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_understanding_fables_multiple_choice__144360ba__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 8.028812990752645, \"nll_base\": 2.0830366952078685, \"eval_ref\": \"bigbench_understanding_fables_multiple_choice__144... | {
"index": "bigbench_understanding_fables_multiple_choice__144360ba__s000__tb65536__v000",
"uid": "bigbench_understanding_fables_multiple_choice__144360ba__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_undo_permutation_multiple_choice__13a6ce0f__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1.7802121519279035, \"nll_base\": 0.5767325436782003, \"eval_ref\": \"bigbench_undo_permutation_multiple_choice__13a6ce0f... | {
"index": "bigbench_undo_permutation_multiple_choice__13a6ce0f__s000__tb131072__v000",
"uid": "bigbench_undo_permutation_multiple_choice__13a6ce0f__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_undo_permutation_multiple_choice__13a6ce0f__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1.7802121519279035, \"nll_base\": 0.5767325436782003, \"eval_ref\": \"bigbench_undo_permutation_multiple_choice__13a6ce0f... | {
"index": "bigbench_undo_permutation_multiple_choice__13a6ce0f__s000__tb262144__v001",
"uid": "bigbench_undo_permutation_multiple_choice__13a6ce0f__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_unit_interpretation_generate_until__e164786c__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 84.10280486915953, \"nll_base\": 4.432039918017988, \"eval_ref\": \"bigbench_unit_interpretation_generate_until__e16478... | {
"index": "bigbench_unit_interpretation_generate_until__e164786c__s000__tb131072__v000",
"uid": "bigbench_unit_interpretation_generate_until__e164786c__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_unit_interpretation_generate_until__e164786c__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 84.10280486915953, \"nll_base\": 4.432039918017988, \"eval_ref\": \"bigbench_unit_interpretation_generate_until__e16478... | {
"index": "bigbench_unit_interpretation_generate_until__e164786c__s000__tb262144__v001",
"uid": "bigbench_unit_interpretation_generate_until__e164786c__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_unit_interpretation_generate_until__e164786c__s000__tb65536__v002\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 84.10280486915953, \"nll_base\": 4.432039918017988, \"eval_ref\": \"bigbench_unit_interpretation_generate_until__e164786c... | {
"index": "bigbench_unit_interpretation_generate_until__e164786c__s000__tb65536__v002",
"uid": "bigbench_unit_interpretation_generate_until__e164786c__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_unit_interpretation_multiple_choice__729d2b39__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 97.53171808974285, \"nll_base\": 4.580177638841712, \"eval_ref\": \"bigbench_unit_interpretation_multiple_choice__729d... | {
"index": "bigbench_unit_interpretation_multiple_choice__729d2b39__s000__tb131072__v001",
"uid": "bigbench_unit_interpretation_multiple_choice__729d2b39__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_unit_interpretation_multiple_choice__729d2b39__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 97.53171808974285, \"nll_base\": 4.580177638841712, \"eval_ref\": \"bigbench_unit_interpretation_multiple_choice__729d2b... | {
"index": "bigbench_unit_interpretation_multiple_choice__729d2b39__s000__tb65536__v000",
"uid": "bigbench_unit_interpretation_multiple_choice__729d2b39__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_unnatural_in_context_learning_generate_until__3b863f86__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 8.30607155936185, \"nll_base\": 2.116986760553324, \"eval_ref\": \"bigbench_unnatural_in_context_learning_gener... | {
"index": "bigbench_unnatural_in_context_learning_generate_until__3b863f86__s000__tb65536__v001",
"uid": "bigbench_unnatural_in_context_learning_generate_until__3b863f86__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_vitaminc_fact_verification_generate_until__52b62c19__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2616.696718702814, \"nll_base\": 7.869668006896973, \"eval_ref\": \"bigbench_vitaminc_fact_verification_generate... | {
"index": "bigbench_vitaminc_fact_verification_generate_until__52b62c19__s000__tb262144__v000",
"uid": "bigbench_vitaminc_fact_verification_generate_until__52b62c19__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_vitaminc_fact_verification_generate_until__52b62c19__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2616.696718702814, \"nll_base\": 7.869668006896973, \"eval_ref\": \"bigbench_vitaminc_fact_verification_generate_u... | {
"index": "bigbench_vitaminc_fact_verification_generate_until__52b62c19__s000__tb65536__v001",
"uid": "bigbench_vitaminc_fact_verification_generate_until__52b62c19__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_vitaminc_fact_verification_multiple_choice__72ed05a7__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2358.355073334525, \"nll_base\": 7.765719652175903, \"eval_ref\": \"bigbench_vitaminc_fact_verification_multiple_... | {
"index": "bigbench_vitaminc_fact_verification_multiple_choice__72ed05a7__s000__tb65536__v000",
"uid": "bigbench_vitaminc_fact_verification_multiple_choice__72ed05a7__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_what_is_the_tao_multiple_choice__376c6e8e__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 18.104626522014566, \"nll_base\": 2.896167514560459, \"eval_ref\": \"bigbench_what_is_the_tao_multiple_choice__376c6e8e__s... | {
"index": "bigbench_what_is_the_tao_multiple_choice__376c6e8e__s000__tb131072__v000",
"uid": "bigbench_what_is_the_tao_multiple_choice__376c6e8e__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_what_is_the_tao_multiple_choice__376c6e8e__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 18.104626522014566, \"nll_base\": 2.896167514560459, \"eval_ref\": \"bigbench_what_is_the_tao_multiple_choice__376c6e8e__s... | {
"index": "bigbench_what_is_the_tao_multiple_choice__376c6e8e__s000__tb262144__v001",
"uid": "bigbench_what_is_the_tao_multiple_choice__376c6e8e__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_what_is_the_tao_multiple_choice__376c6e8e__s000__tb65536__v002\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 18.104626522014566, \"nll_base\": 2.896167514560459, \"eval_ref\": \"bigbench_what_is_the_tao_multiple_choice__376c6e8e__s00... | {
"index": "bigbench_what_is_the_tao_multiple_choice__376c6e8e__s000__tb65536__v002",
"uid": "bigbench_what_is_the_tao_multiple_choice__376c6e8e__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_which_wiki_edit_multiple_choice__7c7c2bb3__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2.5878341201380257, \"nll_base\": 0.9508212788108719, \"eval_ref\": \"bigbench_which_wiki_edit_multiple_choice__7c7c2bb3__... | {
"index": "bigbench_which_wiki_edit_multiple_choice__7c7c2bb3__s000__tb131072__v001",
"uid": "bigbench_which_wiki_edit_multiple_choice__7c7c2bb3__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_winowhy_generate_until__bb76fb45__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4756.064567779753, \"nll_base\": 8.467175833786591, \"eval_ref\": \"bigbench_winowhy_generate_until__bb76fb45__s000__tb131072__v001... | {
"index": "bigbench_winowhy_generate_until__bb76fb45__s000__tb131072__v001",
"uid": "bigbench_winowhy_generate_until__bb76fb45__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_winowhy_generate_until__bb76fb45__s000__tb262144__v002\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 4756.064567779753, \"nll_base\": 8.467175833786591, \"eval_ref\": \"bigbench_winowhy_generate_until__bb76fb45__s000__tb262144__v002... | {
"index": "bigbench_winowhy_generate_until__bb76fb45__s000__tb262144__v002",
"uid": "bigbench_winowhy_generate_until__bb76fb45__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_winowhy_generate_until__bb76fb45__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4756.064567779753, \"nll_base\": 8.467175833786591, \"eval_ref\": \"bigbench_winowhy_generate_until__bb76fb45__s000__tb65536__v000\",... | {
"index": "bigbench_winowhy_generate_until__bb76fb45__s000__tb65536__v000",
"uid": "bigbench_winowhy_generate_until__bb76fb45__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_winowhy_multiple_choice__4c3763ca__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 585.9424974201787, \"nll_base\": 6.3732216574928975, \"eval_ref\": \"bigbench_winowhy_multiple_choice__4c3763ca__s000__tb65536__v001... | {
"index": "bigbench_winowhy_multiple_choice__4c3763ca__s000__tb65536__v001",
"uid": "bigbench_winowhy_multiple_choice__4c3763ca__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_word_unscrambling_generate_until__5f7a0c24__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 22162.607648969202, \"nll_base\": 10.00616180762816, \"eval_ref\": \"bigbench_word_unscrambling_generate_until__5f7a0c24_... | {
"index": "bigbench_word_unscrambling_generate_until__5f7a0c24__s000__tb131072__v000",
"uid": "bigbench_word_unscrambling_generate_until__5f7a0c24__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_word_unscrambling_generate_until__5f7a0c24__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 22162.607648969202, \"nll_base\": 10.00616180762816, \"eval_ref\": \"bigbench_word_unscrambling_generate_until__5f7a0c24_... | {
"index": "bigbench_word_unscrambling_generate_until__5f7a0c24__s000__tb262144__v001",
"uid": "bigbench_word_unscrambling_generate_until__5f7a0c24__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\": \"blimp_adjunct_island__57b04c48__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4.1317650949866405, \"nll_base\": 1.4187046994445145, \"eval_ref\": \"blimp_adjunct_island__57b04c48__s000__tb131072__v000\", \"oracle_delta\"... | {
"index": "blimp_adjunct_island__57b04c48__s000__tb131072__v000",
"uid": "blimp_adjunct_island__57b04c48__s000__tb131072__v000",
"axis": "instruction",
"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\": \"blimp_anaphor_gender_agreement__c9b9d0fe__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 8.384231951420091, \"nll_base\": 2.1263527931573565, \"eval_ref\": \"blimp_anaphor_gender_agreement__c9b9d0fe__s000__tb131072__v000\... | {
"index": "blimp_anaphor_gender_agreement__c9b9d0fe__s000__tb131072__v000",
"uid": "blimp_anaphor_gender_agreement__c9b9d0fe__s000__tb131072__v000",
"axis": "instruction",
"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\": \"blimp_animate_subject_passive__f061cee6__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4.609181008001606, \"nll_base\": 1.528050185694839, \"eval_ref\": \"blimp_animate_subject_passive__f061cee6__s000__tb131072__v000\", ... | {
"index": "blimp_animate_subject_passive__f061cee6__s000__tb131072__v000",
"uid": "blimp_animate_subject_passive__f061cee6__s000__tb131072__v000",
"axis": "instruction",
"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\": \"blimp_causative__e5aa3232__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 6.419367355654049, \"nll_base\": 1.8593195701132015, \"eval_ref\": \"blimp_causative__e5aa3232__s000__tb131072__v000\", \"oracle_delta\": 0.2764325... | {
"index": "blimp_causative__e5aa3232__s000__tb131072__v000",
"uid": "blimp_causative__e5aa3232__s000__tb131072__v000",
"axis": "instruction",
"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\": \"blimp_determiner_noun_agreement_irregular_1__f1ea4e44__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 5.714620425510921, \"nll_base\": 1.7430278778076171, \"eval_ref\": \"blimp_determiner_noun_agreement_irregular_1__f1ea4e4... | {
"index": "blimp_determiner_noun_agreement_irregular_1__f1ea4e44__s000__tb65536__v000",
"uid": "blimp_determiner_noun_agreement_irregular_1__f1ea4e44__s000__tb65536__v000",
"axis": "instruction",
"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\": \"blimp_determiner_noun_agreement_irregular_2__c87da3b0__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 5.707946514668075, \"nll_base\": 1.7418593293303377, \"eval_ref\": \"blimp_determiner_noun_agreement_irregular_2__c87da3b... | {
"index": "blimp_determiner_noun_agreement_irregular_2__c87da3b0__s000__tb65536__v000",
"uid": "blimp_determiner_noun_agreement_irregular_2__c87da3b0__s000__tb65536__v000",
"axis": "instruction",
"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\": \"blimp_determiner_noun_agreement_with_adj_2__af23f893__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4.731308986327667, \"nll_base\": 1.5542019055135226, \"eval_ref\": \"blimp_determiner_noun_agreement_with_adj_2__af23f893_... | {
"index": "blimp_determiner_noun_agreement_with_adj_2__af23f893__s000__tb65536__v000",
"uid": "blimp_determiner_noun_agreement_with_adj_2__af23f893__s000__tb65536__v000",
"axis": "instruction",
"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\": \"blimp_determiner_noun_agreement_with_adj_irregular_2__ce0c91c6__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4.508385992355374, \"nll_base\": 1.5059392163707952, \"eval_ref\": \"blimp_determiner_noun_agreement_with_adj_... | {
"index": "blimp_determiner_noun_agreement_with_adj_irregular_2__ce0c91c6__s000__tb131072__v000",
"uid": "blimp_determiner_noun_agreement_with_adj_irregular_2__ce0c91c6__s000__tb131072__v000",
"axis": "instruction",
"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\": \"blimp_distractor_agreement_relative_clause__bf30ebea__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2.890850096044115, \"nll_base\": 1.061550609750769, \"eval_ref\": \"blimp_distractor_agreement_relative_clause__bf30ebea... | {
"index": "blimp_distractor_agreement_relative_clause__bf30ebea__s000__tb131072__v000",
"uid": "blimp_distractor_agreement_relative_clause__bf30ebea__s000__tb131072__v000",
"axis": "instruction",
"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\": \"blimp_existential_there_quantifiers_2__7b059abc__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4.2771903874356525, \"nll_base\": 1.4532963425013303, \"eval_ref\": \"blimp_existential_there_quantifiers_2__7b059abc__s000__tb... | {
"index": "blimp_existential_there_quantifiers_2__7b059abc__s000__tb65536__v000",
"uid": "blimp_existential_there_quantifiers_2__7b059abc__s000__tb65536__v000",
"axis": "instruction",
"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\": \"blimp_existential_there_subject_raising__4b0ff8e7__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3.114469621331586, \"nll_base\": 1.136058871717338, \"eval_ref\": \"blimp_existential_there_subject_raising__4b0ff8e7__s000... | {
"index": "blimp_existential_there_subject_raising__4b0ff8e7__s000__tb131072__v000",
"uid": "blimp_existential_there_subject_raising__4b0ff8e7__s000__tb131072__v000",
"axis": "instruction",
"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\": \"blimp_inchoative__f7588cd4__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 11.764499932614417, \"nll_base\": 2.465086516610819, \"eval_ref\": \"blimp_inchoative__f7588cd4__s000__tb262144__v000\", \"oracle_delta\": 0.61735... | {
"index": "blimp_inchoative__f7588cd4__s000__tb262144__v000",
"uid": "blimp_inchoative__f7588cd4__s000__tb262144__v000",
"axis": "instruction",
"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\": \"blimp_intransitive__769ba1bd__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 9.903848066255104, \"nll_base\": 2.292923375179893, \"eval_ref\": \"blimp_intransitive__769ba1bd__s000__tb65536__v000\", \"oracle_delta\": 0.26572... | {
"index": "blimp_intransitive__769ba1bd__s000__tb65536__v000",
"uid": "blimp_intransitive__769ba1bd__s000__tb65536__v000",
"axis": "instruction",
"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\": \"blimp_irregular_past_participle_verbs__a4ce0fc8__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 9.839613389130461, \"nll_base\": 2.2864164205697866, \"eval_ref\": \"blimp_irregular_past_participle_verbs__a4ce0fc8__s000__tb6... | {
"index": "blimp_irregular_past_participle_verbs__a4ce0fc8__s000__tb65536__v000",
"uid": "blimp_irregular_past_participle_verbs__a4ce0fc8__s000__tb65536__v000",
"axis": "instruction",
"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\": \"blimp_left_branch_island_echo_question__7e3cffb8__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 7.203595085253112, \"nll_base\": 1.9745802188008248, \"eval_ref\": \"blimp_left_branch_island_echo_question__7e3cffb8__s000__t... | {
"index": "blimp_left_branch_island_echo_question__7e3cffb8__s000__tb65536__v000",
"uid": "blimp_left_branch_island_echo_question__7e3cffb8__s000__tb65536__v000",
"axis": "instruction",
"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\": \"blimp_matrix_question_npi_licensor_present__ee2bdf0d__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 7.089567474536012, \"nll_base\": 1.9586243336811824, \"eval_ref\": \"blimp_matrix_question_npi_licensor_present__ee2bdf0... | {
"index": "blimp_matrix_question_npi_licensor_present__ee2bdf0d__s000__tb131072__v000",
"uid": "blimp_matrix_question_npi_licensor_present__ee2bdf0d__s000__tb131072__v000",
"axis": "instruction",
"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\": \"blimp_nl__argument_structure__ditransitive_nomdat_2__ded350a3__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2.2715282951857865, \"nll_base\": 0.8204528627480526, \"eval_ref\": \"blimp_nl__argument_structure__ditransitiv... | {
"index": "blimp_nl__argument_structure__ditransitive_nomdat_2__ded350a3__s000__tb131072__v000",
"uid": "blimp_nl__argument_structure__ditransitive_nomdat_2__ded350a3__s000__tb131072__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"blimp_nl__auxiliaries__order_2__c308d06b__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1.943949519244785, \"nll_base\": 0.6647217382344125, \"eval_ref\": \"blimp_nl__auxiliaries__order_2__c308d06b__s000__tb131072__v000\... | {
"index": "blimp_nl__auxiliaries__order_2__c308d06b__s000__tb131072__v000",
"uid": "blimp_nl__auxiliaries__order_2__c308d06b__s000__tb131072__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"blimp_nl__complementive__ditransitive__2266ab76__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2.011843696617801, \"nll_base\": 0.6990515636421784, \"eval_ref\": \"blimp_nl__complementive__ditransitive__2266ab76__s000__tb6... | {
"index": "blimp_nl__complementive__ditransitive__2266ab76__s000__tb65536__v000",
"uid": "blimp_nl__complementive__ditransitive__2266ab76__s000__tb65536__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"blimp_nl__complementive__position_adverb__d7f5ae8e__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1.87298951204959, \"nll_base\": 0.6275358238998725, \"eval_ref\": \"blimp_nl__complementive__position_adverb__d7f5ae8e__s0... | {
"index": "blimp_nl__complementive__position_adverb__d7f5ae8e__s000__tb131072__v000",
"uid": "blimp_nl__complementive__position_adverb__d7f5ae8e__s000__tb131072__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"blimp_nl__complementive__position_verb__fb31bff4__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1.8768852703921537, \"nll_base\": 0.6296136318108975, \"eval_ref\": \"blimp_nl__complementive__position_verb__fb31bff4__s000... | {
"index": "blimp_nl__complementive__position_verb__fb31bff4__s000__tb131072__v000",
"uid": "blimp_nl__complementive__position_verb__fb31bff4__s000__tb131072__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"blimp_nl__infinitival_argument_clause__bare_verb_cluster__78e7d04e__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2.175306589377293, \"nll_base\": 0.777169615239869, \"eval_ref\": \"blimp_nl__infinitival_argument_clause_... | {
"index": "blimp_nl__infinitival_argument_clause__bare_verb_cluster__78e7d04e__s000__tb131072__v000",
"uid": "blimp_nl__infinitival_argument_clause__bare_verb_cluster__78e7d04e__s000__tb131072__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"blimp_nl__wh_movement__hierarchy__d20df4e3__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1.375993167187484, \"nll_base\": 0.31917577379205253, \"eval_ref\": \"blimp_nl__wh_movement__hierarchy__d20df4e3__s000__tb65536__v00... | {
"index": "blimp_nl__wh_movement__hierarchy__d20df4e3__s000__tb65536__v000",
"uid": "blimp_nl__wh_movement__hierarchy__d20df4e3__s000__tb65536__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"blimp_nl__wh_movement_restrictions__bridge_verb_2__e3c32965__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2.665689152998116, \"nll_base\": 0.9804626181836843, \"eval_ref\": \"blimp_nl__wh_movement_restrictions__bridge_ver... | {
"index": "blimp_nl__wh_movement_restrictions__bridge_verb_2__e3c32965__s000__tb65536__v000",
"uid": "blimp_nl__wh_movement_restrictions__bridge_verb_2__e3c32965__s000__tb65536__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"blimp_npi_present_1__dbc105f9__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 5.279075962927378, \"nll_base\": 1.6637510753813245, \"eval_ref\": \"blimp_npi_present_1__dbc105f9__s000__tb131072__v000\", \"oracle_delta\": 0... | {
"index": "blimp_npi_present_1__dbc105f9__s000__tb131072__v000",
"uid": "blimp_npi_present_1__dbc105f9__s000__tb131072__v000",
"axis": "instruction",
"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\": \"blimp_passive_2__212d9e4b__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 8.064458484912246, \"nll_base\": 2.0874665654806726, \"eval_ref\": \"blimp_passive_2__212d9e4b__s000__tb65536__v000\", \"oracle_delta\": 0.2002205599... | {
"index": "blimp_passive_2__212d9e4b__s000__tb65536__v000",
"uid": "blimp_passive_2__212d9e4b__s000__tb65536__v000",
"axis": "instruction",
"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\": \"blimp_principle_A_case_2__3b2c7ff0__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4.0866417744580685, \"nll_base\": 1.407723550742629, \"eval_ref\": \"blimp_principle_A_case_2__3b2c7ff0__s000__tb65536__v000\", \"oracle_del... | {
"index": "blimp_principle_A_case_2__3b2c7ff0__s000__tb65536__v000",
"uid": "blimp_principle_A_case_2__3b2c7ff0__s000__tb65536__v000",
"axis": "instruction",
"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\": \"blimp_principle_A_domain_3__f0b4ce65__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4.693015840746115, \"nll_base\": 1.5460754122052873, \"eval_ref\": \"blimp_principle_A_domain_3__f0b4ce65__s000__tb65536__v000\", \"oracle... | {
"index": "blimp_principle_A_domain_3__f0b4ce65__s000__tb65536__v000",
"uid": "blimp_principle_A_domain_3__f0b4ce65__s000__tb65536__v000",
"axis": "instruction",
"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\": \"blimp_regular_plural_subject_verb_agreement_1__3a060261__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 6.808845260092351, \"nll_base\": 1.9182225404520752, \"eval_ref\": \"blimp_regular_plural_subject_verb_agreement_1__3a0... | {
"index": "blimp_regular_plural_subject_verb_agreement_1__3a060261__s000__tb65536__v000",
"uid": "blimp_regular_plural_subject_verb_agreement_1__3a060261__s000__tb65536__v000",
"axis": "instruction",
"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\": \"blimp_sentential_negation_npi_scope__8cd9a856__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2.772581724295938, \"nll_base\": 1.0197789165603213, \"eval_ref\": \"blimp_sentential_negation_npi_scope__8cd9a856__s000__tb131... | {
"index": "blimp_sentential_negation_npi_scope__8cd9a856__s000__tb131072__v000",
"uid": "blimp_sentential_negation_npi_scope__8cd9a856__s000__tb131072__v000",
"axis": "instruction",
"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\": \"blimp_wh_island__873b13ed__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4.6694084023054145, \"nll_base\": 1.541032383207641, \"eval_ref\": \"blimp_wh_island__873b13ed__s000__tb65536__v000\", \"oracle_delta\": 0.1266173434... | {
"index": "blimp_wh_island__873b13ed__s000__tb65536__v000",
"uid": "blimp_wh_island__873b13ed__s000__tb65536__v000",
"axis": "instruction",
"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\": \"blimp_wh_questions_object_gap__af085ca6__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3.8420309160453003, \"nll_base\": 1.3460011111763786, \"eval_ref\": \"blimp_wh_questions_object_gap__af085ca6__s000__tb131072__v000\"... | {
"index": "blimp_wh_questions_object_gap__af085ca6__s000__tb131072__v000",
"uid": "blimp_wh_questions_object_gap__af085ca6__s000__tb131072__v000",
"axis": "instruction",
"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\": \"blimp_wh_questions_subject_gap_long_distance__4d029f4e__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2.547718313082224, \"nll_base\": 0.9351981794330436, \"eval_ref\": \"blimp_wh_questions_subject_gap_long_distance__4d0... | {
"index": "blimp_wh_questions_subject_gap_long_distance__4d029f4e__s000__tb131072__v000",
"uid": "blimp_wh_questions_subject_gap_long_distance__4d029f4e__s000__tb131072__v000",
"axis": "instruction",
"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\": \"blimp_wh_vs_that_no_gap__9cde8b2d__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3.214637882090123, \"nll_base\": 1.1677147174703664, \"eval_ref\": \"blimp_wh_vs_that_no_gap__9cde8b2d__s000__tb65536__v000\", \"oracle_delta... | {
"index": "blimp_wh_vs_that_no_gap__9cde8b2d__s000__tb65536__v000",
"uid": "blimp_wh_vs_that_no_gap__9cde8b2d__s000__tb65536__v000",
"axis": "instruction",
"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\": \"blimp_wh_vs_that_no_gap_long_distance__1908577f__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2.4708329033485357, \"nll_base\": 0.9045553016315175, \"eval_ref\": \"blimp_wh_vs_that_no_gap_long_distance__1908577f__s000__... | {
"index": "blimp_wh_vs_that_no_gap_long_distance__1908577f__s000__tb131072__v000",
"uid": "blimp_wh_vs_that_no_gap_long_distance__1908577f__s000__tb131072__v000",
"axis": "instruction",
"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\": \"blimp_wh_vs_that_with_gap__62676693__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4.429477652789772, \"nll_base\": 1.488281665770819, \"eval_ref\": \"blimp_wh_vs_that_with_gap__62676693__s000__tb131072__v000\", \"oracle... | {
"index": "blimp_wh_vs_that_with_gap__62676693__s000__tb131072__v000",
"uid": "blimp_wh_vs_that_with_gap__62676693__s000__tb131072__v000",
"axis": "instruction",
"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\": \"blimp_wh_vs_that_with_gap_long_distance__6285dd63__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2.5072570272309918, \"nll_base\": 0.9191893377466328, \"eval_ref\": \"blimp_wh_vs_that_with_gap_long_distance__6285dd63__s000... | {
"index": "blimp_wh_vs_that_with_gap_long_distance__6285dd63__s000__tb65536__v000",
"uid": "blimp_wh_vs_that_with_gap_long_distance__6285dd63__s000__tb65536__v000",
"axis": "instruction",
"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\": \"boolq-seq2seq__4c178464__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 29322.149945068875, \"nll_base\": 10.28609848022461, \"eval_ref\": \"boolq-seq2seq__4c178464__s000__tb131072__v001\", \"oracle_delta\": 0.61736917495... | {
"index": "boolq-seq2seq__4c178464__s000__tb131072__v001",
"uid": "boolq-seq2seq__4c178464__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\": \"boolq-seq2seq__4c178464__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 29322.149945068875, \"nll_base\": 10.28609848022461, \"eval_ref\": \"boolq-seq2seq__4c178464__s000__tb65536__v000\", \"oracle_delta\": 0.15045118331909... | {
"index": "boolq-seq2seq__4c178464__s000__tb65536__v000",
"uid": "boolq-seq2seq__4c178464__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\": \"boolq__985d28c6__s000__tb131072__v002\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1692.923521571609, \"nll_base\": 7.4342122077941895, \"eval_ref\": \"boolq__985d28c6__s000__tb131072__v002\", \"oracle_delta\": 0.507042407989502, \"random_d... | {
"index": "boolq__985d28c6__s000__tb131072__v002",
"uid": "boolq__985d28c6__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\": \"boolq__985d28c6__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1692.923521571609, \"nll_base\": 7.4342122077941895, \"eval_ref\": \"boolq__985d28c6__s000__tb262144__v000\", \"oracle_delta\": 1.3581058979034424, \"random_... | {
"index": "boolq__985d28c6__s000__tb262144__v000",
"uid": "boolq__985d28c6__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\": \"boolq__985d28c6__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1692.923521571609, \"nll_base\": 7.4342122077941895, \"eval_ref\": \"boolq__985d28c6__s000__tb65536__v001\", \"oracle_delta\": 0.14672636985778809, \"random_de... | {
"index": "boolq__985d28c6__s000__tb65536__v001",
"uid": "boolq__985d28c6__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\": \"careqa_en__81fa66b8__s000__tb131072__v002\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1492.773524105188, \"nll_base\": 7.308391094207764, \"eval_ref\": \"careqa_en__81fa66b8__s000__tb131072__v002\", \"oracle_delta\": 0.5338902473449707, \"... | {
"index": "careqa_en__81fa66b8__s000__tb131072__v002",
"uid": "careqa_en__81fa66b8__s000__tb131072__v002",
"axis": "domain",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"careqa_en__81fa66b8__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1492.773524105188, \"nll_base\": 7.308391094207764, \"eval_ref\": \"careqa_en__81fa66b8__s000__tb65536__v001\", \"oracle_delta\": 0.24828243255615234, \"ra... | {
"index": "careqa_en__81fa66b8__s000__tb65536__v001",
"uid": "careqa_en__81fa66b8__s000__tb65536__v001",
"axis": "domain",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"careqa_es__94d03ec7__s000__tb131072__v002\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1810.7265279860826, \"nll_base\": 7.50148344039917, \"eval_ref\": \"careqa_es__94d03ec7__s000__tb131072__v002\", \"oracle_delta\": 1.8831665515899658, \"... | {
"index": "careqa_es__94d03ec7__s000__tb131072__v002",
"uid": "careqa_es__94d03ec7__s000__tb131072__v002",
"axis": "domain",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"careqa_es__94d03ec7__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1810.7265279860826, \"nll_base\": 7.50148344039917, \"eval_ref\": \"careqa_es__94d03ec7__s000__tb262144__v000\", \"oracle_delta\": 2.7567038536071777, \"... | {
"index": "careqa_es__94d03ec7__s000__tb262144__v000",
"uid": "careqa_es__94d03ec7__s000__tb262144__v000",
"axis": "domain",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"careqa_es__94d03ec7__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1810.7265279860826, \"nll_base\": 7.50148344039917, \"eval_ref\": \"careqa_es__94d03ec7__s000__tb65536__v001\", \"oracle_delta\": 0.9818828105926514, \"ran... | {
"index": "careqa_es__94d03ec7__s000__tb65536__v001",
"uid": "careqa_es__94d03ec7__s000__tb65536__v001",
"axis": "domain",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"careqa_es__94d03ec7__s001__tb262144__v003\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1318.4815989561769, \"nll_base\": 7.1842360496521, \"eval_ref\": \"careqa_es__94d03ec7__s001__tb262144__v003\", \"oracle_delta\": 2.2188913822174072, \"r... | {
"index": "careqa_es__94d03ec7__s001__tb262144__v003",
"uid": "careqa_es__94d03ec7__s001__tb262144__v003",
"axis": "domain",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"cb__85d56cb4__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2333.4465993169956, \"nll_base\": 7.755101680755615, \"eval_ref\": \"cb__85d56cb4__s000__tb262144__v000\", \"oracle_delta\": 1.958237886428833, \"random_delta\"... | {
"index": "cb__85d56cb4__s000__tb262144__v000",
"uid": "cb__85d56cb4__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\": \"cb__85d56cb4__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2333.4465993169956, \"nll_base\": 7.755101680755615, \"eval_ref\": \"cb__85d56cb4__s000__tb65536__v001\", \"oracle_delta\": 0.617377519607544, \"random_delta\": 0... | {
"index": "cb__85d56cb4__s000__tb65536__v001",
"uid": "cb__85d56cb4__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\": \"ceval-valid_college_programming__33fb307a__s000__tb131072__v002\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 621.9637731789594, \"nll_base\": 6.432881848565463, \"eval_ref\": \"ceval-valid_college_programming__33fb307a__s000__tb131072__v002... | {
"index": "ceval-valid_college_programming__33fb307a__s000__tb131072__v002",
"uid": "ceval-valid_college_programming__33fb307a__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\": \"ceval-valid_college_programming__33fb307a__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 621.9637731789594, \"nll_base\": 6.432881848565463, \"eval_ref\": \"ceval-valid_college_programming__33fb307a__s000__tb65536__v001\",... | {
"index": "ceval-valid_college_programming__33fb307a__s000__tb65536__v001",
"uid": "ceval-valid_college_programming__33fb307a__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\": \"ceval-valid_college_programming__33fb307a__s001__tb65536__v004\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 630.5363610550495, \"nll_base\": 6.446570823932516, \"eval_ref\": \"ceval-valid_college_programming__33fb307a__s001__tb65536__v004\",... | {
"index": "ceval-valid_college_programming__33fb307a__s001__tb65536__v004",
"uid": "ceval-valid_college_programming__33fb307a__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\": \"ceval-valid_college_programming__33fb307a__s002__tb131072__v008\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 687.5781537970823, \"nll_base\": 6.533175501330145, \"eval_ref\": \"ceval-valid_college_programming__33fb307a__s002__tb131072__v008... | {
"index": "ceval-valid_college_programming__33fb307a__s002__tb131072__v008",
"uid": "ceval-valid_college_programming__33fb307a__s002__tb131072__v008",
"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\": \"ceval-valid_college_programming__33fb307a__s002__tb65536__v007\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 687.5781537970823, \"nll_base\": 6.533175501330145, \"eval_ref\": \"ceval-valid_college_programming__33fb307a__s002__tb65536__v007\",... | {
"index": "ceval-valid_college_programming__33fb307a__s002__tb65536__v007",
"uid": "ceval-valid_college_programming__33fb307a__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\": \"ceval-valid_college_programming__33fb307a__s003__tb131072__v011\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 641.0853839740322, \"nll_base\": 6.463162652377425, \"eval_ref\": \"ceval-valid_college_programming__33fb307a__s003__tb131072__v011... | {
"index": "ceval-valid_college_programming__33fb307a__s003__tb131072__v011",
"uid": "ceval-valid_college_programming__33fb307a__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\": \"ceval-valid_college_programming__33fb307a__s003__tb65536__v010\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 641.0853839740322, \"nll_base\": 6.463162652377425, \"eval_ref\": \"ceval-valid_college_programming__33fb307a__s003__tb65536__v010\",... | {
"index": "ceval-valid_college_programming__33fb307a__s003__tb65536__v010",
"uid": "ceval-valid_college_programming__33fb307a__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\": \"ceval-valid_college_programming__33fb307a__s004__tb131072__v014\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 580.0708405028334, \"nll_base\": 6.363150234880118, \"eval_ref\": \"ceval-valid_college_programming__33fb307a__s004__tb131072__v014... | {
"index": "ceval-valid_college_programming__33fb307a__s004__tb131072__v014",
"uid": "ceval-valid_college_programming__33fb307a__s004__tb131072__v014",
"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\": \"ceval-valid_college_programming__33fb307a__s004__tb65536__v013\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 580.0708405028334, \"nll_base\": 6.363150234880118, \"eval_ref\": \"ceval-valid_college_programming__33fb307a__s004__tb65536__v013\",... | {
"index": "ceval-valid_college_programming__33fb307a__s004__tb65536__v013",
"uid": "ceval-valid_college_programming__33fb307a__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\": \"ceval-valid_college_programming__33fb307a__s005__tb131072__v017\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 574.4854940080621, \"nll_base\": 6.353474847201643, \"eval_ref\": \"ceval-valid_college_programming__33fb307a__s005__tb131072__v017... | {
"index": "ceval-valid_college_programming__33fb307a__s005__tb131072__v017",
"uid": "ceval-valid_college_programming__33fb307a__s005__tb131072__v017",
"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\": \"ceval-valid_college_programming__33fb307a__s005__tb65536__v016\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 574.4854940080621, \"nll_base\": 6.353474847201643, \"eval_ref\": \"ceval-valid_college_programming__33fb307a__s005__tb65536__v016\",... | {
"index": "ceval-valid_college_programming__33fb307a__s005__tb65536__v016",
"uid": "ceval-valid_college_programming__33fb307a__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\": \"ceval-valid_college_programming__33fb307a__s006__tb131072__v020\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 593.3455322622633, \"nll_base\": 6.385776914399246, \"eval_ref\": \"ceval-valid_college_programming__33fb307a__s006__tb131072__v020... | {
"index": "ceval-valid_college_programming__33fb307a__s006__tb131072__v020",
"uid": "ceval-valid_college_programming__33fb307a__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\": \"ceval-valid_college_programming__33fb307a__s006__tb65536__v019\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 593.3455322622633, \"nll_base\": 6.385776914399246, \"eval_ref\": \"ceval-valid_college_programming__33fb307a__s006__tb65536__v019\",... | {
"index": "ceval-valid_college_programming__33fb307a__s006__tb65536__v019",
"uid": "ceval-valid_college_programming__33fb307a__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\": \"ceval-valid_college_programming__33fb307a__s008__tb131072__v026\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 740.1219304555855, \"nll_base\": 6.60681494351091, \"eval_ref\": \"ceval-valid_college_programming__33fb307a__s008__tb131072__v026\... | {
"index": "ceval-valid_college_programming__33fb307a__s008__tb131072__v026",
"uid": "ceval-valid_college_programming__33fb307a__s008__tb131072__v026",
"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\": \"ceval-valid_college_programming__33fb307a__s008__tb65536__v025\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 740.1219304555855, \"nll_base\": 6.60681494351091, \"eval_ref\": \"ceval-valid_college_programming__33fb307a__s008__tb65536__v025\", ... | {
"index": "ceval-valid_college_programming__33fb307a__s008__tb65536__v025",
"uid": "ceval-valid_college_programming__33fb307a__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\": \"ceval-valid_college_programming__33fb307a__s009__tb65536__v028\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 634.1199604825036, \"nll_base\": 6.4522381486563845, \"eval_ref\": \"ceval-valid_college_programming__33fb307a__s009__tb65536__v028\"... | {
"index": "ceval-valid_college_programming__33fb307a__s009__tb65536__v028",
"uid": "ceval-valid_college_programming__33fb307a__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\": \"ceval-valid_college_programming__33fb307a__s010__tb131072__v032\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 643.562859442536, \"nll_base\": 6.467019705936827, \"eval_ref\": \"ceval-valid_college_programming__33fb307a__s010__tb131072__v032\... | {
"index": "ceval-valid_college_programming__33fb307a__s010__tb131072__v032",
"uid": "ceval-valid_college_programming__33fb307a__s010__tb131072__v032",
"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\": \"ceval-valid_college_programming__33fb307a__s011__tb131072__v035\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 612.2073989467997, \"nll_base\": 6.417071112271013, \"eval_ref\": \"ceval-valid_college_programming__33fb307a__s011__tb131072__v035... | {
"index": "ceval-valid_college_programming__33fb307a__s011__tb131072__v035",
"uid": "ceval-valid_college_programming__33fb307a__s011__tb131072__v035",
"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\": \"ceval-valid_college_programming__33fb307a__s012__tb65536__v037\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 622.2257965090722, \"nll_base\": 6.433303043760103, \"eval_ref\": \"ceval-valid_college_programming__33fb307a__s012__tb65536__v037\",... | {
"index": "ceval-valid_college_programming__33fb307a__s012__tb65536__v037",
"uid": "ceval-valid_college_programming__33fb307a__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\": \"ceval-valid_college_programming__33fb307a__s013__tb131072__v041\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 599.1210201147767, \"nll_base\": 6.395463614628233, \"eval_ref\": \"ceval-valid_college_programming__33fb307a__s013__tb131072__v041... | {
"index": "ceval-valid_college_programming__33fb307a__s013__tb131072__v041",
"uid": "ceval-valid_college_programming__33fb307a__s013__tb131072__v041",
"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\": \"ceval-valid_college_programming__33fb307a__s013__tb65536__v040\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 599.1210201147767, \"nll_base\": 6.395463614628233, \"eval_ref\": \"ceval-valid_college_programming__33fb307a__s013__tb65536__v040\",... | {
"index": "ceval-valid_college_programming__33fb307a__s013__tb65536__v040",
"uid": "ceval-valid_college_programming__33fb307a__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\": \"ceval-valid_high_school_chinese__42455218__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 638.9736022770852, \"nll_base\": 6.459863142533735, \"eval_ref\": \"ceval-valid_high_school_chinese__42455218__s000__tb131072__v000... | {
"index": "ceval-valid_high_school_chinese__42455218__s000__tb131072__v000",
"uid": "ceval-valid_high_school_chinese__42455218__s000__tb131072__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"ceval-valid_legal_professional__3616fe3a__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1192.3255093642094, \"nll_base\": 7.083660888671875, \"eval_ref\": \"ceval-valid_legal_professional__3616fe3a__s000__tb131072__v001\... | {
"index": "ceval-valid_legal_professional__3616fe3a__s000__tb131072__v001",
"uid": "ceval-valid_legal_professional__3616fe3a__s000__tb131072__v001",
"axis": "domain",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"ceval-valid_legal_professional__3616fe3a__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1192.3255093642094, \"nll_base\": 7.083660888671875, \"eval_ref\": \"ceval-valid_legal_professional__3616fe3a__s000__tb65536__v000\", ... | {
"index": "ceval-valid_legal_professional__3616fe3a__s000__tb65536__v000",
"uid": "ceval-valid_legal_professional__3616fe3a__s000__tb65536__v000",
"axis": "domain",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"ceval-valid_legal_professional__3616fe3a__s001__tb65536__v003\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1165.199484891449, \"nll_base\": 7.060647583007812, \"eval_ref\": \"ceval-valid_legal_professional__3616fe3a__s001__tb65536__v003\", \... | {
"index": "ceval-valid_legal_professional__3616fe3a__s001__tb65536__v003",
"uid": "ceval-valid_legal_professional__3616fe3a__s001__tb65536__v003",
"axis": "domain",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"cmmlu_chinese_teacher_qualification__f6aca9db__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 959.8348204061393, \"nll_base\": 6.866761207580566, \"eval_ref\": \"cmmlu_chinese_teacher_qualification__f6aca9db__s000__tb65536_... | {
"index": "cmmlu_chinese_teacher_qualification__f6aca9db__s000__tb65536__v000",
"uid": "cmmlu_chinese_teacher_qualification__f6aca9db__s000__tb65536__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"cmmlu_clinical_knowledge__17df098b__s000__tb131072__v002\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1433.9398656491855, \"nll_base\": 7.268181085586548, \"eval_ref\": \"cmmlu_clinical_knowledge__17df098b__s000__tb131072__v002\", \"oracle_... | {
"index": "cmmlu_clinical_knowledge__17df098b__s000__tb131072__v002",
"uid": "cmmlu_clinical_knowledge__17df098b__s000__tb131072__v002",
"axis": "domain",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"cmmlu_clinical_knowledge__17df098b__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1433.9398656491855, \"nll_base\": 7.268181085586548, \"eval_ref\": \"cmmlu_clinical_knowledge__17df098b__s000__tb262144__v000\", \"oracle_... | {
"index": "cmmlu_clinical_knowledge__17df098b__s000__tb262144__v000",
"uid": "cmmlu_clinical_knowledge__17df098b__s000__tb262144__v000",
"axis": "domain",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.