\": \"bigbench_code_line_description_multiple_choice_...
{ "index": "bigbench_code_line_description_multiple_choice__4d302ca8__s009__tb131072__v029", "uid": "bigbench_code_line_description_multiple_choice__4d302ca8__s009__tb131072__v029", "axis": "code", "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\": \"bigbench_code_line_description_multiple_choice__4d302ca8__s009__tb65536__v028\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4.111120336633118, \"nll_base\": 1.4136955793513808, \"eval_ref\": \"bigbench_code_line_description_multiple_choice__4...
{ "index": "bigbench_code_line_description_multiple_choice__4d302ca8__s009__tb65536__v028", "uid": "bigbench_code_line_description_multiple_choice__4d302ca8__s009__tb65536__v028", "axis": "code", "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\": \"bigbench_code_line_description_multiple_choice__4d302ca8__s010__tb65536__v031\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4.767528361850376, \"nll_base\": 1.5618280074732476, \"eval_ref\": \"bigbench_code_line_description_multiple_choice__4...
{ "index": "bigbench_code_line_description_multiple_choice__4d302ca8__s010__tb65536__v031", "uid": "bigbench_code_line_description_multiple_choice__4d302ca8__s010__tb65536__v031", "axis": "code", "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\": \"bigbench_code_line_description_multiple_choice__4d302ca8__s011__tb65536__v034\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 5.729455346407893, \"nll_base\": 1.7456204732259115, \"eval_ref\": \"bigbench_code_line_description_multiple_choice__4...
{ "index": "bigbench_code_line_description_multiple_choice__4d302ca8__s011__tb65536__v034", "uid": "bigbench_code_line_description_multiple_choice__4d302ca8__s011__tb65536__v034", "axis": "code", "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\": \"bigbench_code_line_description_multiple_choice__4d302ca8__s012__tb131072__v038\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4.111120336633118, \"nll_base\": 1.4136955793513808, \"eval_ref\": \"bigbench_code_line_description_multiple_choice_...
{ "index": "bigbench_code_line_description_multiple_choice__4d302ca8__s012__tb131072__v038", "uid": "bigbench_code_line_description_multiple_choice__4d302ca8__s012__tb131072__v038", "axis": "code", "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\": \"bigbench_code_line_description_multiple_choice__4d302ca8__s012__tb65536__v037\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4.111120336633118, \"nll_base\": 1.4136955793513808, \"eval_ref\": \"bigbench_code_line_description_multiple_choice__4...
{ "index": "bigbench_code_line_description_multiple_choice__4d302ca8__s012__tb65536__v037", "uid": "bigbench_code_line_description_multiple_choice__4d302ca8__s012__tb65536__v037", "axis": "code", "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\": \"bigbench_code_line_description_multiple_choice__4d302ca8__s013__tb65536__v040\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4.767528361850376, \"nll_base\": 1.5618280074732476, \"eval_ref\": \"bigbench_code_line_description_multiple_choice__4...
{ "index": "bigbench_code_line_description_multiple_choice__4d302ca8__s013__tb65536__v040", "uid": "bigbench_code_line_description_multiple_choice__4d302ca8__s013__tb65536__v040", "axis": "code", "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\": \"bigbench_color_generate_until__6744efaa__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 26824.145811716076, \"nll_base\": 10.197057723999023, \"eval_ref\": \"bigbench_color_generate_until__6744efaa__s000__tb131072__v001\"...
{ "index": "bigbench_color_generate_until__6744efaa__s000__tb131072__v001", "uid": "bigbench_color_generate_until__6744efaa__s000__tb131072__v001", "axis": "reasoning", "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\": \"bigbench_color_generate_until__6744efaa__s000__tb262144__v002\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 26824.145811716076, \"nll_base\": 10.197057723999023, \"eval_ref\": \"bigbench_color_generate_until__6744efaa__s000__tb262144__v002\"...
{ "index": "bigbench_color_generate_until__6744efaa__s000__tb262144__v002", "uid": "bigbench_color_generate_until__6744efaa__s000__tb262144__v002", "axis": "reasoning", "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\": \"bigbench_color_generate_until__6744efaa__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 26824.145811716076, \"nll_base\": 10.197057723999023, \"eval_ref\": \"bigbench_color_generate_until__6744efaa__s000__tb65536__v000\", \...
{ "index": "bigbench_color_generate_until__6744efaa__s000__tb65536__v000", "uid": "bigbench_color_generate_until__6744efaa__s000__tb65536__v000", "axis": "reasoning", "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\": \"bigbench_color_multiple_choice__bb05a6c9__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 22264.343034267848, \"nll_base\": 10.010741710662842, \"eval_ref\": \"bigbench_color_multiple_choice__bb05a6c9__s000__tb262144__v000...
{ "index": "bigbench_color_multiple_choice__bb05a6c9__s000__tb262144__v000", "uid": "bigbench_color_multiple_choice__bb05a6c9__s000__tb262144__v000", "axis": "reasoning", "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\": \"bigbench_color_multiple_choice__bb05a6c9__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 22264.343034267848, \"nll_base\": 10.010741710662842, \"eval_ref\": \"bigbench_color_multiple_choice__bb05a6c9__s000__tb65536__v001\",...
{ "index": "bigbench_color_multiple_choice__bb05a6c9__s000__tb65536__v001", "uid": "bigbench_color_multiple_choice__bb05a6c9__s000__tb65536__v001", "axis": "reasoning", "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\": \"bigbench_common_morpheme_generate_until__b351c2e2__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2972.0002718132587, \"nll_base\": 7.996990497295673, \"eval_ref\": \"bigbench_common_morpheme_generate_until__b351c2e2__s00...
{ "index": "bigbench_common_morpheme_generate_until__b351c2e2__s000__tb131072__v001", "uid": "bigbench_common_morpheme_generate_until__b351c2e2__s000__tb131072__v001", "axis": "reasoning", "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\": \"bigbench_common_morpheme_generate_until__b351c2e2__s000__tb262144__v002\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2972.0002718132587, \"nll_base\": 7.996990497295673, \"eval_ref\": \"bigbench_common_morpheme_generate_until__b351c2e2__s00...
{ "index": "bigbench_common_morpheme_generate_until__b351c2e2__s000__tb262144__v002", "uid": "bigbench_common_morpheme_generate_until__b351c2e2__s000__tb262144__v002", "axis": "reasoning", "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\": \"bigbench_common_morpheme_generate_until__b351c2e2__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2972.0002718132587, \"nll_base\": 7.996990497295673, \"eval_ref\": \"bigbench_common_morpheme_generate_until__b351c2e2__s000_...
{ "index": "bigbench_common_morpheme_generate_until__b351c2e2__s000__tb65536__v000", "uid": "bigbench_common_morpheme_generate_until__b351c2e2__s000__tb65536__v000", "axis": "reasoning", "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\": \"bigbench_common_morpheme_multiple_choice__1a6cb94f__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1720.0820699082315, \"nll_base\": 7.450127283732097, \"eval_ref\": \"bigbench_common_morpheme_multiple_choice__1a6cb94f__s...
{ "index": "bigbench_common_morpheme_multiple_choice__1a6cb94f__s000__tb262144__v000", "uid": "bigbench_common_morpheme_multiple_choice__1a6cb94f__s000__tb262144__v000", "axis": "reasoning", "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\": \"bigbench_common_morpheme_multiple_choice__1a6cb94f__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1720.0820699082315, \"nll_base\": 7.450127283732097, \"eval_ref\": \"bigbench_common_morpheme_multiple_choice__1a6cb94f__s00...
{ "index": "bigbench_common_morpheme_multiple_choice__1a6cb94f__s000__tb65536__v001", "uid": "bigbench_common_morpheme_multiple_choice__1a6cb94f__s000__tb65536__v001", "axis": "reasoning", "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\": \"bigbench_conceptual_combinations_generate_until__a6cb49f9__s000__tb131072__v002\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 6.371384231360026, \"nll_base\": 1.8518167507429064, \"eval_ref\": \"bigbench_conceptual_combinations_generate_unti...
{ "index": "bigbench_conceptual_combinations_generate_until__a6cb49f9__s000__tb131072__v002", "uid": "bigbench_conceptual_combinations_generate_until__a6cb49f9__s000__tb131072__v002", "axis": "reasoning", "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\": \"bigbench_conceptual_combinations_generate_until__a6cb49f9__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 6.371384231360026, \"nll_base\": 1.8518167507429064, \"eval_ref\": \"bigbench_conceptual_combinations_generate_unti...
{ "index": "bigbench_conceptual_combinations_generate_until__a6cb49f9__s000__tb262144__v000", "uid": "bigbench_conceptual_combinations_generate_until__a6cb49f9__s000__tb262144__v000", "axis": "reasoning", "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\": \"bigbench_conceptual_combinations_generate_until__a6cb49f9__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 6.371384231360026, \"nll_base\": 1.8518167507429064, \"eval_ref\": \"bigbench_conceptual_combinations_generate_until_...
{ "index": "bigbench_conceptual_combinations_generate_until__a6cb49f9__s000__tb65536__v001", "uid": "bigbench_conceptual_combinations_generate_until__a6cb49f9__s000__tb65536__v001", "axis": "reasoning", "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\": \"bigbench_conceptual_combinations_multiple_choice__bae3b008__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 5.922866079851087, \"nll_base\": 1.778820466850304, \"eval_ref\": \"bigbench_conceptual_combinations_multiple_choi...
{ "index": "bigbench_conceptual_combinations_multiple_choice__bae3b008__s000__tb131072__v001", "uid": "bigbench_conceptual_combinations_multiple_choice__bae3b008__s000__tb131072__v001", "axis": "reasoning", "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\": \"bigbench_conceptual_combinations_multiple_choice__bae3b008__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 5.922866079851087, \"nll_base\": 1.778820466850304, \"eval_ref\": \"bigbench_conceptual_combinations_multiple_choice...
{ "index": "bigbench_conceptual_combinations_multiple_choice__bae3b008__s000__tb65536__v000", "uid": "bigbench_conceptual_combinations_multiple_choice__bae3b008__s000__tb65536__v000", "axis": "reasoning", "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\": \"bigbench_contextual_parametric_knowledge_conflicts_generate_until__cf55e9fc__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 15.854322096929321, \"nll_base\": 2.763442150654237, \"eval_ref\": \"bigbench_contextual_parametr...
{ "index": "bigbench_contextual_parametric_knowledge_conflicts_generate_until__cf55e9fc__s000__tb131072__v001", "uid": "bigbench_contextual_parametric_knowledge_conflicts_generate_until__cf55e9fc__s000__tb131072__v001", "axis": "reasoning", "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\": \"bigbench_contextual_parametric_knowledge_conflicts_generate_until__cf55e9fc__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 15.854322096929321, \"nll_base\": 2.763442150654237, \"eval_ref\": \"bigbench_contextual_parametric...
{ "index": "bigbench_contextual_parametric_knowledge_conflicts_generate_until__cf55e9fc__s000__tb65536__v000", "uid": "bigbench_contextual_parametric_knowledge_conflicts_generate_until__cf55e9fc__s000__tb65536__v000", "axis": "reasoning", "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\": \"bigbench_contextual_parametric_knowledge_conflicts_multiple_choice__1cf8adb5__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 13.657547758762872, \"nll_base\": 2.6142923181707207, \"eval_ref\": \"bigbench_contextual_parame...
{ "index": "bigbench_contextual_parametric_knowledge_conflicts_multiple_choice__1cf8adb5__s000__tb131072__v000", "uid": "bigbench_contextual_parametric_knowledge_conflicts_multiple_choice__1cf8adb5__s000__tb131072__v000", "axis": "reasoning", "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\": \"bigbench_crash_blossom_multiple_choice__fcee20e7__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4553.158081291059, \"nll_base\": 8.423576354980469, \"eval_ref\": \"bigbench_crash_blossom_multiple_choice__fcee20e7__s000__...
{ "index": "bigbench_crash_blossom_multiple_choice__fcee20e7__s000__tb131072__v001", "uid": "bigbench_crash_blossom_multiple_choice__fcee20e7__s000__tb131072__v001", "axis": "reasoning", "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\": \"bigbench_crass_ai_generate_until__d2b5624c__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 9.24031255763784, \"nll_base\": 2.2235757116656685, \"eval_ref\": \"bigbench_crass_ai_generate_until__d2b5624c__s000__tb65536__v001\...
{ "index": "bigbench_crass_ai_generate_until__d2b5624c__s000__tb65536__v001", "uid": "bigbench_crass_ai_generate_until__d2b5624c__s000__tb65536__v001", "axis": "reasoning", "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\": \"bigbench_cryobiology_spanish_generate_until__35435087__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 15.83999554246162, \"nll_base\": 2.762538104975989, \"eval_ref\": \"bigbench_cryobiology_spanish_generate_until__354350...
{ "index": "bigbench_cryobiology_spanish_generate_until__35435087__s000__tb131072__v001", "uid": "bigbench_cryobiology_spanish_generate_until__35435087__s000__tb131072__v001", "axis": "reasoning", "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\": \"bigbench_cryobiology_spanish_generate_until__35435087__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 15.83999554246162, \"nll_base\": 2.762538104975989, \"eval_ref\": \"bigbench_cryobiology_spanish_generate_until__35435087...
{ "index": "bigbench_cryobiology_spanish_generate_until__35435087__s000__tb65536__v000", "uid": "bigbench_cryobiology_spanish_generate_until__35435087__s000__tb65536__v000", "axis": "reasoning", "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\": \"bigbench_dark_humor_detection_generate_until__15ca4666__s000__tb65536__v002\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 117.07821062395752, \"nll_base\": 4.762842178344727, \"eval_ref\": \"bigbench_dark_humor_detection_generate_until__15ca4...
{ "index": "bigbench_dark_humor_detection_generate_until__15ca4666__s000__tb65536__v002", "uid": "bigbench_dark_humor_detection_generate_until__15ca4666__s000__tb65536__v002", "axis": "reasoning", "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\": \"bigbench_dark_humor_detection_multiple_choice__befd4cf9__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 115.97477810544864, \"nll_base\": 4.753372737339565, \"eval_ref\": \"bigbench_dark_humor_detection_multiple_choice__bef...
{ "index": "bigbench_dark_humor_detection_multiple_choice__befd4cf9__s000__tb65536__v001", "uid": "bigbench_dark_humor_detection_multiple_choice__befd4cf9__s000__tb65536__v001", "axis": "reasoning", "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\": \"bigbench_date_understanding_multiple_choice__8a9ebddb__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3.5739807718636705, \"nll_base\": 1.273680036718195, \"eval_ref\": \"bigbench_date_understanding_multiple_choice__8a9eb...
{ "index": "bigbench_date_understanding_multiple_choice__8a9ebddb__s000__tb131072__v000", "uid": "bigbench_date_understanding_multiple_choice__8a9ebddb__s000__tb131072__v000", "axis": "reasoning", "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\": \"bigbench_disambiguation_qa_generate_until__acce9404__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 23.277543733611797, \"nll_base\": 3.1474891075721154, \"eval_ref\": \"bigbench_disambiguation_qa_generate_until__acce9404...
{ "index": "bigbench_disambiguation_qa_generate_until__acce9404__s000__tb262144__v000", "uid": "bigbench_disambiguation_qa_generate_until__acce9404__s000__tb262144__v000", "axis": "reasoning", "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\": \"bigbench_disambiguation_qa_generate_until__acce9404__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 23.277543733611797, \"nll_base\": 3.1474891075721154, \"eval_ref\": \"bigbench_disambiguation_qa_generate_until__acce9404__...
{ "index": "bigbench_disambiguation_qa_generate_until__acce9404__s000__tb65536__v001", "uid": "bigbench_disambiguation_qa_generate_until__acce9404__s000__tb65536__v001", "axis": "reasoning", "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\": \"bigbench_disambiguation_qa_multiple_choice__bd8216cd__s000__tb131072__v002\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 33.20294190596073, \"nll_base\": 3.502638483621988, \"eval_ref\": \"bigbench_disambiguation_qa_multiple_choice__bd8216cd...
{ "index": "bigbench_disambiguation_qa_multiple_choice__bd8216cd__s000__tb131072__v002", "uid": "bigbench_disambiguation_qa_multiple_choice__bd8216cd__s000__tb131072__v002", "axis": "reasoning", "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\": \"bigbench_disambiguation_qa_multiple_choice__bd8216cd__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 33.20294190596073, \"nll_base\": 3.502638483621988, \"eval_ref\": \"bigbench_disambiguation_qa_multiple_choice__bd8216cd...
{ "index": "bigbench_disambiguation_qa_multiple_choice__bd8216cd__s000__tb262144__v000", "uid": "bigbench_disambiguation_qa_multiple_choice__bd8216cd__s000__tb262144__v000", "axis": "reasoning", "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\": \"bigbench_disambiguation_qa_multiple_choice__bd8216cd__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 33.20294190596073, \"nll_base\": 3.502638483621988, \"eval_ref\": \"bigbench_disambiguation_qa_multiple_choice__bd8216cd__...
{ "index": "bigbench_disambiguation_qa_multiple_choice__bd8216cd__s000__tb65536__v001", "uid": "bigbench_disambiguation_qa_multiple_choice__bd8216cd__s000__tb65536__v001", "axis": "reasoning", "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\": \"bigbench_discourse_marker_prediction_multiple_choice__8b62c92f__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2.0415729348023155, \"nll_base\": 0.7137205572242281, \"eval_ref\": \"bigbench_discourse_marker_prediction_mul...
{ "index": "bigbench_discourse_marker_prediction_multiple_choice__8b62c92f__s000__tb131072__v000", "uid": "bigbench_discourse_marker_prediction_multiple_choice__8b62c92f__s000__tb131072__v000", "axis": "reasoning", "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\": \"bigbench_discourse_marker_prediction_multiple_choice__8b62c92f__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2.0415729348023155, \"nll_base\": 0.7137205572242281, \"eval_ref\": \"bigbench_discourse_marker_prediction_mul...
{ "index": "bigbench_discourse_marker_prediction_multiple_choice__8b62c92f__s000__tb262144__v001", "uid": "bigbench_discourse_marker_prediction_multiple_choice__8b62c92f__s000__tb262144__v001", "axis": "reasoning", "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\": \"bigbench_disfl_qa_generate_until__295ad810__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 99.1246915133236, \"nll_base\": 4.596378567852551, \"eval_ref\": \"bigbench_disfl_qa_generate_until__295ad810__s000__tb131072__v00...
{ "index": "bigbench_disfl_qa_generate_until__295ad810__s000__tb131072__v001", "uid": "bigbench_disfl_qa_generate_until__295ad810__s000__tb131072__v001", "axis": "reasoning", "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\": \"bigbench_disfl_qa_generate_until__295ad810__s000__tb262144__v002\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 99.1246915133236, \"nll_base\": 4.596378567852551, \"eval_ref\": \"bigbench_disfl_qa_generate_until__295ad810__s000__tb262144__v00...
{ "index": "bigbench_disfl_qa_generate_until__295ad810__s000__tb262144__v002", "uid": "bigbench_disfl_qa_generate_until__295ad810__s000__tb262144__v002", "axis": "reasoning", "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\": \"bigbench_disfl_qa_generate_until__295ad810__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 99.1246915133236, \"nll_base\": 4.596378567852551, \"eval_ref\": \"bigbench_disfl_qa_generate_until__295ad810__s000__tb65536__v000\"...
{ "index": "bigbench_disfl_qa_generate_until__295ad810__s000__tb65536__v000", "uid": "bigbench_disfl_qa_generate_until__295ad810__s000__tb65536__v000", "axis": "reasoning", "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\": \"bigbench_dyck_languages_multiple_choice__c3df556e__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 463.53221129603236, \"nll_base\": 6.13887587837551, \"eval_ref\": \"bigbench_dyck_languages_multiple_choice__c3df556e__s000__...
{ "index": "bigbench_dyck_languages_multiple_choice__c3df556e__s000__tb65536__v001", "uid": "bigbench_dyck_languages_multiple_choice__c3df556e__s000__tb65536__v001", "axis": "reasoning", "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\": \"bigbench_elementary_math_qa_generate_until__19db5397__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 55.1069998827056, \"nll_base\": 4.009276747703552, \"eval_ref\": \"bigbench_elementary_math_qa_generate_until__19db5397_...
{ "index": "bigbench_elementary_math_qa_generate_until__19db5397__s000__tb131072__v000", "uid": "bigbench_elementary_math_qa_generate_until__19db5397__s000__tb131072__v000", "axis": "reasoning", "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\": \"bigbench_elementary_math_qa_generate_until__19db5397__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 55.1069998827056, \"nll_base\": 4.009276747703552, \"eval_ref\": \"bigbench_elementary_math_qa_generate_until__19db5397_...
{ "index": "bigbench_elementary_math_qa_generate_until__19db5397__s000__tb262144__v001", "uid": "bigbench_elementary_math_qa_generate_until__19db5397__s000__tb262144__v001", "axis": "reasoning", "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\": \"bigbench_elementary_math_qa_multiple_choice__6b0a4656__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 38.672621510037786, \"nll_base\": 3.6551318951507112, \"eval_ref\": \"bigbench_elementary_math_qa_multiple_choice__6b0a...
{ "index": "bigbench_elementary_math_qa_multiple_choice__6b0a4656__s000__tb131072__v000", "uid": "bigbench_elementary_math_qa_multiple_choice__6b0a4656__s000__tb131072__v000", "axis": "reasoning", "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\": \"bigbench_elementary_math_qa_multiple_choice__6b0a4656__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 38.672621510037786, \"nll_base\": 3.6551318951507112, \"eval_ref\": \"bigbench_elementary_math_qa_multiple_choice__6b0a...
{ "index": "bigbench_elementary_math_qa_multiple_choice__6b0a4656__s000__tb262144__v001", "uid": "bigbench_elementary_math_qa_multiple_choice__6b0a4656__s000__tb262144__v001", "axis": "reasoning", "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\": \"bigbench_emoji_movie_generate_until__890e3b6b__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 104.2869183197833, \"nll_base\": 4.647145930552643, \"eval_ref\": \"bigbench_emoji_movie_generate_until__890e3b6b__s000__tb1310...
{ "index": "bigbench_emoji_movie_generate_until__890e3b6b__s000__tb131072__v000", "uid": "bigbench_emoji_movie_generate_until__890e3b6b__s000__tb131072__v000", "axis": "reasoning", "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\": \"bigbench_emoji_movie_generate_until__890e3b6b__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 104.2869183197833, \"nll_base\": 4.647145930552643, \"eval_ref\": \"bigbench_emoji_movie_generate_until__890e3b6b__s000__tb2621...
{ "index": "bigbench_emoji_movie_generate_until__890e3b6b__s000__tb262144__v001", "uid": "bigbench_emoji_movie_generate_until__890e3b6b__s000__tb262144__v001", "axis": "reasoning", "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\": \"bigbench_emoji_movie_multiple_choice__54ab336e__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 98.32854673565616, \"nll_base\": 4.58831438922242, \"eval_ref\": \"bigbench_emoji_movie_multiple_choice__54ab336e__s000__tb131...
{ "index": "bigbench_emoji_movie_multiple_choice__54ab336e__s000__tb131072__v000", "uid": "bigbench_emoji_movie_multiple_choice__54ab336e__s000__tb131072__v000", "axis": "reasoning", "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\": \"bigbench_emoji_movie_multiple_choice__54ab336e__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 98.32854673565616, \"nll_base\": 4.58831438922242, \"eval_ref\": \"bigbench_emoji_movie_multiple_choice__54ab336e__s000__tb262...
{ "index": "bigbench_emoji_movie_multiple_choice__54ab336e__s000__tb262144__v001", "uid": "bigbench_emoji_movie_multiple_choice__54ab336e__s000__tb262144__v001", "axis": "reasoning", "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\": \"bigbench_emoji_movie_multiple_choice__54ab336e__s000__tb65536__v002\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 98.32854673565616, \"nll_base\": 4.58831438922242, \"eval_ref\": \"bigbench_emoji_movie_multiple_choice__54ab336e__s000__tb65536...
{ "index": "bigbench_emoji_movie_multiple_choice__54ab336e__s000__tb65536__v002", "uid": "bigbench_emoji_movie_multiple_choice__54ab336e__s000__tb65536__v002", "axis": "reasoning", "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\": \"bigbench_emojis_emotion_prediction_generate_until__72b8f3dd__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3470.272187174015, \"nll_base\": 8.151988309972426, \"eval_ref\": \"bigbench_emojis_emotion_prediction_generate_u...
{ "index": "bigbench_emojis_emotion_prediction_generate_until__72b8f3dd__s000__tb131072__v000", "uid": "bigbench_emojis_emotion_prediction_generate_until__72b8f3dd__s000__tb131072__v000", "axis": "reasoning", "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\": \"bigbench_emojis_emotion_prediction_multiple_choice__81176ecb__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1153.8556605152983, \"nll_base\": 7.050864361702128, \"eval_ref\": \"bigbench_emojis_emotion_prediction_multiple...
{ "index": "bigbench_emojis_emotion_prediction_multiple_choice__81176ecb__s000__tb131072__v001", "uid": "bigbench_emojis_emotion_prediction_multiple_choice__81176ecb__s000__tb131072__v001", "axis": "reasoning", "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\": \"bigbench_emojis_emotion_prediction_multiple_choice__81176ecb__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1153.8556605152983, \"nll_base\": 7.050864361702128, \"eval_ref\": \"bigbench_emojis_emotion_prediction_multiple_c...
{ "index": "bigbench_emojis_emotion_prediction_multiple_choice__81176ecb__s000__tb65536__v000", "uid": "bigbench_emojis_emotion_prediction_multiple_choice__81176ecb__s000__tb65536__v000", "axis": "reasoning", "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\": \"bigbench_empirical_judgments_generate_until__053538ec__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 7449.6788376117565, \"nll_base\": 8.915926201398982, \"eval_ref\": \"bigbench_empirical_judgments_generate_until__05353...
{ "index": "bigbench_empirical_judgments_generate_until__053538ec__s000__tb131072__v001", "uid": "bigbench_empirical_judgments_generate_until__053538ec__s000__tb131072__v001", "axis": "reasoning", "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\": \"bigbench_empirical_judgments_generate_until__053538ec__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 7449.6788376117565, \"nll_base\": 8.915926201398982, \"eval_ref\": \"bigbench_empirical_judgments_generate_until__053538e...
{ "index": "bigbench_empirical_judgments_generate_until__053538ec__s000__tb65536__v000", "uid": "bigbench_empirical_judgments_generate_until__053538ec__s000__tb65536__v000", "axis": "reasoning", "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\": \"bigbench_entailed_polarity_generate_until__25c92012__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 9179.427304962192, \"nll_base\": 9.124720096588135, \"eval_ref\": \"bigbench_entailed_polarity_generate_until__25c92012__...
{ "index": "bigbench_entailed_polarity_generate_until__25c92012__s000__tb131072__v001", "uid": "bigbench_entailed_polarity_generate_until__25c92012__s000__tb131072__v001", "axis": "reasoning", "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\": \"bigbench_entailed_polarity_generate_until__25c92012__s000__tb262144__v002\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 9179.427304962192, \"nll_base\": 9.124720096588135, \"eval_ref\": \"bigbench_entailed_polarity_generate_until__25c92012__...
{ "index": "bigbench_entailed_polarity_generate_until__25c92012__s000__tb262144__v002", "uid": "bigbench_entailed_polarity_generate_until__25c92012__s000__tb262144__v002", "axis": "reasoning", "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\": \"bigbench_entailed_polarity_generate_until__25c92012__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 9179.427304962192, \"nll_base\": 9.124720096588135, \"eval_ref\": \"bigbench_entailed_polarity_generate_until__25c92012__s0...
{ "index": "bigbench_entailed_polarity_generate_until__25c92012__s000__tb65536__v000", "uid": "bigbench_entailed_polarity_generate_until__25c92012__s000__tb65536__v000", "axis": "reasoning", "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\": \"bigbench_entailed_polarity_hindi_generate_until__064f5b4c__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 34.8218588859427, \"nll_base\": 3.550245318496436, \"eval_ref\": \"bigbench_entailed_polarity_hindi_generate_until_...
{ "index": "bigbench_entailed_polarity_hindi_generate_until__064f5b4c__s000__tb131072__v000", "uid": "bigbench_entailed_polarity_hindi_generate_until__064f5b4c__s000__tb131072__v000", "axis": "reasoning", "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\": \"bigbench_entailed_polarity_hindi_generate_until__064f5b4c__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 34.8218588859427, \"nll_base\": 3.550245318496436, \"eval_ref\": \"bigbench_entailed_polarity_hindi_generate_until_...
{ "index": "bigbench_entailed_polarity_hindi_generate_until__064f5b4c__s000__tb262144__v001", "uid": "bigbench_entailed_polarity_hindi_generate_until__064f5b4c__s000__tb262144__v001", "axis": "reasoning", "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\": \"bigbench_entailed_polarity_hindi_generate_until__064f5b4c__s000__tb65536__v002\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 34.8218588859427, \"nll_base\": 3.550245318496436, \"eval_ref\": \"bigbench_entailed_polarity_hindi_generate_until__0...
{ "index": "bigbench_entailed_polarity_hindi_generate_until__064f5b4c__s000__tb65536__v002", "uid": "bigbench_entailed_polarity_hindi_generate_until__064f5b4c__s000__tb65536__v002", "axis": "reasoning", "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\": \"bigbench_entailed_polarity_hindi_multiple_choice__02152c73__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 30.799322402303645, \"nll_base\": 3.4274926898123206, \"eval_ref\": \"bigbench_entailed_polarity_hindi_multiple_ch...
{ "index": "bigbench_entailed_polarity_hindi_multiple_choice__02152c73__s000__tb262144__v000", "uid": "bigbench_entailed_polarity_hindi_multiple_choice__02152c73__s000__tb262144__v000", "axis": "reasoning", "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\": \"bigbench_entailed_polarity_hindi_multiple_choice__02152c73__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 30.799322402303645, \"nll_base\": 3.4274926898123206, \"eval_ref\": \"bigbench_entailed_polarity_hindi_multiple_choi...
{ "index": "bigbench_entailed_polarity_hindi_multiple_choice__02152c73__s000__tb65536__v001", "uid": "bigbench_entailed_polarity_hindi_multiple_choice__02152c73__s000__tb65536__v001", "axis": "reasoning", "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\": \"bigbench_entailed_polarity_multiple_choice__2ab4de8e__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 9054.336628227904, \"nll_base\": 9.11099910736084, \"eval_ref\": \"bigbench_entailed_polarity_multiple_choice__2ab4de8e_...
{ "index": "bigbench_entailed_polarity_multiple_choice__2ab4de8e__s000__tb262144__v000", "uid": "bigbench_entailed_polarity_multiple_choice__2ab4de8e__s000__tb262144__v000", "axis": "reasoning", "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\": \"bigbench_entailed_polarity_multiple_choice__2ab4de8e__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 9054.336628227904, \"nll_base\": 9.11099910736084, \"eval_ref\": \"bigbench_entailed_polarity_multiple_choice__2ab4de8e__s...
{ "index": "bigbench_entailed_polarity_multiple_choice__2ab4de8e__s000__tb65536__v001", "uid": "bigbench_entailed_polarity_multiple_choice__2ab4de8e__s000__tb65536__v001", "axis": "reasoning", "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\": \"bigbench_epistemic_reasoning_multiple_choice__d2b6eae7__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 21.492516946827255, \"nll_base\": 3.0677048255657327, \"eval_ref\": \"bigbench_epistemic_reasoning_multiple_choice__d2b6...
{ "index": "bigbench_epistemic_reasoning_multiple_choice__d2b6eae7__s000__tb65536__v000", "uid": "bigbench_epistemic_reasoning_multiple_choice__d2b6eae7__s000__tb65536__v000", "axis": "reasoning", "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\": \"bigbench_evaluating_information_essentiality_generate_until__30b755c0__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4.347386220735399, \"nll_base\": 1.4695747957101317, \"eval_ref\": \"bigbench_evaluating_information_es...
{ "index": "bigbench_evaluating_information_essentiality_generate_until__30b755c0__s000__tb131072__v000", "uid": "bigbench_evaluating_information_essentiality_generate_until__30b755c0__s000__tb131072__v000", "axis": "reasoning", "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\": \"bigbench_evaluating_information_essentiality_multiple_choice__32082ada__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4.287069544496869, \"nll_base\": 1.4556034096668764, \"eval_ref\": \"bigbench_evaluating_information_e...
{ "index": "bigbench_evaluating_information_essentiality_multiple_choice__32082ada__s000__tb131072__v000", "uid": "bigbench_evaluating_information_essentiality_multiple_choice__32082ada__s000__tb131072__v000", "axis": "reasoning", "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\": \"bigbench_evaluating_information_essentiality_multiple_choice__32082ada__s000__tb65536__v002\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4.287069544496869, \"nll_base\": 1.4556034096668764, \"eval_ref\": \"bigbench_evaluating_information_ess...
{ "index": "bigbench_evaluating_information_essentiality_multiple_choice__32082ada__s000__tb65536__v002", "uid": "bigbench_evaluating_information_essentiality_multiple_choice__32082ada__s000__tb65536__v002", "axis": "reasoning", "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\": \"bigbench_fact_checker_generate_until__e721a68d__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2834.3848048638706, \"nll_base\": 7.949580192565918, \"eval_ref\": \"bigbench_fact_checker_generate_until__e721a68d__s000__tb655...
{ "index": "bigbench_fact_checker_generate_until__e721a68d__s000__tb65536__v001", "uid": "bigbench_fact_checker_generate_until__e721a68d__s000__tb65536__v001", "axis": "reasoning", "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\": \"bigbench_fact_checker_multiple_choice__b696717b__s000__tb65536__v002\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2768.7289920417456, \"nll_base\": 7.926143646240234, \"eval_ref\": \"bigbench_fact_checker_multiple_choice__b696717b__s000__tb6...
{ "index": "bigbench_fact_checker_multiple_choice__b696717b__s000__tb65536__v002", "uid": "bigbench_fact_checker_multiple_choice__b696717b__s000__tb65536__v002", "axis": "reasoning", "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\": \"bigbench_few_shot_nlg_generate_until__810e8fdc__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 14.20732965336048, \"nll_base\": 2.653758004207739, \"eval_ref\": \"bigbench_few_shot_nlg_generate_until__810e8fdc__s000__tb26...
{ "index": "bigbench_few_shot_nlg_generate_until__810e8fdc__s000__tb262144__v000", "uid": "bigbench_few_shot_nlg_generate_until__810e8fdc__s000__tb262144__v000", "axis": "reasoning", "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\": \"bigbench_few_shot_nlg_generate_until__810e8fdc__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 14.20732965336048, \"nll_base\": 2.653758004207739, \"eval_ref\": \"bigbench_few_shot_nlg_generate_until__810e8fdc__s000__tb6553...
{ "index": "bigbench_few_shot_nlg_generate_until__810e8fdc__s000__tb65536__v001", "uid": "bigbench_few_shot_nlg_generate_until__810e8fdc__s000__tb65536__v001", "axis": "reasoning", "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\": \"bigbench_figure_of_speech_detection_generate_until__bea027a3__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 337.3211608049602, \"nll_base\": 5.821035475957961, \"eval_ref\": \"bigbench_figure_of_speech_detection_generate_u...
{ "index": "bigbench_figure_of_speech_detection_generate_until__bea027a3__s000__tb65536__v001", "uid": "bigbench_figure_of_speech_detection_generate_until__bea027a3__s000__tb65536__v001", "axis": "reasoning", "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\": \"bigbench_formal_fallacies_syllogisms_negation_generate_until__5eaf95ec__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 28177.491453797928, \"nll_base\": 10.246278762817383, \"eval_ref\": \"bigbench_formal_fallacies_syllog...
{ "index": "bigbench_formal_fallacies_syllogisms_negation_generate_until__5eaf95ec__s000__tb262144__v000", "uid": "bigbench_formal_fallacies_syllogisms_negation_generate_until__5eaf95ec__s000__tb262144__v000", "axis": "reasoning", "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\": \"bigbench_formal_fallacies_syllogisms_negation_generate_until__5eaf95ec__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 28177.491453797928, \"nll_base\": 10.246278762817383, \"eval_ref\": \"bigbench_formal_fallacies_syllogis...
{ "index": "bigbench_formal_fallacies_syllogisms_negation_generate_until__5eaf95ec__s000__tb65536__v001", "uid": "bigbench_formal_fallacies_syllogisms_negation_generate_until__5eaf95ec__s000__tb65536__v001", "axis": "reasoning", "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\": \"bigbench_formal_fallacies_syllogisms_negation_multiple_choice__510f2401__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 24519.733573805777, \"nll_base\": 10.10723352432251, \"eval_ref\": \"bigbench_formal_fallacies_syllog...
{ "index": "bigbench_formal_fallacies_syllogisms_negation_multiple_choice__510f2401__s000__tb262144__v000", "uid": "bigbench_formal_fallacies_syllogisms_negation_multiple_choice__510f2401__s000__tb262144__v000", "axis": "reasoning", "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\": \"bigbench_formal_fallacies_syllogisms_negation_multiple_choice__510f2401__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 24519.733573805777, \"nll_base\": 10.10723352432251, \"eval_ref\": \"bigbench_formal_fallacies_syllogis...
{ "index": "bigbench_formal_fallacies_syllogisms_negation_multiple_choice__510f2401__s000__tb65536__v001", "uid": "bigbench_formal_fallacies_syllogisms_negation_multiple_choice__510f2401__s000__tb65536__v001", "axis": "reasoning", "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\": \"bigbench_gender_inclusive_sentences_german_generate_until__77393b30__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2.253055247157267, \"nll_base\": 0.8122871827461143, \"eval_ref\": \"bigbench_gender_inclusive_sentences_...
{ "index": "bigbench_gender_inclusive_sentences_german_generate_until__77393b30__s000__tb262144__v000", "uid": "bigbench_gender_inclusive_sentences_german_generate_until__77393b30__s000__tb262144__v000", "axis": "reasoning", "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\": \"bigbench_gender_inclusive_sentences_german_generate_until__77393b30__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2.253055247157267, \"nll_base\": 0.8122871827461143, \"eval_ref\": \"bigbench_gender_inclusive_sentences_ge...
{ "index": "bigbench_gender_inclusive_sentences_german_generate_until__77393b30__s000__tb65536__v001", "uid": "bigbench_gender_inclusive_sentences_german_generate_until__77393b30__s000__tb65536__v001", "axis": "reasoning", "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\": \"bigbench_general_knowledge_multiple_choice__29420800__s000__tb131072__v002\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1759.0900239270036, \"nll_base\": 7.472551922465479, \"eval_ref\": \"bigbench_general_knowledge_multiple_choice__2942080...
{ "index": "bigbench_general_knowledge_multiple_choice__29420800__s000__tb131072__v002", "uid": "bigbench_general_knowledge_multiple_choice__29420800__s000__tb131072__v002", "axis": "reasoning", "budget": 2048, "token_budget": 131072, "variant": "filter" }