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_redux_college_computer_science_generative__973fc314__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1933.7536775959893, \"nll_base\": 7.56721830368042, \"eval_ref\": \"mmlu_redux_college_computer_science_generative__... | {
"index": "mmlu_redux_college_computer_science_generative__973fc314__s000__tb131072__v000",
"uid": "mmlu_redux_college_computer_science_generative__973fc314__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_redux_college_computer_science_generative__973fc314__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1933.7536775959893, \"nll_base\": 7.56721830368042, \"eval_ref\": \"mmlu_redux_college_computer_science_generative__... | {
"index": "mmlu_redux_college_computer_science_generative__973fc314__s000__tb262144__v001",
"uid": "mmlu_redux_college_computer_science_generative__973fc314__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_redux_college_mathematics_generative__e36cc287__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1740.6480541670073, \"nll_base\": 7.462012767791748, \"eval_ref\": \"mmlu_redux_college_mathematics_generative__e36cc287_... | {
"index": "mmlu_redux_college_mathematics_generative__e36cc287__s000__tb262144__v000",
"uid": "mmlu_redux_college_mathematics_generative__e36cc287__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_redux_college_mathematics_generative__e36cc287__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1740.6480541670073, \"nll_base\": 7.462012767791748, \"eval_ref\": \"mmlu_redux_college_mathematics_generative__e36cc287__s... | {
"index": "mmlu_redux_college_mathematics_generative__e36cc287__s000__tb65536__v001",
"uid": "mmlu_redux_college_mathematics_generative__e36cc287__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_redux_college_medicine_generative__2af63346__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2496.043907118928, \"nll_base\": 7.822462320327759, \"eval_ref\": \"mmlu_redux_college_medicine_generative__2af63346__s000__tb... | {
"index": "mmlu_redux_college_medicine_generative__2af63346__s000__tb65536__v001",
"uid": "mmlu_redux_college_medicine_generative__2af63346__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_redux_conceptual_physics_generative__05500900__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1696.7267001799823, \"nll_base\": 7.436456203460693, \"eval_ref\": \"mmlu_redux_conceptual_physics_generative__05500900__s... | {
"index": "mmlu_redux_conceptual_physics_generative__05500900__s000__tb262144__v000",
"uid": "mmlu_redux_conceptual_physics_generative__05500900__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_redux_conceptual_physics_generative__05500900__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1696.7267001799823, \"nll_base\": 7.436456203460693, \"eval_ref\": \"mmlu_redux_conceptual_physics_generative__05500900__s00... | {
"index": "mmlu_redux_conceptual_physics_generative__05500900__s000__tb65536__v001",
"uid": "mmlu_redux_conceptual_physics_generative__05500900__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_redux_elementary_mathematics_generative__8e724a8c__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2228.0347085970825, \"nll_base\": 7.7088751792907715, \"eval_ref\": \"mmlu_redux_elementary_mathematics_generative__8e72... | {
"index": "mmlu_redux_elementary_mathematics_generative__8e724a8c__s000__tb65536__v000",
"uid": "mmlu_redux_elementary_mathematics_generative__8e724a8c__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_redux_formal_logic_generative__4eb42234__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3020.443029796301, \"nll_base\": 8.013158798217773, \"eval_ref\": \"mmlu_redux_formal_logic_generative__4eb42234__s000__tb131072... | {
"index": "mmlu_redux_formal_logic_generative__4eb42234__s000__tb131072__v000",
"uid": "mmlu_redux_formal_logic_generative__4eb42234__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_redux_global_facts_generative__e4d644e9__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2661.7282513363675, \"nll_base\": 7.886730909347534, \"eval_ref\": \"mmlu_redux_global_facts_generative__e4d644e9__s000__tb26214... | {
"index": "mmlu_redux_global_facts_generative__e4d644e9__s000__tb262144__v000",
"uid": "mmlu_redux_global_facts_generative__e4d644e9__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_redux_global_facts_generative__e4d644e9__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2661.7282513363675, \"nll_base\": 7.886730909347534, \"eval_ref\": \"mmlu_redux_global_facts_generative__e4d644e9__s000__tb65536__... | {
"index": "mmlu_redux_global_facts_generative__e4d644e9__s000__tb65536__v001",
"uid": "mmlu_redux_global_facts_generative__e4d644e9__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_redux_high_school_chemistry_generative__fd7ebf2c__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1825.1018440386374, \"nll_base\": 7.5093910694122314, \"eval_ref\": \"mmlu_redux_high_school_chemistry_generative__fd7e... | {
"index": "mmlu_redux_high_school_chemistry_generative__fd7ebf2c__s000__tb131072__v000",
"uid": "mmlu_redux_high_school_chemistry_generative__fd7ebf2c__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_redux_high_school_computer_science_generative__c23a2573__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1191.7195374628675, \"nll_base\": 7.083152532577515, \"eval_ref\": \"mmlu_redux_high_school_computer_science_gener... | {
"index": "mmlu_redux_high_school_computer_science_generative__c23a2573__s000__tb65536__v000",
"uid": "mmlu_redux_high_school_computer_science_generative__c23a2573__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_redux_high_school_macroeconomics_generative__ac481b31__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2087.391168490026, \"nll_base\": 7.643670320510864, \"eval_ref\": \"mmlu_redux_high_school_macroeconomics_generative... | {
"index": "mmlu_redux_high_school_macroeconomics_generative__ac481b31__s000__tb65536__v001",
"uid": "mmlu_redux_high_school_macroeconomics_generative__ac481b31__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_redux_high_school_microeconomics_generative__98000977__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1566.851458591175, \"nll_base\": 7.356823444366455, \"eval_ref\": \"mmlu_redux_high_school_microeconomics_generati... | {
"index": "mmlu_redux_high_school_microeconomics_generative__98000977__s000__tb131072__v000",
"uid": "mmlu_redux_high_school_microeconomics_generative__98000977__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_redux_high_school_physics_generative__0f31cc0b__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1232.4461346087683, \"nll_base\": 7.116756200790405, \"eval_ref\": \"mmlu_redux_high_school_physics_generative__0f31cc0b__s... | {
"index": "mmlu_redux_high_school_physics_generative__0f31cc0b__s000__tb65536__v001",
"uid": "mmlu_redux_high_school_physics_generative__0f31cc0b__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_redux_high_school_world_history_generative__87b9a1f2__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2081.690659089486, \"nll_base\": 7.640935659408569, \"eval_ref\": \"mmlu_redux_high_school_world_history_generative... | {
"index": "mmlu_redux_high_school_world_history_generative__87b9a1f2__s000__tb131072__v000",
"uid": "mmlu_redux_high_school_world_history_generative__87b9a1f2__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_redux_human_sexuality_generative__63afc650__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3737.652346086899, \"nll_base\": 8.226212978363037, \"eval_ref\": \"mmlu_redux_human_sexuality_generative__63afc650__s000__tb... | {
"index": "mmlu_redux_human_sexuality_generative__63afc650__s000__tb262144__v000",
"uid": "mmlu_redux_human_sexuality_generative__63afc650__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_redux_human_sexuality_generative__63afc650__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3737.652346086899, \"nll_base\": 8.226212978363037, \"eval_ref\": \"mmlu_redux_human_sexuality_generative__63afc650__s000__tb65... | {
"index": "mmlu_redux_human_sexuality_generative__63afc650__s000__tb65536__v001",
"uid": "mmlu_redux_human_sexuality_generative__63afc650__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_redux_jurisprudence_generative__5a00fefd__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1965.8540486628133, \"nll_base\": 7.583682060241699, \"eval_ref\": \"mmlu_redux_jurisprudence_generative__5a00fefd__s000__tb65536... | {
"index": "mmlu_redux_jurisprudence_generative__5a00fefd__s000__tb65536__v000",
"uid": "mmlu_redux_jurisprudence_generative__5a00fefd__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_redux_logical_fallacies_generative__fb3bdfd8__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1720.546091746334, \"nll_base\": 7.45039701461792, \"eval_ref\": \"mmlu_redux_logical_fallacies_generative__fb3bdfd8__s000_... | {
"index": "mmlu_redux_logical_fallacies_generative__fb3bdfd8__s000__tb131072__v000",
"uid": "mmlu_redux_logical_fallacies_generative__fb3bdfd8__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_redux_marketing_generative__bc3353c6__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2091.4631127367825, \"nll_base\": 7.64561915397644, \"eval_ref\": \"mmlu_redux_marketing_generative__bc3353c6__s000__tb131072__v001... | {
"index": "mmlu_redux_marketing_generative__bc3353c6__s000__tb131072__v001",
"uid": "mmlu_redux_marketing_generative__bc3353c6__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_redux_marketing_generative__bc3353c6__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2091.4631127367825, \"nll_base\": 7.64561915397644, \"eval_ref\": \"mmlu_redux_marketing_generative__bc3353c6__s000__tb65536__v000\",... | {
"index": "mmlu_redux_marketing_generative__bc3353c6__s000__tb65536__v000",
"uid": "mmlu_redux_marketing_generative__bc3353c6__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_redux_medical_genetics_generative__7bb646af__s000__tb131072__v002\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2410.566275943501, \"nll_base\": 7.787616968154907, \"eval_ref\": \"mmlu_redux_medical_genetics_generative__7bb646af__s000__... | {
"index": "mmlu_redux_medical_genetics_generative__7bb646af__s000__tb131072__v002",
"uid": "mmlu_redux_medical_genetics_generative__7bb646af__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_redux_medical_genetics_generative__7bb646af__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2410.566275943501, \"nll_base\": 7.787616968154907, \"eval_ref\": \"mmlu_redux_medical_genetics_generative__7bb646af__s000__tb... | {
"index": "mmlu_redux_medical_genetics_generative__7bb646af__s000__tb65536__v001",
"uid": "mmlu_redux_medical_genetics_generative__7bb646af__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_redux_miscellaneous_generative__90f3426a__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3546.8519955045126, \"nll_base\": 8.173815727233887, \"eval_ref\": \"mmlu_redux_miscellaneous_generative__90f3426a__s000__tb262... | {
"index": "mmlu_redux_miscellaneous_generative__90f3426a__s000__tb262144__v000",
"uid": "mmlu_redux_miscellaneous_generative__90f3426a__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_redux_miscellaneous_generative__90f3426a__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3546.8519955045126, \"nll_base\": 8.173815727233887, \"eval_ref\": \"mmlu_redux_miscellaneous_generative__90f3426a__s000__tb65536... | {
"index": "mmlu_redux_miscellaneous_generative__90f3426a__s000__tb65536__v001",
"uid": "mmlu_redux_miscellaneous_generative__90f3426a__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_redux_moral_disputes_generative__2906765d__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1710.8959880943046, \"nll_base\": 7.444772481918335, \"eval_ref\": \"mmlu_redux_moral_disputes_generative__2906765d__s000__tb1... | {
"index": "mmlu_redux_moral_disputes_generative__2906765d__s000__tb131072__v001",
"uid": "mmlu_redux_moral_disputes_generative__2906765d__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_redux_moral_disputes_generative__2906765d__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1710.8959880943046, \"nll_base\": 7.444772481918335, \"eval_ref\": \"mmlu_redux_moral_disputes_generative__2906765d__s000__tb655... | {
"index": "mmlu_redux_moral_disputes_generative__2906765d__s000__tb65536__v000",
"uid": "mmlu_redux_moral_disputes_generative__2906765d__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_redux_moral_scenarios_generative__62d142c5__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 6555.412740821174, \"nll_base\": 8.78804636001587, \"eval_ref\": \"mmlu_redux_moral_scenarios_generative__62d142c5__s000__tb1... | {
"index": "mmlu_redux_moral_scenarios_generative__62d142c5__s000__tb131072__v001",
"uid": "mmlu_redux_moral_scenarios_generative__62d142c5__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_redux_moral_scenarios_generative__62d142c5__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 6555.412740821174, \"nll_base\": 8.78804636001587, \"eval_ref\": \"mmlu_redux_moral_scenarios_generative__62d142c5__s000__tb655... | {
"index": "mmlu_redux_moral_scenarios_generative__62d142c5__s000__tb65536__v000",
"uid": "mmlu_redux_moral_scenarios_generative__62d142c5__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_redux_nutrition_generative__7cd12067__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2325.072655013427, \"nll_base\": 7.751506567001343, \"eval_ref\": \"mmlu_redux_nutrition_generative__7cd12067__s000__tb65536__v001\",... | {
"index": "mmlu_redux_nutrition_generative__7cd12067__s000__tb65536__v001",
"uid": "mmlu_redux_nutrition_generative__7cd12067__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_redux_philosophy_generative__ef09aefb__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2690.4732368210903, \"nll_base\": 7.897472381591797, \"eval_ref\": \"mmlu_redux_philosophy_generative__ef09aefb__s000__tb262144__v... | {
"index": "mmlu_redux_philosophy_generative__ef09aefb__s000__tb262144__v000",
"uid": "mmlu_redux_philosophy_generative__ef09aefb__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_redux_philosophy_generative__ef09aefb__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2690.4732368210903, \"nll_base\": 7.897472381591797, \"eval_ref\": \"mmlu_redux_philosophy_generative__ef09aefb__s000__tb65536__v001... | {
"index": "mmlu_redux_philosophy_generative__ef09aefb__s000__tb65536__v001",
"uid": "mmlu_redux_philosophy_generative__ef09aefb__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_redux_prehistory_generative__5e5c01aa__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2994.089581654692, \"nll_base\": 8.004395484924316, \"eval_ref\": \"mmlu_redux_prehistory_generative__5e5c01aa__s000__tb65536__v000\... | {
"index": "mmlu_redux_prehistory_generative__5e5c01aa__s000__tb65536__v000",
"uid": "mmlu_redux_prehistory_generative__5e5c01aa__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_redux_professional_medicine_generative__9eccb865__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2173.5365791591626, \"nll_base\": 7.684110879898071, \"eval_ref\": \"mmlu_redux_professional_medicine_generative__9eccb... | {
"index": "mmlu_redux_professional_medicine_generative__9eccb865__s000__tb262144__v000",
"uid": "mmlu_redux_professional_medicine_generative__9eccb865__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_redux_professional_medicine_generative__9eccb865__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2173.5365791591626, \"nll_base\": 7.684110879898071, \"eval_ref\": \"mmlu_redux_professional_medicine_generative__9eccb86... | {
"index": "mmlu_redux_professional_medicine_generative__9eccb865__s000__tb65536__v001",
"uid": "mmlu_redux_professional_medicine_generative__9eccb865__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_redux_public_relations_generative__d511c66f__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2689.1021473718797, \"nll_base\": 7.896962642669678, \"eval_ref\": \"mmlu_redux_public_relations_generative__d511c66f__s000__t... | {
"index": "mmlu_redux_public_relations_generative__d511c66f__s000__tb65536__v000",
"uid": "mmlu_redux_public_relations_generative__d511c66f__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_redux_sociology_generative__8bb59f2c__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2465.8156860105305, \"nll_base\": 7.810277938842773, \"eval_ref\": \"mmlu_redux_sociology_generative__8bb59f2c__s000__tb65536__v000\"... | {
"index": "mmlu_redux_sociology_generative__8bb59f2c__s000__tb65536__v000",
"uid": "mmlu_redux_sociology_generative__8bb59f2c__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_redux_virology_generative__a3a2e721__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1973.6644083335675, \"nll_base\": 7.587647199630737, \"eval_ref\": \"mmlu_redux_virology_generative__a3a2e721__s000__tb65536__v000\", ... | {
"index": "mmlu_redux_virology_generative__a3a2e721__s000__tb65536__v000",
"uid": "mmlu_redux_virology_generative__a3a2e721__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_security_studies_generative_spanish__95f747ee__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2912.136059065996, \"nll_base\": 7.97664213180542, \"eval_ref\": \"mmlu_security_studies_generative_spanish__95f747ee__s00... | {
"index": "mmlu_security_studies_generative_spanish__95f747ee__s000__tb262144__v000",
"uid": "mmlu_security_studies_generative_spanish__95f747ee__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_security_studies_generative_spanish__95f747ee__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2912.136059065996, \"nll_base\": 7.97664213180542, \"eval_ref\": \"mmlu_security_studies_generative_spanish__95f747ee__s000_... | {
"index": "mmlu_security_studies_generative_spanish__95f747ee__s000__tb65536__v001",
"uid": "mmlu_security_studies_generative_spanish__95f747ee__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_sociology__8a3f5be1__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 800.1127879860272, \"nll_base\": 6.684752702713013, \"eval_ref\": \"mmlu_sociology__8a3f5be1__s000__tb131072__v001\", \"oracle_delta\": 0.5441663265... | {
"index": "mmlu_sociology__8a3f5be1__s000__tb131072__v001",
"uid": "mmlu_sociology__8a3f5be1__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_sociology__8a3f5be1__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 800.1127879860272, \"nll_base\": 6.684752702713013, \"eval_ref\": \"mmlu_sociology__8a3f5be1__s000__tb65536__v000\", \"oracle_delta\": 0.1710584163665... | {
"index": "mmlu_sociology__8a3f5be1__s000__tb65536__v000",
"uid": "mmlu_sociology__8a3f5be1__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_sociology_continuation__cc81744a__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 6.7974335241580315, \"nll_base\": 1.9165451180221689, \"eval_ref\": \"mmlu_sociology_continuation__cc81744a__s000__tb262144__v000\", \"... | {
"index": "mmlu_sociology_continuation__cc81744a__s000__tb262144__v000",
"uid": "mmlu_sociology_continuation__cc81744a__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_sociology_continuation__cc81744a__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 6.7974335241580315, \"nll_base\": 1.9165451180221689, \"eval_ref\": \"mmlu_sociology_continuation__cc81744a__s000__tb65536__v001\", \"ora... | {
"index": "mmlu_sociology_continuation__cc81744a__s000__tb65536__v001",
"uid": "mmlu_sociology_continuation__cc81744a__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_sociology_generative__dc7ed342__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 895.0510198292311, \"nll_base\": 6.796880722045898, \"eval_ref\": \"mmlu_sociology_generative__dc7ed342__s000__tb262144__v000\", \"oracle... | {
"index": "mmlu_sociology_generative__dc7ed342__s000__tb262144__v000",
"uid": "mmlu_sociology_generative__dc7ed342__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_th_llama_clinical_knowledge__2930f1a0__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 218.77275637958564, \"nll_base\": 5.388033548990886, \"eval_ref\": \"mmlu_th_llama_clinical_knowledge__2930f1a0__s000__tb131072__v... | {
"index": "mmlu_th_llama_clinical_knowledge__2930f1a0__s000__tb131072__v001",
"uid": "mmlu_th_llama_clinical_knowledge__2930f1a0__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_th_llama_clinical_knowledge__2930f1a0__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 218.77275637958564, \"nll_base\": 5.388033548990886, \"eval_ref\": \"mmlu_th_llama_clinical_knowledge__2930f1a0__s000__tb65536__v000... | {
"index": "mmlu_th_llama_clinical_knowledge__2930f1a0__s000__tb65536__v000",
"uid": "mmlu_th_llama_clinical_knowledge__2930f1a0__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_th_llama_college_computer_science__fddb735d__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 197.23354657438804, \"nll_base\": 5.284388542175293, \"eval_ref\": \"mmlu_th_llama_college_computer_science__fddb735d__s000_... | {
"index": "mmlu_th_llama_college_computer_science__fddb735d__s000__tb131072__v000",
"uid": "mmlu_th_llama_college_computer_science__fddb735d__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_th_llama_computer_security__60b14a6e__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 201.42700837218115, \"nll_base\": 5.305427074432373, \"eval_ref\": \"mmlu_th_llama_computer_security__60b14a6e__s000__tb65536__v000\"... | {
"index": "mmlu_th_llama_computer_security__60b14a6e__s000__tb65536__v000",
"uid": "mmlu_th_llama_computer_security__60b14a6e__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_th_llama_econometrics__c1e4e82c__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 192.118572874027, \"nll_base\": 5.258112748463948, \"eval_ref\": \"mmlu_th_llama_econometrics__c1e4e82c__s000__tb262144__v000\", \"oracl... | {
"index": "mmlu_th_llama_econometrics__c1e4e82c__s000__tb262144__v000",
"uid": "mmlu_th_llama_econometrics__c1e4e82c__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_th_llama_electrical_engineering__0dc94c82__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 234.0501879502667, \"nll_base\": 5.4555355707804365, \"eval_ref\": \"mmlu_th_llama_electrical_engineering__0dc94c82__s000__tb655... | {
"index": "mmlu_th_llama_electrical_engineering__0dc94c82__s000__tb65536__v000",
"uid": "mmlu_th_llama_electrical_engineering__0dc94c82__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_th_llama_global_facts__5bc84897__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 275.04562735167707, \"nll_base\": 5.616937001546224, \"eval_ref\": \"mmlu_th_llama_global_facts__5bc84897__s000__tb131072__v000\", \"ora... | {
"index": "mmlu_th_llama_global_facts__5bc84897__s000__tb131072__v000",
"uid": "mmlu_th_llama_global_facts__5bc84897__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_th_llama_high_school_chemistry__841ba610__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 224.85001181785486, \"nll_base\": 5.415433565775554, \"eval_ref\": \"mmlu_th_llama_high_school_chemistry__841ba610__s000__tb262... | {
"index": "mmlu_th_llama_high_school_chemistry__841ba610__s000__tb262144__v000",
"uid": "mmlu_th_llama_high_school_chemistry__841ba610__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_th_llama_high_school_mathematics__07d19171__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 260.6662894996453, \"nll_base\": 5.563241004943848, \"eval_ref\": \"mmlu_th_llama_high_school_mathematics__07d19171__s000__tb... | {
"index": "mmlu_th_llama_high_school_mathematics__07d19171__s000__tb262144__v000",
"uid": "mmlu_th_llama_high_school_mathematics__07d19171__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_th_llama_high_school_statistics__0d723e1b__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 197.63170946948404, \"nll_base\": 5.286405245463054, \"eval_ref\": \"mmlu_th_llama_high_school_statistics__0d723e1b__s000__tb2... | {
"index": "mmlu_th_llama_high_school_statistics__0d723e1b__s000__tb262144__v000",
"uid": "mmlu_th_llama_high_school_statistics__0d723e1b__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_th_llama_human_aging__5c455f4d__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 232.25958139650817, \"nll_base\": 5.447855631510417, \"eval_ref\": \"mmlu_th_llama_human_aging__5c455f4d__s000__tb131072__v000\", \"oracl... | {
"index": "mmlu_th_llama_human_aging__5c455f4d__s000__tb131072__v000",
"uid": "mmlu_th_llama_human_aging__5c455f4d__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_th_llama_jurisprudence__a470d00d__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 188.46385249913328, \"nll_base\": 5.238906224568685, \"eval_ref\": \"mmlu_th_llama_jurisprudence__a470d00d__s000__tb262144__v000\", \"o... | {
"index": "mmlu_th_llama_jurisprudence__a470d00d__s000__tb262144__v000",
"uid": "mmlu_th_llama_jurisprudence__a470d00d__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_th_llama_logical_fallacies__7b1a551b__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 249.31348738473122, \"nll_base\": 5.518711090087891, \"eval_ref\": \"mmlu_th_llama_logical_fallacies__7b1a551b__s000__tb65536__v000\"... | {
"index": "mmlu_th_llama_logical_fallacies__7b1a551b__s000__tb65536__v000",
"uid": "mmlu_th_llama_logical_fallacies__7b1a551b__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_th_llama_moral_scenarios__d993a252__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 246.11908544664462, \"nll_base\": 5.505815505981445, \"eval_ref\": \"mmlu_th_llama_moral_scenarios__d993a252__s000__tb262144__v000\",... | {
"index": "mmlu_th_llama_moral_scenarios__d993a252__s000__tb262144__v000",
"uid": "mmlu_th_llama_moral_scenarios__d993a252__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_th_llama_nutrition__39e6e5ae__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 221.86796290391501, \"nll_base\": 5.402082443237305, \"eval_ref\": \"mmlu_th_llama_nutrition__39e6e5ae__s000__tb131072__v000\", \"oracle_de... | {
"index": "mmlu_th_llama_nutrition__39e6e5ae__s000__tb131072__v000",
"uid": "mmlu_th_llama_nutrition__39e6e5ae__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_th_llama_professional_accounting__84bddc23__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 206.33819991855833, \"nll_base\": 5.329516569773356, \"eval_ref\": \"mmlu_th_llama_professional_accounting__84bddc23__s000__t... | {
"index": "mmlu_th_llama_professional_accounting__84bddc23__s000__tb131072__v000",
"uid": "mmlu_th_llama_professional_accounting__84bddc23__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_th_llama_virology__f09ee4d2__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 214.27313974537265, \"nll_base\": 5.367251555124919, \"eval_ref\": \"mmlu_th_llama_virology__f09ee4d2__s000__tb65536__v000\", \"oracle_delta\"... | {
"index": "mmlu_th_llama_virology__f09ee4d2__s000__tb65536__v000",
"uid": "mmlu_th_llama_virology__f09ee4d2__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_th_llama_world_religions__599cf525__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 219.8466000891158, \"nll_base\": 5.392930030822754, \"eval_ref\": \"mmlu_th_llama_world_religions__599cf525__s000__tb131072__v000\", ... | {
"index": "mmlu_th_llama_world_religions__599cf525__s000__tb131072__v000",
"uid": "mmlu_th_llama_world_religions__599cf525__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_us_foreign_policy__5e886108__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 947.0745563385334, \"nll_base\": 6.853377819061279, \"eval_ref\": \"mmlu_us_foreign_policy__5e886108__s000__tb65536__v001\", \"oracle_delta\":... | {
"index": "mmlu_us_foreign_policy__5e886108__s000__tb65536__v001",
"uid": "mmlu_us_foreign_policy__5e886108__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_us_foreign_policy_generative__02e22763__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 829.9125972281174, \"nll_base\": 6.721320390701294, \"eval_ref\": \"mmlu_us_foreign_policy_generative__02e22763__s000__tb131072__... | {
"index": "mmlu_us_foreign_policy_generative__02e22763__s000__tb131072__v000",
"uid": "mmlu_us_foreign_policy_generative__02e22763__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_us_foreign_policy_generative__02e22763__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 829.9125972281174, \"nll_base\": 6.721320390701294, \"eval_ref\": \"mmlu_us_foreign_policy_generative__02e22763__s000__tb262144__... | {
"index": "mmlu_us_foreign_policy_generative__02e22763__s000__tb262144__v001",
"uid": "mmlu_us_foreign_policy_generative__02e22763__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_us_foreign_policy_generative_spanish__fecdeb8a__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 5504.295745624716, \"nll_base\": 8.61328411102295, \"eval_ref\": \"mmlu_us_foreign_policy_generative_spanish__fecdeb8a__s... | {
"index": "mmlu_us_foreign_policy_generative_spanish__fecdeb8a__s000__tb131072__v001",
"uid": "mmlu_us_foreign_policy_generative_spanish__fecdeb8a__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_us_foreign_policy_generative_spanish__fecdeb8a__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 5504.295745624716, \"nll_base\": 8.61328411102295, \"eval_ref\": \"mmlu_us_foreign_policy_generative_spanish__fecdeb8a__s00... | {
"index": "mmlu_us_foreign_policy_generative_spanish__fecdeb8a__s000__tb65536__v000",
"uid": "mmlu_us_foreign_policy_generative_spanish__fecdeb8a__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_virology__5aba30f6__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 712.6624410064854, \"nll_base\": 6.569007873535156, \"eval_ref\": \"mmlu_virology__5aba30f6__s000__tb262144__v000\", \"oracle_delta\": 0.722830295562... | {
"index": "mmlu_virology__5aba30f6__s000__tb262144__v000",
"uid": "mmlu_virology__5aba30f6__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_virology__5aba30f6__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 712.6624410064854, \"nll_base\": 6.569007873535156, \"eval_ref\": \"mmlu_virology__5aba30f6__s000__tb65536__v001\", \"oracle_delta\": 0.344781637191772... | {
"index": "mmlu_virology__5aba30f6__s000__tb65536__v001",
"uid": "mmlu_virology__5aba30f6__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_virology_continuation__d22d5f8b__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 13.242337021482673, \"nll_base\": 2.5834190471289937, \"eval_ref\": \"mmlu_virology_continuation__d22d5f8b__s000__tb262144__v000\", \"or... | {
"index": "mmlu_virology_continuation__d22d5f8b__s000__tb262144__v000",
"uid": "mmlu_virology_continuation__d22d5f8b__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_virology_continuation__d22d5f8b__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 13.242337021482673, \"nll_base\": 2.5834190471289937, \"eval_ref\": \"mmlu_virology_continuation__d22d5f8b__s000__tb65536__v001\", \"oracl... | {
"index": "mmlu_virology_continuation__d22d5f8b__s000__tb65536__v001",
"uid": "mmlu_virology_continuation__d22d5f8b__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_virology_generative__62d12586__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 968.6317734675439, \"nll_base\": 6.875884532928467, \"eval_ref\": \"mmlu_virology_generative__62d12586__s000__tb131072__v000\", \"oracle_d... | {
"index": "mmlu_virology_generative__62d12586__s000__tb131072__v000",
"uid": "mmlu_virology_generative__62d12586__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_world_religions_continuation__49988ac6__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 57.246143199727335, \"nll_base\": 4.047360272474692, \"eval_ref\": \"mmlu_world_religions_continuation__49988ac6__s000__tb65536__v0... | {
"index": "mmlu_world_religions_continuation__49988ac6__s000__tb65536__v000",
"uid": "mmlu_world_religions_continuation__49988ac6__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_world_religions_generative__44ec5ad1__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3201.1051622426453, \"nll_base\": 8.071251392364502, \"eval_ref\": \"mmlu_world_religions_generative__44ec5ad1__s000__tb65536__v000\"... | {
"index": "mmlu_world_religions_generative__44ec5ad1__s000__tb65536__v000",
"uid": "mmlu_world_religions_generative__44ec5ad1__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_world_religions_generative_spanish__e212cda6__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 6318.619243073494, \"nll_base\": 8.751255989074707, \"eval_ref\": \"mmlu_world_religions_generative_spanish__e212cda6__s000__... | {
"index": "mmlu_world_religions_generative_spanish__e212cda6__s000__tb65536__v000",
"uid": "mmlu_world_religions_generative_spanish__e212cda6__s000__tb65536__v000",
"axis": "knowledge",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlusr_answer_only_anatomy__d1cf47cd__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 568.4359225661076, \"nll_base\": 6.342888593673706, \"eval_ref\": \"mmlusr_answer_only_anatomy__d1cf47cd__s000__tb65536__v000\", \"oracle_... | {
"index": "mmlusr_answer_only_anatomy__d1cf47cd__s000__tb65536__v000",
"uid": "mmlusr_answer_only_anatomy__d1cf47cd__s000__tb65536__v000",
"axis": "knowledge",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlusr_answer_only_astronomy__c40e25bb__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 864.8099941366428, \"nll_base\": 6.762509822845459, \"eval_ref\": \"mmlusr_answer_only_astronomy__c40e25bb__s000__tb262144__v000\", \"... | {
"index": "mmlusr_answer_only_astronomy__c40e25bb__s000__tb262144__v000",
"uid": "mmlusr_answer_only_astronomy__c40e25bb__s000__tb262144__v000",
"axis": "knowledge",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlusr_answer_only_astronomy__c40e25bb__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 864.8099941366428, \"nll_base\": 6.762509822845459, \"eval_ref\": \"mmlusr_answer_only_astronomy__c40e25bb__s000__tb65536__v001\", \"ora... | {
"index": "mmlusr_answer_only_astronomy__c40e25bb__s000__tb65536__v001",
"uid": "mmlusr_answer_only_astronomy__c40e25bb__s000__tb65536__v001",
"axis": "knowledge",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlusr_answer_only_clinical_knowledge__fed34f54__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 676.1920328745573, \"nll_base\": 6.516477108001709, \"eval_ref\": \"mmlusr_answer_only_clinical_knowledge__fed34f54__s000__tb... | {
"index": "mmlusr_answer_only_clinical_knowledge__fed34f54__s000__tb131072__v000",
"uid": "mmlusr_answer_only_clinical_knowledge__fed34f54__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\": \"mmlusr_answer_only_clinical_knowledge__fed34f54__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 676.1920328745573, \"nll_base\": 6.516477108001709, \"eval_ref\": \"mmlusr_answer_only_clinical_knowledge__fed34f54__s000__tb... | {
"index": "mmlusr_answer_only_clinical_knowledge__fed34f54__s000__tb262144__v001",
"uid": "mmlusr_answer_only_clinical_knowledge__fed34f54__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\": \"mmlusr_answer_only_clinical_knowledge__fed34f54__s000__tb65536__v002\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 676.1920328745573, \"nll_base\": 6.516477108001709, \"eval_ref\": \"mmlusr_answer_only_clinical_knowledge__fed34f54__s000__tb65... | {
"index": "mmlusr_answer_only_clinical_knowledge__fed34f54__s000__tb65536__v002",
"uid": "mmlusr_answer_only_clinical_knowledge__fed34f54__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\": \"mmlusr_answer_only_college_biology__c4ba1b34__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 614.659337474136, \"nll_base\": 6.42106819152832, \"eval_ref\": \"mmlusr_answer_only_college_biology__c4ba1b34__s000__tb131072__... | {
"index": "mmlusr_answer_only_college_biology__c4ba1b34__s000__tb131072__v001",
"uid": "mmlusr_answer_only_college_biology__c4ba1b34__s000__tb131072__v001",
"axis": "knowledge",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlusr_answer_only_college_biology__c4ba1b34__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 614.659337474136, \"nll_base\": 6.42106819152832, \"eval_ref\": \"mmlusr_answer_only_college_biology__c4ba1b34__s000__tb65536__v00... | {
"index": "mmlusr_answer_only_college_biology__c4ba1b34__s000__tb65536__v000",
"uid": "mmlusr_answer_only_college_biology__c4ba1b34__s000__tb65536__v000",
"axis": "knowledge",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlusr_answer_only_college_computer_science__0e3c1897__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 720.809378691307, \"nll_base\": 6.580374717712402, \"eval_ref\": \"mmlusr_answer_only_college_computer_science__0e3c189... | {
"index": "mmlusr_answer_only_college_computer_science__0e3c1897__s000__tb131072__v000",
"uid": "mmlusr_answer_only_college_computer_science__0e3c1897__s000__tb131072__v000",
"axis": "knowledge",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlusr_answer_only_college_computer_science__0e3c1897__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 720.809378691307, \"nll_base\": 6.580374717712402, \"eval_ref\": \"mmlusr_answer_only_college_computer_science__0e3c189... | {
"index": "mmlusr_answer_only_college_computer_science__0e3c1897__s000__tb262144__v001",
"uid": "mmlusr_answer_only_college_computer_science__0e3c1897__s000__tb262144__v001",
"axis": "knowledge",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlusr_answer_only_college_mathematics__6f41ce93__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 866.3648012514358, \"nll_base\": 6.76430606842041, \"eval_ref\": \"mmlusr_answer_only_college_mathematics__6f41ce93__s000__t... | {
"index": "mmlusr_answer_only_college_mathematics__6f41ce93__s000__tb262144__v000",
"uid": "mmlusr_answer_only_college_mathematics__6f41ce93__s000__tb262144__v000",
"axis": "knowledge",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlusr_answer_only_college_medicine__358bb393__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 856.1237571382567, \"nll_base\": 6.75241494178772, \"eval_ref\": \"mmlusr_answer_only_college_medicine__358bb393__s000__tb65536__... | {
"index": "mmlusr_answer_only_college_medicine__358bb393__s000__tb65536__v000",
"uid": "mmlusr_answer_only_college_medicine__358bb393__s000__tb65536__v000",
"axis": "knowledge",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlusr_answer_only_college_physics__417c3d83__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 645.1574159812934, \"nll_base\": 6.469494342803955, \"eval_ref\": \"mmlusr_answer_only_college_physics__417c3d83__s000__tb131072... | {
"index": "mmlusr_answer_only_college_physics__417c3d83__s000__tb131072__v001",
"uid": "mmlusr_answer_only_college_physics__417c3d83__s000__tb131072__v001",
"axis": "knowledge",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlusr_answer_only_college_physics__417c3d83__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 645.1574159812934, \"nll_base\": 6.469494342803955, \"eval_ref\": \"mmlusr_answer_only_college_physics__417c3d83__s000__tb65536__v... | {
"index": "mmlusr_answer_only_college_physics__417c3d83__s000__tb65536__v000",
"uid": "mmlusr_answer_only_college_physics__417c3d83__s000__tb65536__v000",
"axis": "knowledge",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlusr_answer_only_computer_security__5bcb6ba7__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 826.5866129046789, \"nll_base\": 6.717304706573486, \"eval_ref\": \"mmlusr_answer_only_computer_security__5bcb6ba7__s000__tb6553... | {
"index": "mmlusr_answer_only_computer_security__5bcb6ba7__s000__tb65536__v000",
"uid": "mmlusr_answer_only_computer_security__5bcb6ba7__s000__tb65536__v000",
"axis": "knowledge",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlusr_answer_only_conceptual_physics__83123649__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 598.8397424272343, \"nll_base\": 6.394994020462036, \"eval_ref\": \"mmlusr_answer_only_conceptual_physics__83123649__s000__tb... | {
"index": "mmlusr_answer_only_conceptual_physics__83123649__s000__tb262144__v000",
"uid": "mmlusr_answer_only_conceptual_physics__83123649__s000__tb262144__v000",
"axis": "knowledge",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlusr_answer_only_conceptual_physics__83123649__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 598.8397424272343, \"nll_base\": 6.394994020462036, \"eval_ref\": \"mmlusr_answer_only_conceptual_physics__83123649__s000__tb65... | {
"index": "mmlusr_answer_only_conceptual_physics__83123649__s000__tb65536__v001",
"uid": "mmlusr_answer_only_conceptual_physics__83123649__s000__tb65536__v001",
"axis": "knowledge",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlusr_answer_only_electrical_engineering__dadbeb6e__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 637.9348853132946, \"nll_base\": 6.458236217498779, \"eval_ref\": \"mmlusr_answer_only_electrical_engineering__dadbeb6e__... | {
"index": "mmlusr_answer_only_electrical_engineering__dadbeb6e__s000__tb131072__v000",
"uid": "mmlusr_answer_only_electrical_engineering__dadbeb6e__s000__tb131072__v000",
"axis": "knowledge",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlusr_answer_only_elementary_mathematics__d60234ec__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 934.9566049077616, \"nll_base\": 6.840500116348267, \"eval_ref\": \"mmlusr_answer_only_elementary_mathematics__d60234ec__... | {
"index": "mmlusr_answer_only_elementary_mathematics__d60234ec__s000__tb131072__v000",
"uid": "mmlusr_answer_only_elementary_mathematics__d60234ec__s000__tb131072__v000",
"axis": "knowledge",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlusr_answer_only_global_facts__14619982__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1002.9051465612506, \"nll_base\": 6.910656213760376, \"eval_ref\": \"mmlusr_answer_only_global_facts__14619982__s000__tb262144__v00... | {
"index": "mmlusr_answer_only_global_facts__14619982__s000__tb262144__v000",
"uid": "mmlusr_answer_only_global_facts__14619982__s000__tb262144__v000",
"axis": "knowledge",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlusr_answer_only_global_facts__14619982__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1002.9051465612506, \"nll_base\": 6.910656213760376, \"eval_ref\": \"mmlusr_answer_only_global_facts__14619982__s000__tb65536__v001\"... | {
"index": "mmlusr_answer_only_global_facts__14619982__s000__tb65536__v001",
"uid": "mmlusr_answer_only_global_facts__14619982__s000__tb65536__v001",
"axis": "knowledge",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlusr_answer_only_high_school_biology__67fd5858__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 430.27119385884504, \"nll_base\": 6.064415693283081, \"eval_ref\": \"mmlusr_answer_only_high_school_biology__67fd5858__s000_... | {
"index": "mmlusr_answer_only_high_school_biology__67fd5858__s000__tb262144__v000",
"uid": "mmlusr_answer_only_high_school_biology__67fd5858__s000__tb262144__v000",
"axis": "knowledge",
"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.