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\": \"mmlu_it_llama_college_physics__0a850d65__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 296.12714464774103, \"nll_base\": 5.690788904825847, \"eval_ref\": \"mmlu_it_llama_college_physics__0a850d65__s000__tb131072__v000\",... | {
"index": "mmlu_it_llama_college_physics__0a850d65__s000__tb131072__v000",
"uid": "mmlu_it_llama_college_physics__0a850d65__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\": \"mmlu_it_llama_computer_security__45cb4d9a__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 246.32407910367093, \"nll_base\": 5.506648063659668, \"eval_ref\": \"mmlu_it_llama_computer_security__45cb4d9a__s000__tb131072__v00... | {
"index": "mmlu_it_llama_computer_security__45cb4d9a__s000__tb131072__v000",
"uid": "mmlu_it_llama_computer_security__45cb4d9a__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\": \"mmlu_it_llama_econometrics__3609a80d__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 231.04856044354122, \"nll_base\": 5.442627906799316, \"eval_ref\": \"mmlu_it_llama_econometrics__3609a80d__s000__tb131072__v000\", \"ora... | {
"index": "mmlu_it_llama_econometrics__3609a80d__s000__tb131072__v000",
"uid": "mmlu_it_llama_econometrics__3609a80d__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\": \"mmlu_it_llama_elementary_mathematics__baa25ec0__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 304.85276536465216, \"nll_base\": 5.719828923543294, \"eval_ref\": \"mmlu_it_llama_elementary_mathematics__baa25ec0__s000__tb1... | {
"index": "mmlu_it_llama_elementary_mathematics__baa25ec0__s000__tb131072__v000",
"uid": "mmlu_it_llama_elementary_mathematics__baa25ec0__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\": \"mmlu_it_llama_formal_logic__20897c69__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 241.0867181333137, \"nll_base\": 5.485156695048015, \"eval_ref\": \"mmlu_it_llama_formal_logic__20897c69__s000__tb65536__v000\", \"oracle_... | {
"index": "mmlu_it_llama_formal_logic__20897c69__s000__tb65536__v000",
"uid": "mmlu_it_llama_formal_logic__20897c69__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\": \"mmlu_it_llama_high_school_chemistry__98655755__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 263.85919570619745, \"nll_base\": 5.57541561126709, \"eval_ref\": \"mmlu_it_llama_high_school_chemistry__98655755__s000__tb65536_... | {
"index": "mmlu_it_llama_high_school_chemistry__98655755__s000__tb65536__v000",
"uid": "mmlu_it_llama_high_school_chemistry__98655755__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\": \"mmlu_it_llama_high_school_geography__f4362716__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 301.57978313681974, \"nll_base\": 5.709034601847331, \"eval_ref\": \"mmlu_it_llama_high_school_geography__f4362716__s000__tb131... | {
"index": "mmlu_it_llama_high_school_geography__f4362716__s000__tb131072__v000",
"uid": "mmlu_it_llama_high_school_geography__f4362716__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\": \"mmlu_it_llama_high_school_microeconomics__ee67664a__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 265.2265923405355, \"nll_base\": 5.580584526062012, \"eval_ref\": \"mmlu_it_llama_high_school_microeconomics__ee67664a__s0... | {
"index": "mmlu_it_llama_high_school_microeconomics__ee67664a__s000__tb131072__v000",
"uid": "mmlu_it_llama_high_school_microeconomics__ee67664a__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\": \"mmlu_it_llama_high_school_physics__8222e8e2__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 261.3022963633894, \"nll_base\": 5.565677960713704, \"eval_ref\": \"mmlu_it_llama_high_school_physics__8222e8e2__s000__tb65536__v00... | {
"index": "mmlu_it_llama_high_school_physics__8222e8e2__s000__tb65536__v000",
"uid": "mmlu_it_llama_high_school_physics__8222e8e2__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\": \"mmlu_it_llama_high_school_statistics__3aef9db4__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 259.1215018530028, \"nll_base\": 5.557297070821126, \"eval_ref\": \"mmlu_it_llama_high_school_statistics__3aef9db4__s000__tb13... | {
"index": "mmlu_it_llama_high_school_statistics__3aef9db4__s000__tb131072__v000",
"uid": "mmlu_it_llama_high_school_statistics__3aef9db4__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\": \"mmlu_it_llama_high_school_world_history__3989724b__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 158.2182130146344, \"nll_base\": 5.063975175221761, \"eval_ref\": \"mmlu_it_llama_high_school_world_history__3989724b__s000__... | {
"index": "mmlu_it_llama_high_school_world_history__3989724b__s000__tb65536__v000",
"uid": "mmlu_it_llama_high_school_world_history__3989724b__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\": \"mmlu_it_llama_logical_fallacies__5d82243d__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 292.8467260992515, \"nll_base\": 5.679649353027344, \"eval_ref\": \"mmlu_it_llama_logical_fallacies__5d82243d__s000__tb131072__v000... | {
"index": "mmlu_it_llama_logical_fallacies__5d82243d__s000__tb131072__v000",
"uid": "mmlu_it_llama_logical_fallacies__5d82243d__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\": \"mmlu_it_llama_machine_learning__23dea8d8__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 240.96420286719987, \"nll_base\": 5.48464838663737, \"eval_ref\": \"mmlu_it_llama_machine_learning__23dea8d8__s000__tb65536__v000\", \... | {
"index": "mmlu_it_llama_machine_learning__23dea8d8__s000__tb65536__v000",
"uid": "mmlu_it_llama_machine_learning__23dea8d8__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\": \"mmlu_it_llama_miscellaneous__ca154f72__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 301.37718282012634, \"nll_base\": 5.708362579345703, \"eval_ref\": \"mmlu_it_llama_miscellaneous__ca154f72__s000__tb65536__v000\", \"orac... | {
"index": "mmlu_it_llama_miscellaneous__ca154f72__s000__tb65536__v000",
"uid": "mmlu_it_llama_miscellaneous__ca154f72__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\": \"mmlu_it_llama_moral_disputes__176684de__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 237.45293408427648, \"nll_base\": 5.469969431559245, \"eval_ref\": \"mmlu_it_llama_moral_disputes__176684de__s000__tb131072__v000\", \... | {
"index": "mmlu_it_llama_moral_disputes__176684de__s000__tb131072__v000",
"uid": "mmlu_it_llama_moral_disputes__176684de__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\": \"mmlu_it_llama_nutrition__a1b698cc__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 278.7174316166754, \"nll_base\": 5.6301984786987305, \"eval_ref\": \"mmlu_it_llama_nutrition__a1b698cc__s000__tb131072__v000\", \"oracle_de... | {
"index": "mmlu_it_llama_nutrition__a1b698cc__s000__tb131072__v000",
"uid": "mmlu_it_llama_nutrition__a1b698cc__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\": \"mmlu_it_llama_philosophy__c5c05fd7__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 270.3394407716151, \"nll_base\": 5.59967835744222, \"eval_ref\": \"mmlu_it_llama_philosophy__c5c05fd7__s000__tb65536__v000\", \"oracle_delta... | {
"index": "mmlu_it_llama_philosophy__c5c05fd7__s000__tb65536__v000",
"uid": "mmlu_it_llama_philosophy__c5c05fd7__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\": \"mmlu_it_llama_professional_accounting__b102d4bc__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 209.16068263883744, \"nll_base\": 5.343102773030599, \"eval_ref\": \"mmlu_it_llama_professional_accounting__b102d4bc__s000__tb6... | {
"index": "mmlu_it_llama_professional_accounting__b102d4bc__s000__tb65536__v000",
"uid": "mmlu_it_llama_professional_accounting__b102d4bc__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\": \"mmlu_it_llama_professional_law__3eb8cef8__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 129.14687350789077, \"nll_base\": 4.860950311024983, \"eval_ref\": \"mmlu_it_llama_professional_law__3eb8cef8__s000__tb65536__v000\", ... | {
"index": "mmlu_it_llama_professional_law__3eb8cef8__s000__tb65536__v000",
"uid": "mmlu_it_llama_professional_law__3eb8cef8__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\": \"mmlu_it_llama_professional_medicine__4f7928a4__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 283.51962048615474, \"nll_base\": 5.647281328837077, \"eval_ref\": \"mmlu_it_llama_professional_medicine__4f7928a4__s000__tb131... | {
"index": "mmlu_it_llama_professional_medicine__4f7928a4__s000__tb131072__v000",
"uid": "mmlu_it_llama_professional_medicine__4f7928a4__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\": \"mmlu_it_llama_professional_psychology__7d41fc83__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 247.08849492592017, \"nll_base\": 5.509746551513672, \"eval_ref\": \"mmlu_it_llama_professional_psychology__7d41fc83__s000__t... | {
"index": "mmlu_it_llama_professional_psychology__7d41fc83__s000__tb131072__v000",
"uid": "mmlu_it_llama_professional_psychology__7d41fc83__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\": \"mmlu_it_llama_public_relations__f6b41185__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 328.20911304970247, \"nll_base\": 5.793650945027669, \"eval_ref\": \"mmlu_it_llama_public_relations__f6b41185__s000__tb262144__v000\... | {
"index": "mmlu_it_llama_public_relations__f6b41185__s000__tb262144__v000",
"uid": "mmlu_it_llama_public_relations__f6b41185__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\": \"mmlu_it_llama_sociology__315cec94__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 254.72187381810656, \"nll_base\": 5.540172259012858, \"eval_ref\": \"mmlu_it_llama_sociology__315cec94__s000__tb262144__v000\", \"oracle_de... | {
"index": "mmlu_it_llama_sociology__315cec94__s000__tb262144__v000",
"uid": "mmlu_it_llama_sociology__315cec94__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\": \"mmlu_it_llama_us_foreign_policy__6146603a__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 295.3247804439847, \"nll_base\": 5.688075701395671, \"eval_ref\": \"mmlu_it_llama_us_foreign_policy__6146603a__s000__tb131072__v000... | {
"index": "mmlu_it_llama_us_foreign_policy__6146603a__s000__tb131072__v000",
"uid": "mmlu_it_llama_us_foreign_policy__6146603a__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\": \"mmlu_jurisprudence__433dd0b5__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1123.2055715820848, \"nll_base\": 7.023941993713379, \"eval_ref\": \"mmlu_jurisprudence__433dd0b5__s000__tb131072__v000\", \"oracle_delta\": 0.8... | {
"index": "mmlu_jurisprudence__433dd0b5__s000__tb131072__v000",
"uid": "mmlu_jurisprudence__433dd0b5__s000__tb131072__v000",
"axis": "knowledge",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_jurisprudence_generative__ef907898__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1046.2228337484732, \"nll_base\": 6.952941656112671, \"eval_ref\": \"mmlu_jurisprudence_generative__ef907898__s000__tb262144__v000\",... | {
"index": "mmlu_jurisprudence_generative__ef907898__s000__tb262144__v000",
"uid": "mmlu_jurisprudence_generative__ef907898__s000__tb262144__v000",
"axis": "knowledge",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_llama_astronomy__4f5fa693__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 282.181496477822, \"nll_base\": 5.642550468444824, \"eval_ref\": \"mmlu_llama_astronomy__4f5fa693__s000__tb131072__v000\", \"oracle_delta\": 0... | {
"index": "mmlu_llama_astronomy__4f5fa693__s000__tb131072__v000",
"uid": "mmlu_llama_astronomy__4f5fa693__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\": \"mmlu_llama_clinical_knowledge__46eb0b45__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 245.23547765628737, \"nll_base\": 5.502218882242839, \"eval_ref\": \"mmlu_llama_clinical_knowledge__46eb0b45__s000__tb131072__v001\",... | {
"index": "mmlu_llama_clinical_knowledge__46eb0b45__s000__tb131072__v001",
"uid": "mmlu_llama_clinical_knowledge__46eb0b45__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\": \"mmlu_llama_clinical_knowledge__46eb0b45__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 245.23547765628737, \"nll_base\": 5.502218882242839, \"eval_ref\": \"mmlu_llama_clinical_knowledge__46eb0b45__s000__tb65536__v000\", \"... | {
"index": "mmlu_llama_clinical_knowledge__46eb0b45__s000__tb65536__v000",
"uid": "mmlu_llama_clinical_knowledge__46eb0b45__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\": \"mmlu_llama_clinical_knowledge__46eb0b45__s001__tb65536__v003\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 226.30901424031697, \"nll_base\": 5.421901384989421, \"eval_ref\": \"mmlu_llama_clinical_knowledge__46eb0b45__s001__tb65536__v003\", \"... | {
"index": "mmlu_llama_clinical_knowledge__46eb0b45__s001__tb65536__v003",
"uid": "mmlu_llama_clinical_knowledge__46eb0b45__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\": \"mmlu_llama_college_biology__1cd8f27f__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 202.05190273150419, \"nll_base\": 5.3085246086120605, \"eval_ref\": \"mmlu_llama_college_biology__1cd8f27f__s000__tb131072__v000\", \"or... | {
"index": "mmlu_llama_college_biology__1cd8f27f__s000__tb131072__v000",
"uid": "mmlu_llama_college_biology__1cd8f27f__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\": \"mmlu_llama_college_chemistry__17508e89__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 227.36903200788265, \"nll_base\": 5.4265743891398115, \"eval_ref\": \"mmlu_llama_college_chemistry__17508e89__s000__tb131072__v000\", ... | {
"index": "mmlu_llama_college_chemistry__17508e89__s000__tb131072__v000",
"uid": "mmlu_llama_college_chemistry__17508e89__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\": \"mmlu_llama_college_physics__5ffae102__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 233.80285584165927, \"nll_base\": 5.4544782638549805, \"eval_ref\": \"mmlu_llama_college_physics__5ffae102__s000__tb131072__v000\", \"or... | {
"index": "mmlu_llama_college_physics__5ffae102__s000__tb131072__v000",
"uid": "mmlu_llama_college_physics__5ffae102__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\": \"mmlu_llama_econometrics__d60b149a__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 182.599069928039, \"nll_base\": 5.207292874654134, \"eval_ref\": \"mmlu_llama_econometrics__d60b149a__s000__tb131072__v000\", \"oracle_delt... | {
"index": "mmlu_llama_econometrics__d60b149a__s000__tb131072__v000",
"uid": "mmlu_llama_econometrics__d60b149a__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\": \"mmlu_llama_electrical_engineering__5ba8e103__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 221.97858139461908, \"nll_base\": 5.402580897013347, \"eval_ref\": \"mmlu_llama_electrical_engineering__5ba8e103__s000__tb131072_... | {
"index": "mmlu_llama_electrical_engineering__5ba8e103__s000__tb131072__v000",
"uid": "mmlu_llama_electrical_engineering__5ba8e103__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\": \"mmlu_llama_high_school_computer_science__7df55758__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 213.67900928587366, \"nll_base\": 5.364474932352702, \"eval_ref\": \"mmlu_llama_high_school_computer_science__7df55758__s000_... | {
"index": "mmlu_llama_high_school_computer_science__7df55758__s000__tb65536__v000",
"uid": "mmlu_llama_high_school_computer_science__7df55758__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\": \"mmlu_llama_high_school_mathematics__25e4e6e5__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 207.99907507034126, \"nll_base\": 5.337533632914226, \"eval_ref\": \"mmlu_llama_high_school_mathematics__25e4e6e5__s000__tb13107... | {
"index": "mmlu_llama_high_school_mathematics__25e4e6e5__s000__tb131072__v000",
"uid": "mmlu_llama_high_school_mathematics__25e4e6e5__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\": \"mmlu_llama_high_school_world_history__8c9af912__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 288.53027163983205, \"nll_base\": 5.6648000081380205, \"eval_ref\": \"mmlu_llama_high_school_world_history__8c9af912__s000__tb65... | {
"index": "mmlu_llama_high_school_world_history__8c9af912__s000__tb65536__v000",
"uid": "mmlu_llama_high_school_world_history__8c9af912__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\": \"mmlu_llama_jurisprudence__236dfd88__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 197.22514510663913, \"nll_base\": 5.284345944722493, \"eval_ref\": \"mmlu_llama_jurisprudence__236dfd88__s000__tb65536__v000\", \"oracle_del... | {
"index": "mmlu_llama_jurisprudence__236dfd88__s000__tb65536__v000",
"uid": "mmlu_llama_jurisprudence__236dfd88__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\": \"mmlu_llama_logical_fallacies__d98eb3bd__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 299.68257195723385, \"nll_base\": 5.7027238210042315, \"eval_ref\": \"mmlu_llama_logical_fallacies__d98eb3bd__s000__tb131072__v000\", ... | {
"index": "mmlu_llama_logical_fallacies__d98eb3bd__s000__tb131072__v000",
"uid": "mmlu_llama_logical_fallacies__d98eb3bd__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\": \"mmlu_llama_miscellaneous__07294076__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 236.77567572590166, \"nll_base\": 5.467113176981608, \"eval_ref\": \"mmlu_llama_miscellaneous__07294076__s000__tb65536__v000\", \"oracle_del... | {
"index": "mmlu_llama_miscellaneous__07294076__s000__tb65536__v000",
"uid": "mmlu_llama_miscellaneous__07294076__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\": \"mmlu_llama_professional_accounting__d01fb9ac__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 187.1798593245749, \"nll_base\": 5.232069969177246, \"eval_ref\": \"mmlu_llama_professional_accounting__d01fb9ac__s000__tb131072... | {
"index": "mmlu_llama_professional_accounting__d01fb9ac__s000__tb131072__v000",
"uid": "mmlu_llama_professional_accounting__d01fb9ac__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\": \"mmlu_llama_professional_law__7f0bd81a__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 185.62166386607174, \"nll_base\": 5.223710536956787, \"eval_ref\": \"mmlu_llama_professional_law__7f0bd81a__s000__tb131072__v000\", \"o... | {
"index": "mmlu_llama_professional_law__7f0bd81a__s000__tb131072__v000",
"uid": "mmlu_llama_professional_law__7f0bd81a__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\": \"mmlu_llama_public_relations__1a1dc65c__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 236.39219018296674, \"nll_base\": 5.465492248535156, \"eval_ref\": \"mmlu_llama_public_relations__1a1dc65c__s000__tb65536__v000\", \"orac... | {
"index": "mmlu_llama_public_relations__1a1dc65c__s000__tb65536__v000",
"uid": "mmlu_llama_public_relations__1a1dc65c__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\": \"mmlu_llama_us_foreign_policy__d365ef1a__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 208.30901711611247, \"nll_base\": 5.339022636413574, \"eval_ref\": \"mmlu_llama_us_foreign_policy__d365ef1a__s000__tb65536__v000\", \"or... | {
"index": "mmlu_llama_us_foreign_policy__d365ef1a__s000__tb65536__v000",
"uid": "mmlu_llama_us_foreign_policy__d365ef1a__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\": \"mmlu_logical_fallacies__c5b41de8__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 636.9240995748551, \"nll_base\": 6.456650495529175, \"eval_ref\": \"mmlu_logical_fallacies__c5b41de8__s000__tb131072__v000\", \"oracle_delta... | {
"index": "mmlu_logical_fallacies__c5b41de8__s000__tb131072__v000",
"uid": "mmlu_logical_fallacies__c5b41de8__s000__tb131072__v000",
"axis": "knowledge",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_logical_fallacies_continuation__ee168e5e__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 13.462094612600925, \"nll_base\": 2.5998779296875, \"eval_ref\": \"mmlu_logical_fallacies_continuation__ee168e5e__s000__tb13107... | {
"index": "mmlu_logical_fallacies_continuation__ee168e5e__s000__tb131072__v001",
"uid": "mmlu_logical_fallacies_continuation__ee168e5e__s000__tb131072__v001",
"axis": "knowledge",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_logical_fallacies_continuation__ee168e5e__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 13.462094612600925, \"nll_base\": 2.5998779296875, \"eval_ref\": \"mmlu_logical_fallacies_continuation__ee168e5e__s000__tb65536__... | {
"index": "mmlu_logical_fallacies_continuation__ee168e5e__s000__tb65536__v000",
"uid": "mmlu_logical_fallacies_continuation__ee168e5e__s000__tb65536__v000",
"axis": "knowledge",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_logical_fallacies_generative__281a7b88__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 826.4973433820751, \"nll_base\": 6.717196702957153, \"eval_ref\": \"mmlu_logical_fallacies_generative__281a7b88__s000__tb131072__... | {
"index": "mmlu_logical_fallacies_generative__281a7b88__s000__tb131072__v000",
"uid": "mmlu_logical_fallacies_generative__281a7b88__s000__tb131072__v000",
"axis": "knowledge",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_logical_fallacies_generative_spanish__7e9ce444__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3121.580290659888, \"nll_base\": 8.0460946559906, \"eval_ref\": \"mmlu_logical_fallacies_generative_spanish__7e9ce444__s0... | {
"index": "mmlu_logical_fallacies_generative_spanish__7e9ce444__s000__tb131072__v000",
"uid": "mmlu_logical_fallacies_generative_spanish__7e9ce444__s000__tb131072__v000",
"axis": "knowledge",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_machine_learning_generative_spanish__a46353f3__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2719.9819097013974, \"nll_base\": 7.908380508422852, \"eval_ref\": \"mmlu_machine_learning_generative_spanish__a46353f3__s... | {
"index": "mmlu_machine_learning_generative_spanish__a46353f3__s000__tb131072__v001",
"uid": "mmlu_machine_learning_generative_spanish__a46353f3__s000__tb131072__v001",
"axis": "knowledge",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_machine_learning_generative_spanish__a46353f3__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2719.9819097013974, \"nll_base\": 7.908380508422852, \"eval_ref\": \"mmlu_machine_learning_generative_spanish__a46353f3__s00... | {
"index": "mmlu_machine_learning_generative_spanish__a46353f3__s000__tb65536__v000",
"uid": "mmlu_machine_learning_generative_spanish__a46353f3__s000__tb65536__v000",
"axis": "knowledge",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_management__382e6768__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1807.3868744799183, \"nll_base\": 7.4996373653411865, \"eval_ref\": \"mmlu_management__382e6768__s000__tb131072__v000\", \"oracle_delta\": 0.679131... | {
"index": "mmlu_management__382e6768__s000__tb131072__v000",
"uid": "mmlu_management__382e6768__s000__tb131072__v000",
"axis": "knowledge",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_management_continuation__228746a8__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 37.90530765055013, \"nll_base\": 3.6350911458333335, \"eval_ref\": \"mmlu_management_continuation__228746a8__s000__tb65536__v001\", \"or... | {
"index": "mmlu_management_continuation__228746a8__s000__tb65536__v001",
"uid": "mmlu_management_continuation__228746a8__s000__tb65536__v001",
"axis": "knowledge",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_management_generative__5e9157aa__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1190.2909460558365, \"nll_base\": 7.081953048706055, \"eval_ref\": \"mmlu_management_generative__5e9157aa__s000__tb65536__v000\", \"oracle... | {
"index": "mmlu_management_generative__5e9157aa__s000__tb65536__v000",
"uid": "mmlu_management_generative__5e9157aa__s000__tb65536__v000",
"axis": "knowledge",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_marketing__e3205d75__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 977.9161192288507, \"nll_base\": 6.885423898696899, \"eval_ref\": \"mmlu_marketing__e3205d75__s000__tb65536__v001\", \"oracle_delta\": 0.0621273517608... | {
"index": "mmlu_marketing__e3205d75__s000__tb65536__v001",
"uid": "mmlu_marketing__e3205d75__s000__tb65536__v001",
"axis": "knowledge",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_marketing_continuation__5d0410dc__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 16.552026924717293, \"nll_base\": 2.8065085671164773, \"eval_ref\": \"mmlu_marketing_continuation__5d0410dc__s000__tb131072__v000\", \"... | {
"index": "mmlu_marketing_continuation__5d0410dc__s000__tb131072__v000",
"uid": "mmlu_marketing_continuation__5d0410dc__s000__tb131072__v000",
"axis": "knowledge",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_marketing_generative__b4534206__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 979.7623361064151, \"nll_base\": 6.887310028076172, \"eval_ref\": \"mmlu_marketing_generative__b4534206__s000__tb65536__v000\", \"oracle_de... | {
"index": "mmlu_marketing_generative__b4534206__s000__tb65536__v000",
"uid": "mmlu_marketing_generative__b4534206__s000__tb65536__v000",
"axis": "knowledge",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_marketing_generative_spanish__8fb1951d__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1833.2146979311033, \"nll_base\": 7.513826370239258, \"eval_ref\": \"mmlu_marketing_generative_spanish__8fb1951d__s000__tb131072_... | {
"index": "mmlu_marketing_generative_spanish__8fb1951d__s000__tb131072__v000",
"uid": "mmlu_marketing_generative_spanish__8fb1951d__s000__tb131072__v000",
"axis": "knowledge",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_marketing_generative_spanish__8fb1951d__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1833.2146979311033, \"nll_base\": 7.513826370239258, \"eval_ref\": \"mmlu_marketing_generative_spanish__8fb1951d__s000__tb262144_... | {
"index": "mmlu_marketing_generative_spanish__8fb1951d__s000__tb262144__v001",
"uid": "mmlu_marketing_generative_spanish__8fb1951d__s000__tb262144__v001",
"axis": "knowledge",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_medical_genetics__1a6e574f__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1014.4169684763353, \"nll_base\": 6.922069311141968, \"eval_ref\": \"mmlu_medical_genetics__1a6e574f__s000__tb131072__v000\", \"oracle_delta\... | {
"index": "mmlu_medical_genetics__1a6e574f__s000__tb131072__v000",
"uid": "mmlu_medical_genetics__1a6e574f__s000__tb131072__v000",
"axis": "domain",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_medical_genetics__1a6e574f__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1014.4169684763353, \"nll_base\": 6.922069311141968, \"eval_ref\": \"mmlu_medical_genetics__1a6e574f__s000__tb262144__v001\", \"oracle_delta\... | {
"index": "mmlu_medical_genetics__1a6e574f__s000__tb262144__v001",
"uid": "mmlu_medical_genetics__1a6e574f__s000__tb262144__v001",
"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\": \"mmlu_medical_genetics__1a6e574f__s000__tb65536__v002\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1014.4169684763353, \"nll_base\": 6.922069311141968, \"eval_ref\": \"mmlu_medical_genetics__1a6e574f__s000__tb65536__v002\", \"oracle_delta\": ... | {
"index": "mmlu_medical_genetics__1a6e574f__s000__tb65536__v002",
"uid": "mmlu_medical_genetics__1a6e574f__s000__tb65536__v002",
"axis": "domain",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_medical_genetics_continuation__77d84573__s000__tb131072__v002\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 14.160969367034609, \"nll_base\": 2.6504895440463363, \"eval_ref\": \"mmlu_medical_genetics_continuation__77d84573__s000__tb1310... | {
"index": "mmlu_medical_genetics_continuation__77d84573__s000__tb131072__v002",
"uid": "mmlu_medical_genetics_continuation__77d84573__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\": \"mmlu_medical_genetics_continuation__77d84573__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 14.160969367034609, \"nll_base\": 2.6504895440463363, \"eval_ref\": \"mmlu_medical_genetics_continuation__77d84573__s000__tb65536_... | {
"index": "mmlu_medical_genetics_continuation__77d84573__s000__tb65536__v001",
"uid": "mmlu_medical_genetics_continuation__77d84573__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\": \"mmlu_medical_genetics_continuation__77d84573__s001__tb262144__v003\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 8.233604844494948, \"nll_base\": 2.108223931486254, \"eval_ref\": \"mmlu_medical_genetics_continuation__77d84573__s001__tb262144... | {
"index": "mmlu_medical_genetics_continuation__77d84573__s001__tb262144__v003",
"uid": "mmlu_medical_genetics_continuation__77d84573__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\": \"mmlu_medical_genetics_generative__3e36d408__s000__tb131072__v002\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1093.7705402113659, \"nll_base\": 6.99738621711731, \"eval_ref\": \"mmlu_medical_genetics_generative__3e36d408__s000__tb131072__v0... | {
"index": "mmlu_medical_genetics_generative__3e36d408__s000__tb131072__v002",
"uid": "mmlu_medical_genetics_generative__3e36d408__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\": \"mmlu_medical_genetics_generative__3e36d408__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1093.7705402113659, \"nll_base\": 6.99738621711731, \"eval_ref\": \"mmlu_medical_genetics_generative__3e36d408__s000__tb262144__v0... | {
"index": "mmlu_medical_genetics_generative__3e36d408__s000__tb262144__v000",
"uid": "mmlu_medical_genetics_generative__3e36d408__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\": \"mmlu_medical_genetics_generative__3e36d408__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1093.7705402113659, \"nll_base\": 6.99738621711731, \"eval_ref\": \"mmlu_medical_genetics_generative__3e36d408__s000__tb65536__v001\... | {
"index": "mmlu_medical_genetics_generative__3e36d408__s000__tb65536__v001",
"uid": "mmlu_medical_genetics_generative__3e36d408__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\": \"mmlu_medical_genetics_generative__3e36d408__s001__tb262144__v003\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1133.5655009954746, \"nll_base\": 7.033123254776001, \"eval_ref\": \"mmlu_medical_genetics_generative__3e36d408__s001__tb262144__v... | {
"index": "mmlu_medical_genetics_generative__3e36d408__s001__tb262144__v003",
"uid": "mmlu_medical_genetics_generative__3e36d408__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\": \"mmlu_medical_genetics_generative_spanish__45c09a28__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4614.075308108806, \"nll_base\": 8.436866760253906, \"eval_ref\": \"mmlu_medical_genetics_generative_spanish__45c09a28__s0... | {
"index": "mmlu_medical_genetics_generative_spanish__45c09a28__s000__tb131072__v001",
"uid": "mmlu_medical_genetics_generative_spanish__45c09a28__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\": \"mmlu_medical_genetics_generative_spanish__45c09a28__s000__tb262144__v002\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 4614.075308108806, \"nll_base\": 8.436866760253906, \"eval_ref\": \"mmlu_medical_genetics_generative_spanish__45c09a28__s0... | {
"index": "mmlu_medical_genetics_generative_spanish__45c09a28__s000__tb262144__v002",
"uid": "mmlu_medical_genetics_generative_spanish__45c09a28__s000__tb262144__v002",
"axis": "domain",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_medical_genetics_generative_spanish__45c09a28__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4614.075308108806, \"nll_base\": 8.436866760253906, \"eval_ref\": \"mmlu_medical_genetics_generative_spanish__45c09a28__s000... | {
"index": "mmlu_medical_genetics_generative_spanish__45c09a28__s000__tb65536__v000",
"uid": "mmlu_medical_genetics_generative_spanish__45c09a28__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\": \"mmlu_medical_genetics_generative_spanish__45c09a28__s001__tb65536__v003\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4874.787435368523, \"nll_base\": 8.49183177947998, \"eval_ref\": \"mmlu_medical_genetics_generative_spanish__45c09a28__s001_... | {
"index": "mmlu_medical_genetics_generative_spanish__45c09a28__s001__tb65536__v003",
"uid": "mmlu_medical_genetics_generative_spanish__45c09a28__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\": \"mmlu_miscellaneous__a2bb3a3d__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1398.56930837562, \"nll_base\": 7.2432050704956055, \"eval_ref\": \"mmlu_miscellaneous__a2bb3a3d__s000__tb262144__v000\", \"oracle_delta\": 1.00... | {
"index": "mmlu_miscellaneous__a2bb3a3d__s000__tb262144__v000",
"uid": "mmlu_miscellaneous__a2bb3a3d__s000__tb262144__v000",
"axis": "knowledge",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_miscellaneous__a2bb3a3d__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1398.56930837562, \"nll_base\": 7.2432050704956055, \"eval_ref\": \"mmlu_miscellaneous__a2bb3a3d__s000__tb65536__v001\", \"oracle_delta\": 0.36471... | {
"index": "mmlu_miscellaneous__a2bb3a3d__s000__tb65536__v001",
"uid": "mmlu_miscellaneous__a2bb3a3d__s000__tb65536__v001",
"axis": "knowledge",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_miscellaneous_continuation__63934577__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 38.59898268918997, \"nll_base\": 3.653225920921148, \"eval_ref\": \"mmlu_miscellaneous_continuation__63934577__s000__tb65536__v001\",... | {
"index": "mmlu_miscellaneous_continuation__63934577__s000__tb65536__v001",
"uid": "mmlu_miscellaneous_continuation__63934577__s000__tb65536__v001",
"axis": "knowledge",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_miscellaneous_generative__a03680ad__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1672.6746684700245, \"nll_base\": 7.422179222106934, \"eval_ref\": \"mmlu_miscellaneous_generative__a03680ad__s000__tb131072__v000\",... | {
"index": "mmlu_miscellaneous_generative__a03680ad__s000__tb131072__v000",
"uid": "mmlu_miscellaneous_generative__a03680ad__s000__tb131072__v000",
"axis": "knowledge",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_miscellaneous_generative__a03680ad__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1672.6746684700245, \"nll_base\": 7.422179222106934, \"eval_ref\": \"mmlu_miscellaneous_generative__a03680ad__s000__tb262144__v001\",... | {
"index": "mmlu_miscellaneous_generative__a03680ad__s000__tb262144__v001",
"uid": "mmlu_miscellaneous_generative__a03680ad__s000__tb262144__v001",
"axis": "knowledge",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_miscellaneous_generative_spanish__6417d1b0__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 5192.520771178551, \"nll_base\": 8.554974555969238, \"eval_ref\": \"mmlu_miscellaneous_generative_spanish__6417d1b0__s000__tb... | {
"index": "mmlu_miscellaneous_generative_spanish__6417d1b0__s000__tb131072__v001",
"uid": "mmlu_miscellaneous_generative_spanish__6417d1b0__s000__tb131072__v001",
"axis": "knowledge",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_miscellaneous_generative_spanish__6417d1b0__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 5192.520771178551, \"nll_base\": 8.554974555969238, \"eval_ref\": \"mmlu_miscellaneous_generative_spanish__6417d1b0__s000__tb65... | {
"index": "mmlu_miscellaneous_generative_spanish__6417d1b0__s000__tb65536__v000",
"uid": "mmlu_miscellaneous_generative_spanish__6417d1b0__s000__tb65536__v000",
"axis": "knowledge",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_moral_disputes__a19764e8__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 877.4338951491636, \"nll_base\": 6.777001619338989, \"eval_ref\": \"mmlu_moral_disputes__a19764e8__s000__tb262144__v000\", \"oracle_delta\": 0.... | {
"index": "mmlu_moral_disputes__a19764e8__s000__tb262144__v000",
"uid": "mmlu_moral_disputes__a19764e8__s000__tb262144__v000",
"axis": "knowledge",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_moral_disputes__a19764e8__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 877.4338951491636, \"nll_base\": 6.777001619338989, \"eval_ref\": \"mmlu_moral_disputes__a19764e8__s000__tb65536__v001\", \"oracle_delta\": 0.407... | {
"index": "mmlu_moral_disputes__a19764e8__s000__tb65536__v001",
"uid": "mmlu_moral_disputes__a19764e8__s000__tb65536__v001",
"axis": "knowledge",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_moral_disputes_continuation__ab872124__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 9.28462215924902, \"nll_base\": 2.2283595003039633, \"eval_ref\": \"mmlu_moral_disputes_continuation__ab872124__s000__tb131072__v0... | {
"index": "mmlu_moral_disputes_continuation__ab872124__s000__tb131072__v000",
"uid": "mmlu_moral_disputes_continuation__ab872124__s000__tb131072__v000",
"axis": "knowledge",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_moral_disputes_generative__b9a9c0c1__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 824.5042016684264, \"nll_base\": 6.714782238006592, \"eval_ref\": \"mmlu_moral_disputes_generative__b9a9c0c1__s000__tb262144__v000\"... | {
"index": "mmlu_moral_disputes_generative__b9a9c0c1__s000__tb262144__v000",
"uid": "mmlu_moral_disputes_generative__b9a9c0c1__s000__tb262144__v000",
"axis": "knowledge",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_moral_disputes_generative__b9a9c0c1__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 824.5042016684264, \"nll_base\": 6.714782238006592, \"eval_ref\": \"mmlu_moral_disputes_generative__b9a9c0c1__s000__tb65536__v001\", \... | {
"index": "mmlu_moral_disputes_generative__b9a9c0c1__s000__tb65536__v001",
"uid": "mmlu_moral_disputes_generative__b9a9c0c1__s000__tb65536__v001",
"axis": "knowledge",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_moral_disputes_generative_spanish__90121a25__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3740.0841279012257, \"nll_base\": 8.226863384246826, \"eval_ref\": \"mmlu_moral_disputes_generative_spanish__90121a25__s000__t... | {
"index": "mmlu_moral_disputes_generative_spanish__90121a25__s000__tb65536__v000",
"uid": "mmlu_moral_disputes_generative_spanish__90121a25__s000__tb65536__v000",
"axis": "knowledge",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_moral_scenarios__7074dfe3__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2742.9291573360206, \"nll_base\": 7.916781663894653, \"eval_ref\": \"mmlu_moral_scenarios__7074dfe3__s000__tb262144__v000\", \"oracle_delta\":... | {
"index": "mmlu_moral_scenarios__7074dfe3__s000__tb262144__v000",
"uid": "mmlu_moral_scenarios__7074dfe3__s000__tb262144__v000",
"axis": "knowledge",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_moral_scenarios__7074dfe3__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2742.9291573360206, \"nll_base\": 7.916781663894653, \"eval_ref\": \"mmlu_moral_scenarios__7074dfe3__s000__tb65536__v001\", \"oracle_delta\": 0.... | {
"index": "mmlu_moral_scenarios__7074dfe3__s000__tb65536__v001",
"uid": "mmlu_moral_scenarios__7074dfe3__s000__tb65536__v001",
"axis": "knowledge",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_moral_scenarios_continuation__23f65b3c__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 26.500782619600486, \"nll_base\": 3.277174265371204, \"eval_ref\": \"mmlu_moral_scenarios_continuation__23f65b3c__s000__tb65536__v0... | {
"index": "mmlu_moral_scenarios_continuation__23f65b3c__s000__tb65536__v001",
"uid": "mmlu_moral_scenarios_continuation__23f65b3c__s000__tb65536__v001",
"axis": "knowledge",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_moral_scenarios_generative__dab79934__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2476.391627581001, \"nll_base\": 7.814557790756226, \"eval_ref\": \"mmlu_moral_scenarios_generative__dab79934__s000__tb131072__v000... | {
"index": "mmlu_moral_scenarios_generative__dab79934__s000__tb131072__v000",
"uid": "mmlu_moral_scenarios_generative__dab79934__s000__tb131072__v000",
"axis": "knowledge",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_moral_scenarios_generative_spanish__a3ca4c67__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 12040.880542553175, \"nll_base\": 9.396062850952148, \"eval_ref\": \"mmlu_moral_scenarios_generative_spanish__a3ca4c67__s00... | {
"index": "mmlu_moral_scenarios_generative_spanish__a3ca4c67__s000__tb262144__v000",
"uid": "mmlu_moral_scenarios_generative_spanish__a3ca4c67__s000__tb262144__v000",
"axis": "knowledge",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_moral_scenarios_generative_spanish__a3ca4c67__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 12040.880542553175, \"nll_base\": 9.396062850952148, \"eval_ref\": \"mmlu_moral_scenarios_generative_spanish__a3ca4c67__s000_... | {
"index": "mmlu_moral_scenarios_generative_spanish__a3ca4c67__s000__tb65536__v001",
"uid": "mmlu_moral_scenarios_generative_spanish__a3ca4c67__s000__tb65536__v001",
"axis": "knowledge",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_nutrition__3a994573__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 912.7273858761937, \"nll_base\": 6.816437244415283, \"eval_ref\": \"mmlu_nutrition__3a994573__s000__tb131072__v000\", \"oracle_delta\": 0.6511363983... | {
"index": "mmlu_nutrition__3a994573__s000__tb131072__v000",
"uid": "mmlu_nutrition__3a994573__s000__tb131072__v000",
"axis": "knowledge",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_nutrition_continuation__07d8a6c1__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 6.05060471845801, \"nll_base\": 1.8001582205118694, \"eval_ref\": \"mmlu_nutrition_continuation__07d8a6c1__s000__tb131072__v001\", \"or... | {
"index": "mmlu_nutrition_continuation__07d8a6c1__s000__tb131072__v001",
"uid": "mmlu_nutrition_continuation__07d8a6c1__s000__tb131072__v001",
"axis": "knowledge",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_nutrition_continuation__07d8a6c1__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 6.05060471845801, \"nll_base\": 1.8001582205118694, \"eval_ref\": \"mmlu_nutrition_continuation__07d8a6c1__s000__tb65536__v000\", \"oracl... | {
"index": "mmlu_nutrition_continuation__07d8a6c1__s000__tb65536__v000",
"uid": "mmlu_nutrition_continuation__07d8a6c1__s000__tb65536__v000",
"axis": "knowledge",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_nutrition_generative__f6c24b6a__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 915.0891236187249, \"nll_base\": 6.819021463394165, \"eval_ref\": \"mmlu_nutrition_generative__f6c24b6a__s000__tb262144__v000\", \"oracle... | {
"index": "mmlu_nutrition_generative__f6c24b6a__s000__tb262144__v000",
"uid": "mmlu_nutrition_generative__f6c24b6a__s000__tb262144__v000",
"axis": "knowledge",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_nutrition_generative_spanish__98ba76f8__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3414.5755277446165, \"nll_base\": 8.13580846786499, \"eval_ref\": \"mmlu_nutrition_generative_spanish__98ba76f8__s000__tb131072__... | {
"index": "mmlu_nutrition_generative_spanish__98ba76f8__s000__tb131072__v000",
"uid": "mmlu_nutrition_generative_spanish__98ba76f8__s000__tb131072__v000",
"axis": "knowledge",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_nutrition_generative_spanish__98ba76f8__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3414.5755277446165, \"nll_base\": 8.13580846786499, \"eval_ref\": \"mmlu_nutrition_generative_spanish__98ba76f8__s000__tb262144__... | {
"index": "mmlu_nutrition_generative_spanish__98ba76f8__s000__tb262144__v001",
"uid": "mmlu_nutrition_generative_spanish__98ba76f8__s000__tb262144__v001",
"axis": "knowledge",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_philosophy__e6f6ddc5__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1218.1237141873821, \"nll_base\": 7.105067014694214, \"eval_ref\": \"mmlu_philosophy__e6f6ddc5__s000__tb131072__v000\", \"oracle_delta\": 0.8424985... | {
"index": "mmlu_philosophy__e6f6ddc5__s000__tb131072__v000",
"uid": "mmlu_philosophy__e6f6ddc5__s000__tb131072__v000",
"axis": "knowledge",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.