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\": \"agieval_lsat_ar__d837162a__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4.461145134237497, \"nll_base\": 1.495405489631142, \"eval_ref\": \"agieval_lsat_ar__d837162a__s000__tb65536__v000\", \"oracle_delta\": 0.12888647619... | {
"index": "agieval_lsat_ar__d837162a__s000__tb65536__v000",
"uid": "agieval_lsat_ar__d837162a__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\": \"agieval_lsat_lr__fb28d856__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2.275343773718684, \"nll_base\": 0.8221311503139435, \"eval_ref\": \"agieval_lsat_lr__fb28d856__s000__tb131072__v000\", \"oracle_delta\": 0.1504358... | {
"index": "agieval_lsat_lr__fb28d856__s000__tb131072__v000",
"uid": "agieval_lsat_lr__fb28d856__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\": \"agieval_lsat_lr__fb28d856__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2.275343773718684, \"nll_base\": 0.8221311503139435, \"eval_ref\": \"agieval_lsat_lr__fb28d856__s000__tb262144__v001\", \"oracle_delta\": 0.1763543... | {
"index": "agieval_lsat_lr__fb28d856__s000__tb262144__v001",
"uid": "agieval_lsat_lr__fb28d856__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\": \"agieval_math__af8894f0__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 37.713292620792636, \"nll_base\": 3.6300126217198714, \"eval_ref\": \"agieval_math__af8894f0__s000__tb131072__v001\", \"oracle_delta\": 0.197360536128... | {
"index": "agieval_math__af8894f0__s000__tb131072__v001",
"uid": "agieval_math__af8894f0__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\": \"agieval_math__af8894f0__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 37.713292620792636, \"nll_base\": 3.6300126217198714, \"eval_ref\": \"agieval_math__af8894f0__s000__tb65536__v000\", \"oracle_delta\": 0.090582596628289... | {
"index": "agieval_math__af8894f0__s000__tb65536__v000",
"uid": "agieval_math__af8894f0__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\": \"agieval_sat_en__ca1faa27__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2.643268962544357, \"nll_base\": 0.9720163945551519, \"eval_ref\": \"agieval_sat_en__ca1faa27__s000__tb131072__v001\", \"oracle_delta\": 0.090019892... | {
"index": "agieval_sat_en__ca1faa27__s000__tb131072__v001",
"uid": "agieval_sat_en__ca1faa27__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\": \"agieval_sat_en__ca1faa27__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2.643268962544357, \"nll_base\": 0.9720163945551519, \"eval_ref\": \"agieval_sat_en__ca1faa27__s000__tb65536__v000\", \"oracle_delta\": 0.053556162160... | {
"index": "agieval_sat_en__ca1faa27__s000__tb65536__v000",
"uid": "agieval_sat_en__ca1faa27__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\": \"agieval_sat_en_without_passage__651df265__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3.259089463190163, \"nll_base\": 1.1814478505956065, \"eval_ref\": \"agieval_sat_en_without_passage__651df265__s000__tb65536__v001\", ... | {
"index": "agieval_sat_en_without_passage__651df265__s000__tb65536__v001",
"uid": "agieval_sat_en_without_passage__651df265__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\": \"agieval_sat_math__d14a1cd5__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 6.631557588529889, \"nll_base\": 1.8918397070026118, \"eval_ref\": \"agieval_sat_math__d14a1cd5__s000__tb262144__v000\", \"oracle_delta\": 0.45004... | {
"index": "agieval_sat_math__d14a1cd5__s000__tb262144__v000",
"uid": "agieval_sat_math__d14a1cd5__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\": \"agieval_sat_math__d14a1cd5__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 6.631557588529889, \"nll_base\": 1.8918397070026118, \"eval_ref\": \"agieval_sat_math__d14a1cd5__s000__tb65536__v001\", \"oracle_delta\": 0.20530123... | {
"index": "agieval_sat_math__d14a1cd5__s000__tb65536__v001",
"uid": "agieval_sat_math__d14a1cd5__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\": \"aime__2c85af57__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 208.90139457337318, \"nll_base\": 5.3418623443342685, \"eval_ref\": \"aime__2c85af57__s000__tb131072__v001\", \"oracle_delta\": 0.3841949202056627, \"random_d... | {
"index": "aime__2c85af57__s000__tb131072__v001",
"uid": "aime__2c85af57__s000__tb131072__v001",
"axis": "math",
"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\": \"aime__2c85af57__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 208.90139457337318, \"nll_base\": 5.3418623443342685, \"eval_ref\": \"aime__2c85af57__s000__tb65536__v000\", \"oracle_delta\": 0.18540315546541208, \"random_del... | {
"index": "aime__2c85af57__s000__tb65536__v000",
"uid": "aime__2c85af57__s000__tb65536__v000",
"axis": "math",
"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\": \"anagrams1__e88f0ada__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 174629.96650636027, \"nll_base\": 12.07042453713613, \"eval_ref\": \"anagrams1__e88f0ada__s000__tb262144__v000\", \"oracle_delta\": 1.3055917400203345, \... | {
"index": "anagrams1__e88f0ada__s000__tb262144__v000",
"uid": "anagrams1__e88f0ada__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\": \"anli_r1__ddb4dc48__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2178.9711091594745, \"nll_base\": 7.686608076095581, \"eval_ref\": \"anli_r1__ddb4dc48__s000__tb131072__v000\", \"oracle_delta\": 1.061990737915039, \"rand... | {
"index": "anli_r1__ddb4dc48__s000__tb131072__v000",
"uid": "anli_r1__ddb4dc48__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\": \"anli_r3__1c92a45e__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1696.0496501217349, \"nll_base\": 7.436057090759277, \"eval_ref\": \"anli_r3__1c92a45e__s000__tb131072__v000\", \"oracle_delta\": 1.3761682510375977, \"ran... | {
"index": "anli_r3__1c92a45e__s000__tb131072__v000",
"uid": "anli_r3__1c92a45e__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\": \"arab_culture_completion_egypt__b307eeed__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1.68733086185459, \"nll_base\": 0.5231479087290897, \"eval_ref\": \"arab_culture_completion_egypt__b307eeed__s000__tb131072__v000\", ... | {
"index": "arab_culture_completion_egypt__b307eeed__s000__tb131072__v000",
"uid": "arab_culture_completion_egypt__b307eeed__s000__tb131072__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"arab_culture_completion_yemen__43d655a3__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1.671250540465854, \"nll_base\": 0.513572172824153, \"eval_ref\": \"arab_culture_completion_yemen__43d655a3__s000__tb131072__v000\", ... | {
"index": "arab_culture_completion_yemen__43d655a3__s000__tb131072__v000",
"uid": "arab_culture_completion_yemen__43d655a3__s000__tb131072__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"arabic_leaderboard_acva_Algeria_light__3903734c__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 11.337614633874503, \"nll_base\": 2.4281259264264787, \"eval_ref\": \"arabic_leaderboard_acva_Algeria_light__3903734c__s000__tb... | {
"index": "arabic_leaderboard_acva_Algeria_light__3903734c__s000__tb65536__v000",
"uid": "arabic_leaderboard_acva_Algeria_light__3903734c__s000__tb65536__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"arabic_leaderboard_acva_Arabic_Calligraphy_light__dcf6f6c1__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 10.455486638731404, \"nll_base\": 2.347126877826193, \"eval_ref\": \"arabic_leaderboard_acva_Arabic_Calligraphy_li... | {
"index": "arabic_leaderboard_acva_Arabic_Calligraphy_light__dcf6f6c1__s000__tb131072__v000",
"uid": "arabic_leaderboard_acva_Arabic_Calligraphy_light__dcf6f6c1__s000__tb131072__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"arabic_leaderboard_acva_Arabic_Clothing_light__b7f1a143__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 8.644289653869283, \"nll_base\": 2.1568989473230697, \"eval_ref\": \"arabic_leaderboard_acva_Arabic_Clothing_light__b... | {
"index": "arabic_leaderboard_acva_Arabic_Clothing_light__b7f1a143__s000__tb131072__v000",
"uid": "arabic_leaderboard_acva_Arabic_Clothing_light__b7f1a143__s000__tb131072__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"arabic_leaderboard_acva_Arabic_Philosophy_light__d86c4e0e__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 10.308087117804282, \"nll_base\": 2.3329287442294033, \"eval_ref\": \"arabic_leaderboard_acva_Arabic_Philosophy_lig... | {
"index": "arabic_leaderboard_acva_Arabic_Philosophy_light__d86c4e0e__s000__tb262144__v000",
"uid": "arabic_leaderboard_acva_Arabic_Philosophy_light__d86c4e0e__s000__tb262144__v000",
"axis": "multilingual",
"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\": \"arabic_leaderboard_acva_Islam_Education_light__d7963e70__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 10.813976966500215, \"nll_base\": 2.380839460987156, \"eval_ref\": \"arabic_leaderboard_acva_Islam_Education_light__d... | {
"index": "arabic_leaderboard_acva_Islam_Education_light__d7963e70__s000__tb131072__v000",
"uid": "arabic_leaderboard_acva_Islam_Education_light__d7963e70__s000__tb131072__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"arabic_leaderboard_acva_Islamic_law_system_light__aabb0ac2__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 9.576598016982777, \"nll_base\": 2.2593224158653844, \"eval_ref\": \"arabic_leaderboard_acva_Islamic_law_system_li... | {
"index": "arabic_leaderboard_acva_Islamic_law_system_light__aabb0ac2__s000__tb262144__v000",
"uid": "arabic_leaderboard_acva_Islamic_law_system_light__aabb0ac2__s000__tb262144__v000",
"axis": "multilingual",
"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\": \"arabic_leaderboard_acva_Sudan__791dd114__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 11.014929919994954, \"nll_base\": 2.399251618000291, \"eval_ref\": \"arabic_leaderboard_acva_Sudan__791dd114__s000__tb65536__v000\", \"... | {
"index": "arabic_leaderboard_acva_Sudan__791dd114__s000__tb65536__v000",
"uid": "arabic_leaderboard_acva_Sudan__791dd114__s000__tb65536__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"arabic_leaderboard_arabic_mmlu_clinical_knowledge__cbfb2b81__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 478.6971849045068, \"nll_base\": 6.171068215671974, \"eval_ref\": \"arabic_leaderboard_arabic_mmlu_clinical_knowl... | {
"index": "arabic_leaderboard_arabic_mmlu_clinical_knowledge__cbfb2b81__s000__tb131072__v001",
"uid": "arabic_leaderboard_arabic_mmlu_clinical_knowledge__cbfb2b81__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\": \"arabic_leaderboard_arabic_mmlu_clinical_knowledge__cbfb2b81__s000__tb262144__v002\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 478.6971849045068, \"nll_base\": 6.171068215671974, \"eval_ref\": \"arabic_leaderboard_arabic_mmlu_clinical_knowl... | {
"index": "arabic_leaderboard_arabic_mmlu_clinical_knowledge__cbfb2b81__s000__tb262144__v002",
"uid": "arabic_leaderboard_arabic_mmlu_clinical_knowledge__cbfb2b81__s000__tb262144__v002",
"axis": "domain",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"arabic_leaderboard_arabic_mmlu_clinical_knowledge__cbfb2b81__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 478.6971849045068, \"nll_base\": 6.171068215671974, \"eval_ref\": \"arabic_leaderboard_arabic_mmlu_clinical_knowled... | {
"index": "arabic_leaderboard_arabic_mmlu_clinical_knowledge__cbfb2b81__s000__tb65536__v000",
"uid": "arabic_leaderboard_arabic_mmlu_clinical_knowledge__cbfb2b81__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\": \"arabic_leaderboard_arabic_mmlu_clinical_knowledge__cbfb2b81__s001__tb65536__v003\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 692.5737694230867, \"nll_base\": 6.540414758630701, \"eval_ref\": \"arabic_leaderboard_arabic_mmlu_clinical_knowled... | {
"index": "arabic_leaderboard_arabic_mmlu_clinical_knowledge__cbfb2b81__s001__tb65536__v003",
"uid": "arabic_leaderboard_arabic_mmlu_clinical_knowledge__cbfb2b81__s001__tb65536__v003",
"axis": "domain",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"arabic_leaderboard_arabic_mmlu_clinical_knowledge_light__8749350e__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 597.6910292480496, \"nll_base\": 6.393073946930642, \"eval_ref\": \"arabic_leaderboard_arabic_mmlu_clinical... | {
"index": "arabic_leaderboard_arabic_mmlu_clinical_knowledge_light__8749350e__s000__tb131072__v000",
"uid": "arabic_leaderboard_arabic_mmlu_clinical_knowledge_light__8749350e__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\": \"arabic_leaderboard_arabic_mmlu_clinical_knowledge_light__8749350e__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 597.6910292480496, \"nll_base\": 6.393073946930642, \"eval_ref\": \"arabic_leaderboard_arabic_mmlu_clinical... | {
"index": "arabic_leaderboard_arabic_mmlu_clinical_knowledge_light__8749350e__s000__tb262144__v001",
"uid": "arabic_leaderboard_arabic_mmlu_clinical_knowledge_light__8749350e__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\": \"arabic_leaderboard_arabic_mmlu_clinical_knowledge_light__8749350e__s000__tb65536__v002\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 597.6910292480496, \"nll_base\": 6.393073946930642, \"eval_ref\": \"arabic_leaderboard_arabic_mmlu_clinical_k... | {
"index": "arabic_leaderboard_arabic_mmlu_clinical_knowledge_light__8749350e__s000__tb65536__v002",
"uid": "arabic_leaderboard_arabic_mmlu_clinical_knowledge_light__8749350e__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\": \"arabic_leaderboard_arabic_mmlu_clinical_knowledge_light__8749350e__s001__tb131072__v003\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 663.8664671901797, \"nll_base\": 6.49808102562314, \"eval_ref\": \"arabic_leaderboard_arabic_mmlu_clinical_... | {
"index": "arabic_leaderboard_arabic_mmlu_clinical_knowledge_light__8749350e__s001__tb131072__v003",
"uid": "arabic_leaderboard_arabic_mmlu_clinical_knowledge_light__8749350e__s001__tb131072__v003",
"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\": \"arabic_leaderboard_arabic_mmlu_college_chemistry__cbbcba80__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 498.56428719453123, \"nll_base\": 6.211732542359984, \"eval_ref\": \"arabic_leaderboard_arabic_mmlu_college_chemistr... | {
"index": "arabic_leaderboard_arabic_mmlu_college_chemistry__cbbcba80__s000__tb65536__v000",
"uid": "arabic_leaderboard_arabic_mmlu_college_chemistry__cbbcba80__s000__tb65536__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"arabic_leaderboard_arabic_mmlu_college_physics__6687641c__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 384.31427701488604, \"nll_base\": 5.951460647583008, \"eval_ref\": \"arabic_leaderboard_arabic_mmlu_college_physics__6... | {
"index": "arabic_leaderboard_arabic_mmlu_college_physics__6687641c__s000__tb65536__v000",
"uid": "arabic_leaderboard_arabic_mmlu_college_physics__6687641c__s000__tb65536__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"arabic_leaderboard_arabic_mmlu_computer_security__821e930d__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 341.7850438316582, \"nll_base\": 5.834182012648809, \"eval_ref\": \"arabic_leaderboard_arabic_mmlu_computer_security... | {
"index": "arabic_leaderboard_arabic_mmlu_computer_security__821e930d__s000__tb65536__v000",
"uid": "arabic_leaderboard_arabic_mmlu_computer_security__821e930d__s000__tb65536__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"arabic_leaderboard_arabic_mmlu_elementary_mathematics_light__3a0bb5d6__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 501.6219371286974, \"nll_base\": 6.2178467226700045, \"eval_ref\": \"arabic_leaderboard_arabic_mmlu_ele... | {
"index": "arabic_leaderboard_arabic_mmlu_elementary_mathematics_light__3a0bb5d6__s000__tb262144__v000",
"uid": "arabic_leaderboard_arabic_mmlu_elementary_mathematics_light__3a0bb5d6__s000__tb262144__v000",
"axis": "multilingual",
"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\": \"arabic_leaderboard_arabic_mmlu_formal_logic_light__ea81b74b__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 183.55531956602843, \"nll_base\": 5.2125160910866475, \"eval_ref\": \"arabic_leaderboard_arabic_mmlu_formal_logic... | {
"index": "arabic_leaderboard_arabic_mmlu_formal_logic_light__ea81b74b__s000__tb131072__v000",
"uid": "arabic_leaderboard_arabic_mmlu_formal_logic_light__ea81b74b__s000__tb131072__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"arabic_leaderboard_arabic_mmlu_high_school_biology__13aa9e3a__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 372.3524848177067, \"nll_base\": 5.9198409454732, \"eval_ref\": \"arabic_leaderboard_arabic_mmlu_high_school_bio... | {
"index": "arabic_leaderboard_arabic_mmlu_high_school_biology__13aa9e3a__s000__tb262144__v000",
"uid": "arabic_leaderboard_arabic_mmlu_high_school_biology__13aa9e3a__s000__tb262144__v000",
"axis": "multilingual",
"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\": \"arabic_leaderboard_arabic_mmlu_high_school_european_history_light__9be1e01f__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 629.5103365424039, \"nll_base\": 6.444942273591694, \"eval_ref\": \"arabic_leaderboard_arabic_mml... | {
"index": "arabic_leaderboard_arabic_mmlu_high_school_european_history_light__9be1e01f__s000__tb262144__v000",
"uid": "arabic_leaderboard_arabic_mmlu_high_school_european_history_light__9be1e01f__s000__tb262144__v000",
"axis": "multilingual",
"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\": \"arabic_leaderboard_arabic_mmlu_high_school_government_and_politics__4d257716__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 392.01698625925684, \"nll_base\": 5.971305171145668, \"eval_ref\": \"arabic_leaderboard_arabic_m... | {
"index": "arabic_leaderboard_arabic_mmlu_high_school_government_and_politics__4d257716__s000__tb262144__v000",
"uid": "arabic_leaderboard_arabic_mmlu_high_school_government_and_politics__4d257716__s000__tb262144__v000",
"axis": "multilingual",
"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\": \"arabic_leaderboard_arabic_mmlu_high_school_mathematics_light__ba593dca__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 721.3844764069381, \"nll_base\": 6.581172249533913, \"eval_ref\": \"arabic_leaderboard_arabic_mmlu_hig... | {
"index": "arabic_leaderboard_arabic_mmlu_high_school_mathematics_light__ba593dca__s000__tb131072__v000",
"uid": "arabic_leaderboard_arabic_mmlu_high_school_mathematics_light__ba593dca__s000__tb131072__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"arabic_leaderboard_arabic_mmlu_high_school_physics__193452be__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 380.5436162691257, \"nll_base\": 5.941600799560547, \"eval_ref\": \"arabic_leaderboard_arabic_mmlu_high_school_phy... | {
"index": "arabic_leaderboard_arabic_mmlu_high_school_physics__193452be__s000__tb65536__v000",
"uid": "arabic_leaderboard_arabic_mmlu_high_school_physics__193452be__s000__tb65536__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"arabic_leaderboard_arabic_mmlu_high_school_psychology_light__0d6f15cc__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 577.2531491091713, \"nll_base\": 6.358280903584248, \"eval_ref\": \"arabic_leaderboard_arabic_mmlu_high_s... | {
"index": "arabic_leaderboard_arabic_mmlu_high_school_psychology_light__0d6f15cc__s000__tb65536__v000",
"uid": "arabic_leaderboard_arabic_mmlu_high_school_psychology_light__0d6f15cc__s000__tb65536__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"arabic_leaderboard_arabic_mmlu_jurisprudence__d3b90595__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 511.78823242090783, \"nll_base\": 6.237910930926983, \"eval_ref\": \"arabic_leaderboard_arabic_mmlu_jurisprudence__d3b... | {
"index": "arabic_leaderboard_arabic_mmlu_jurisprudence__d3b90595__s000__tb131072__v000",
"uid": "arabic_leaderboard_arabic_mmlu_jurisprudence__d3b90595__s000__tb131072__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"arabic_leaderboard_arabic_mmlu_medical_genetics__b9c89a3d__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 460.2047814591954, \"nll_base\": 6.131671567506428, \"eval_ref\": \"arabic_leaderboard_arabic_mmlu_medical_genetics... | {
"index": "arabic_leaderboard_arabic_mmlu_medical_genetics__b9c89a3d__s000__tb131072__v001",
"uid": "arabic_leaderboard_arabic_mmlu_medical_genetics__b9c89a3d__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\": \"arabic_leaderboard_arabic_mmlu_medical_genetics__b9c89a3d__s000__tb262144__v002\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 460.2047814591954, \"nll_base\": 6.131671567506428, \"eval_ref\": \"arabic_leaderboard_arabic_mmlu_medical_genetics... | {
"index": "arabic_leaderboard_arabic_mmlu_medical_genetics__b9c89a3d__s000__tb262144__v002",
"uid": "arabic_leaderboard_arabic_mmlu_medical_genetics__b9c89a3d__s000__tb262144__v002",
"axis": "domain",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"arabic_leaderboard_arabic_mmlu_public_relations__c5ac8154__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 476.91958684281275, \"nll_base\": 6.167347895634639, \"eval_ref\": \"arabic_leaderboard_arabic_mmlu_public_relation... | {
"index": "arabic_leaderboard_arabic_mmlu_public_relations__c5ac8154__s000__tb131072__v000",
"uid": "arabic_leaderboard_arabic_mmlu_public_relations__c5ac8154__s000__tb131072__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"arabic_leaderboard_arabic_mmlu_world_religions__b40f24f5__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 300.5736720269645, \"nll_base\": 5.70569288874247, \"eval_ref\": \"arabic_leaderboard_arabic_mmlu_world_religions__b... | {
"index": "arabic_leaderboard_arabic_mmlu_world_religions__b40f24f5__s000__tb131072__v000",
"uid": "arabic_leaderboard_arabic_mmlu_world_religions__b40f24f5__s000__tb131072__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"arabic_mt_toxigen_light__aa8490e2__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 80.52425527734268, \"nll_base\": 4.3885584468335175, \"eval_ref\": \"arabic_mt_toxigen_light__aa8490e2__s000__tb65536__v000\", \"oracle_delta... | {
"index": "arabic_mt_toxigen_light__aa8490e2__s000__tb65536__v000",
"uid": "arabic_mt_toxigen_light__aa8490e2__s000__tb65536__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"arabicmmlu_general_knowledge_primary_school__ab5c3e53__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 988.8954012727764, \"nll_base\": 6.896588563919067, \"eval_ref\": \"arabicmmlu_general_knowledge_primary_school__ab5c3e... | {
"index": "arabicmmlu_general_knowledge_primary_school__ab5c3e53__s000__tb131072__v000",
"uid": "arabicmmlu_general_knowledge_primary_school__ab5c3e53__s000__tb131072__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"arabicmmlu_geography_high_school__136b1b8d__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 757.7604290720243, \"nll_base\": 6.630367279052734, \"eval_ref\": \"arabicmmlu_geography_high_school__136b1b8d__s000__tb262144__v0... | {
"index": "arabicmmlu_geography_high_school__136b1b8d__s000__tb262144__v000",
"uid": "arabicmmlu_geography_high_school__136b1b8d__s000__tb262144__v000",
"axis": "multilingual",
"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\": \"arabicmmlu_social_science_primary_school__b94fe179__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1042.3211691344081, \"nll_base\": 6.94920539855957, \"eval_ref\": \"arabicmmlu_social_science_primary_school__b94fe179__s0... | {
"index": "arabicmmlu_social_science_primary_school__b94fe179__s000__tb131072__v000",
"uid": "arabicmmlu_social_science_primary_school__b94fe179__s000__tb131072__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"arc_challenge_llama__de8cd699__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 5338.878143692579, \"nll_base\": 8.582770824432373, \"eval_ref\": \"arc_challenge_llama__de8cd699__s000__tb65536__v000\", \"oracle_delta\": 0.607... | {
"index": "arc_challenge_llama__de8cd699__s000__tb65536__v000",
"uid": "arc_challenge_llama__de8cd699__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\": \"arc_challenge_mt_de__1e89f8a6__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3.6251345636842442, \"nll_base\": 1.2878914086340134, \"eval_ref\": \"arc_challenge_mt_de__1e89f8a6__s000__tb262144__v000\", \"oracle_delta\": ... | {
"index": "arc_challenge_mt_de__1e89f8a6__s000__tb262144__v000",
"uid": "arc_challenge_mt_de__1e89f8a6__s000__tb262144__v000",
"axis": "multilingual",
"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\": \"arc_challenge_mt_nb__439340c1__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 6.086812753212236, \"nll_base\": 1.8061245872860863, \"eval_ref\": \"arc_challenge_mt_nb__439340c1__s000__tb131072__v000\", \"oracle_delta\": 0... | {
"index": "arc_challenge_mt_nb__439340c1__s000__tb131072__v000",
"uid": "arc_challenge_mt_nb__439340c1__s000__tb131072__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"arc_challenge_mt_pt__eb5a56dd__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 7.067737355137765, \"nll_base\": 1.9555403940524896, \"eval_ref\": \"arc_challenge_mt_pt__eb5a56dd__s000__tb65536__v000\", \"oracle_delta\": 0.08... | {
"index": "arc_challenge_mt_pt__eb5a56dd__s000__tb65536__v000",
"uid": "arc_challenge_mt_pt__eb5a56dd__s000__tb65536__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"arc_challenge_mt_sv__854ba45d__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 5.856921079428544, \"nll_base\": 1.767624052444307, \"eval_ref\": \"arc_challenge_mt_sv__854ba45d__s000__tb262144__v000\", \"oracle_delta\": 0.... | {
"index": "arc_challenge_mt_sv__854ba45d__s000__tb262144__v000",
"uid": "arc_challenge_mt_sv__854ba45d__s000__tb262144__v000",
"axis": "multilingual",
"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\": \"arc_es__17c0d1c4__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4.429451010616176, \"nll_base\": 1.4882756510088522, \"eval_ref\": \"arc_es__17c0d1c4__s000__tb65536__v000\", \"oracle_delta\": 0.032550565658077124, \"random... | {
"index": "arc_es__17c0d1c4__s000__tb65536__v000",
"uid": "arc_es__17c0d1c4__s000__tb65536__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"arc_eu_easy__b333ca25__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 7.1304029609875075, \"nll_base\": 1.964367749096753, \"eval_ref\": \"arc_eu_easy__b333ca25__s000__tb65536__v000\", \"oracle_delta\": 0.049029192766985696... | {
"index": "arc_eu_easy__b333ca25__s000__tb65536__v000",
"uid": "arc_eu_easy__b333ca25__s000__tb65536__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"arc_zh__93ef7e45__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3.787173739105373, \"nll_base\": 1.3316200256347657, \"eval_ref\": \"arc_zh__93ef7e45__s000__tb131072__v000\", \"oracle_delta\": 0.023991775512695268, \"ran... | {
"index": "arc_zh__93ef7e45__s000__tb131072__v000",
"uid": "arc_zh__93ef7e45__s000__tb131072__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"arithmetic_1dc__b69b9c98__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 907.294468771368, \"nll_base\": 6.810467059795673, \"eval_ref\": \"arithmetic_1dc__b69b9c98__s000__tb131072__v000\", \"oracle_delta\": 1.08307823768... | {
"index": "arithmetic_1dc__b69b9c98__s000__tb131072__v000",
"uid": "arithmetic_1dc__b69b9c98__s000__tb131072__v000",
"axis": "math",
"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\": \"arithmetic_1dc__b69b9c98__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 907.294468771368, \"nll_base\": 6.810467059795673, \"eval_ref\": \"arithmetic_1dc__b69b9c98__s000__tb262144__v001\", \"oracle_delta\": 1.63456217447... | {
"index": "arithmetic_1dc__b69b9c98__s000__tb262144__v001",
"uid": "arithmetic_1dc__b69b9c98__s000__tb262144__v001",
"axis": "math",
"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\": \"arithmetic_2ds__9383e51c__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 739.2216197547206, \"nll_base\": 6.605597767403455, \"eval_ref\": \"arithmetic_2ds__9383e51c__s000__tb65536__v001\", \"oracle_delta\": 0.2138788486883... | {
"index": "arithmetic_2ds__9383e51c__s000__tb65536__v001",
"uid": "arithmetic_2ds__9383e51c__s000__tb65536__v001",
"axis": "math",
"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\": \"arithmetic_3da__c8aae1dd__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 194.674304822522, \"nll_base\": 5.271327930492359, \"eval_ref\": \"arithmetic_3da__c8aae1dd__s000__tb65536__v000\", \"oracle_delta\": 0.24346135736821... | {
"index": "arithmetic_3da__c8aae1dd__s000__tb65536__v000",
"uid": "arithmetic_3da__c8aae1dd__s000__tb65536__v000",
"axis": "math",
"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\": \"arithmetic_4da__237460ed__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 150.37722532006202, \"nll_base\": 5.01314697265625, \"eval_ref\": \"arithmetic_4da__237460ed__s000__tb131072__v000\", \"oracle_delta\": 0.4929133717... | {
"index": "arithmetic_4da__237460ed__s000__tb131072__v000",
"uid": "arithmetic_4da__237460ed__s000__tb131072__v000",
"axis": "math",
"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\": \"arithmetic_4da__237460ed__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 150.37722532006202, \"nll_base\": 5.01314697265625, \"eval_ref\": \"arithmetic_4da__237460ed__s000__tb262144__v001\", \"oracle_delta\": 0.7590443680... | {
"index": "arithmetic_4da__237460ed__s000__tb262144__v001",
"uid": "arithmetic_4da__237460ed__s000__tb262144__v001",
"axis": "math",
"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\": \"arithmetic_4ds__f83b67cc__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 189.0647164739215, \"nll_base\": 5.242089371654869, \"eval_ref\": \"arithmetic_4ds__f83b67cc__s000__tb131072__v001\", \"oracle_delta\": 0.2864373138... | {
"index": "arithmetic_4ds__f83b67cc__s000__tb131072__v001",
"uid": "arithmetic_4ds__f83b67cc__s000__tb131072__v001",
"axis": "math",
"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\": \"arithmetic_4ds__f83b67cc__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 189.0647164739215, \"nll_base\": 5.242089371654869, \"eval_ref\": \"arithmetic_4ds__f83b67cc__s000__tb65536__v000\", \"oracle_delta\": 0.0843429986943... | {
"index": "arithmetic_4ds__f83b67cc__s000__tb65536__v000",
"uid": "arithmetic_4ds__f83b67cc__s000__tb65536__v000",
"axis": "math",
"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\": \"arithmetic_5da__461932bc__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 94.81630844580691, \"nll_base\": 4.5519414244845695, \"eval_ref\": \"arithmetic_5da__461932bc__s000__tb131072__v000\", \"oracle_delta\": 0.390416062... | {
"index": "arithmetic_5da__461932bc__s000__tb131072__v000",
"uid": "arithmetic_5da__461932bc__s000__tb131072__v000",
"axis": "math",
"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\": \"arithmetic_5da__461932bc__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 94.81630844580691, \"nll_base\": 4.5519414244845695, \"eval_ref\": \"arithmetic_5da__461932bc__s000__tb262144__v001\", \"oracle_delta\": 0.602004181... | {
"index": "arithmetic_5da__461932bc__s000__tb262144__v001",
"uid": "arithmetic_5da__461932bc__s000__tb262144__v001",
"axis": "math",
"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\": \"arithmetic_5ds__40308e30__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 135.8582427552631, \"nll_base\": 4.9116120094476745, \"eval_ref\": \"arithmetic_5ds__40308e30__s000__tb65536__v001\", \"oracle_delta\": 0.041156999454... | {
"index": "arithmetic_5ds__40308e30__s000__tb65536__v001",
"uid": "arithmetic_5ds__40308e30__s000__tb65536__v001",
"axis": "math",
"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\": \"asdiv__f74d0873__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 239.78701440260454, \"nll_base\": 5.4797510893448536, \"eval_ref\": \"asdiv__f74d0873__s000__tb131072__v001\", \"oracle_delta\": 0.5562555064325743, \"random... | {
"index": "asdiv__f74d0873__s000__tb131072__v001",
"uid": "asdiv__f74d0873__s000__tb131072__v001",
"axis": "math",
"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\": \"asdiv__f74d0873__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 239.78701440260454, \"nll_base\": 5.4797510893448536, \"eval_ref\": \"asdiv__f74d0873__s000__tb65536__v000\", \"oracle_delta\": 0.28764393018639556, \"random_d... | {
"index": "asdiv__f74d0873__s000__tb65536__v000",
"uid": "asdiv__f74d0873__s000__tb65536__v000",
"axis": "math",
"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\": \"asdiv_cot_llama__5f1f456b__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 303.7774021374713, \"nll_base\": 5.716295203384088, \"eval_ref\": \"asdiv_cot_llama__5f1f456b__s000__tb131072__v000\", \"oracle_delta\": 0.65675945... | {
"index": "asdiv_cot_llama__5f1f456b__s000__tb131072__v000",
"uid": "asdiv_cot_llama__5f1f456b__s000__tb131072__v000",
"axis": "math",
"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\": \"asdiv_cot_llama__5f1f456b__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 303.7774021374713, \"nll_base\": 5.716295203384088, \"eval_ref\": \"asdiv_cot_llama__5f1f456b__s000__tb262144__v001\", \"oracle_delta\": 0.61471183... | {
"index": "asdiv_cot_llama__5f1f456b__s000__tb262144__v001",
"uid": "asdiv_cot_llama__5f1f456b__s000__tb262144__v001",
"axis": "math",
"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\": \"babilong_qa10__39062171__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 5820.629543829227, \"nll_base\": 8.669163703918457, \"eval_ref\": \"babilong_qa10__39062171__s000__tb131072__v001\", \"oracle_delta\": 1.254761934280... | {
"index": "babilong_qa10__39062171__s000__tb131072__v001",
"uid": "babilong_qa10__39062171__s000__tb131072__v001",
"axis": "reasoning",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"babilong_qa10__39062171__s000__tb262144__v002\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 5820.629543829227, \"nll_base\": 8.669163703918457, \"eval_ref\": \"babilong_qa10__39062171__s000__tb262144__v002\", \"oracle_delta\": 1.088204145431... | {
"index": "babilong_qa10__39062171__s000__tb262144__v002",
"uid": "babilong_qa10__39062171__s000__tb262144__v002",
"axis": "reasoning",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"babilong_qa10__39062171__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 5820.629543829227, \"nll_base\": 8.669163703918457, \"eval_ref\": \"babilong_qa10__39062171__s000__tb65536__v000\", \"oracle_delta\": 0.771395206451416... | {
"index": "babilong_qa10__39062171__s000__tb65536__v000",
"uid": "babilong_qa10__39062171__s000__tb65536__v000",
"axis": "reasoning",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"babilong_qa11__53723708__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1836.3032224272604, \"nll_base\": 7.515509711371528, \"eval_ref\": \"babilong_qa11__53723708__s000__tb131072__v000\", \"oracle_delta\": 0.58821445041... | {
"index": "babilong_qa11__53723708__s000__tb131072__v000",
"uid": "babilong_qa11__53723708__s000__tb131072__v000",
"axis": "reasoning",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"babilong_qa12__f2d0ab19__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1993.2781902867612, \"nll_base\": 7.597535894158181, \"eval_ref\": \"babilong_qa12__f2d0ab19__s000__tb131072__v000\", \"oracle_delta\": 0.45523997103... | {
"index": "babilong_qa12__f2d0ab19__s000__tb131072__v000",
"uid": "babilong_qa12__f2d0ab19__s000__tb131072__v000",
"axis": "reasoning",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"babilong_qa12__f2d0ab19__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1993.2781902867612, \"nll_base\": 7.597535894158181, \"eval_ref\": \"babilong_qa12__f2d0ab19__s000__tb262144__v001\", \"oracle_delta\": 0.49117699098... | {
"index": "babilong_qa12__f2d0ab19__s000__tb262144__v001",
"uid": "babilong_qa12__f2d0ab19__s000__tb262144__v001",
"axis": "reasoning",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"babilong_qa12__f2d0ab19__s000__tb65536__v002\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1993.2781902867612, \"nll_base\": 7.597535894158181, \"eval_ref\": \"babilong_qa12__f2d0ab19__s000__tb65536__v002\", \"oracle_delta\": 0.29769348830319... | {
"index": "babilong_qa12__f2d0ab19__s000__tb65536__v002",
"uid": "babilong_qa12__f2d0ab19__s000__tb65536__v002",
"axis": "reasoning",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"babilong_qa13__b19d4e7f__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2680.951377066643, \"nll_base\": 7.893927001953125, \"eval_ref\": \"babilong_qa13__b19d4e7f__s000__tb131072__v000\", \"oracle_delta\": 0.681724378797... | {
"index": "babilong_qa13__b19d4e7f__s000__tb131072__v000",
"uid": "babilong_qa13__b19d4e7f__s000__tb131072__v000",
"axis": "reasoning",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"babilong_qa13__b19d4e7f__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2680.951377066643, \"nll_base\": 7.893927001953125, \"eval_ref\": \"babilong_qa13__b19d4e7f__s000__tb262144__v001\", \"oracle_delta\": 0.680114746093... | {
"index": "babilong_qa13__b19d4e7f__s000__tb262144__v001",
"uid": "babilong_qa13__b19d4e7f__s000__tb262144__v001",
"axis": "reasoning",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"babilong_qa14__797a93f9__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 7166.238865420177, \"nll_base\": 8.87713623046875, \"eval_ref\": \"babilong_qa14__797a93f9__s000__tb131072__v000\", \"oracle_delta\": 0.4912647247314... | {
"index": "babilong_qa14__797a93f9__s000__tb131072__v000",
"uid": "babilong_qa14__797a93f9__s000__tb131072__v000",
"axis": "reasoning",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"babilong_qa14__797a93f9__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 7166.238865420177, \"nll_base\": 8.87713623046875, \"eval_ref\": \"babilong_qa14__797a93f9__s000__tb262144__v001\", \"oracle_delta\": 0.6725120544433... | {
"index": "babilong_qa14__797a93f9__s000__tb262144__v001",
"uid": "babilong_qa14__797a93f9__s000__tb262144__v001",
"axis": "reasoning",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"babilong_qa14__797a93f9__s000__tb65536__v002\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 7166.238865420177, \"nll_base\": 8.87713623046875, \"eval_ref\": \"babilong_qa14__797a93f9__s000__tb65536__v002\", \"oracle_delta\": 0.2512634277343757... | {
"index": "babilong_qa14__797a93f9__s000__tb65536__v002",
"uid": "babilong_qa14__797a93f9__s000__tb65536__v002",
"axis": "reasoning",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"babilong_qa15__0b5ee37d__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4526.280671277963, \"nll_base\": 8.417655837367958, \"eval_ref\": \"babilong_qa15__0b5ee37d__s000__tb131072__v001\", \"oracle_delta\": 1.549432029186... | {
"index": "babilong_qa15__0b5ee37d__s000__tb131072__v001",
"uid": "babilong_qa15__0b5ee37d__s000__tb131072__v001",
"axis": "reasoning",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"babilong_qa15__0b5ee37d__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4526.280671277963, \"nll_base\": 8.417655837367958, \"eval_ref\": \"babilong_qa15__0b5ee37d__s000__tb65536__v000\", \"oracle_delta\": 0.90715907996809,... | {
"index": "babilong_qa15__0b5ee37d__s000__tb65536__v000",
"uid": "babilong_qa15__0b5ee37d__s000__tb65536__v000",
"axis": "reasoning",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"babilong_qa16__680a486e__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 12264.17014522866, \"nll_base\": 9.414437294006348, \"eval_ref\": \"babilong_qa16__680a486e__s000__tb131072__v000\", \"oracle_delta\": 1.175781726837... | {
"index": "babilong_qa16__680a486e__s000__tb131072__v000",
"uid": "babilong_qa16__680a486e__s000__tb131072__v000",
"axis": "reasoning",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"babilong_qa16__680a486e__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 12264.17014522866, \"nll_base\": 9.414437294006348, \"eval_ref\": \"babilong_qa16__680a486e__s000__tb262144__v001\", \"oracle_delta\": 1.293859481811... | {
"index": "babilong_qa16__680a486e__s000__tb262144__v001",
"uid": "babilong_qa16__680a486e__s000__tb262144__v001",
"axis": "reasoning",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"babilong_qa16__680a486e__s000__tb65536__v002\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 12264.17014522866, \"nll_base\": 9.414437294006348, \"eval_ref\": \"babilong_qa16__680a486e__s000__tb65536__v002\", \"oracle_delta\": 0.647219181060791... | {
"index": "babilong_qa16__680a486e__s000__tb65536__v002",
"uid": "babilong_qa16__680a486e__s000__tb65536__v002",
"axis": "reasoning",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"babilong_qa17__5d186e46__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1959.7288473389335, \"nll_base\": 7.580561399459839, \"eval_ref\": \"babilong_qa17__5d186e46__s000__tb131072__v000\", \"oracle_delta\": 0.46572828292... | {
"index": "babilong_qa17__5d186e46__s000__tb131072__v000",
"uid": "babilong_qa17__5d186e46__s000__tb131072__v000",
"axis": "reasoning",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"babilong_qa17__5d186e46__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1959.7288473389335, \"nll_base\": 7.580561399459839, \"eval_ref\": \"babilong_qa17__5d186e46__s000__tb262144__v001\", \"oracle_delta\": 0.78467845916... | {
"index": "babilong_qa17__5d186e46__s000__tb262144__v001",
"uid": "babilong_qa17__5d186e46__s000__tb262144__v001",
"axis": "reasoning",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"babilong_qa18__ff842f79__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1864.0918329487351, \"nll_base\": 7.530529260635376, \"eval_ref\": \"babilong_qa18__ff842f79__s000__tb131072__v001\", \"oracle_delta\": 1.33350419998... | {
"index": "babilong_qa18__ff842f79__s000__tb131072__v001",
"uid": "babilong_qa18__ff842f79__s000__tb131072__v001",
"axis": "reasoning",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"babilong_qa18__ff842f79__s000__tb262144__v002\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1864.0918329487351, \"nll_base\": 7.530529260635376, \"eval_ref\": \"babilong_qa18__ff842f79__s000__tb262144__v002\", \"oracle_delta\": 1.52063202857... | {
"index": "babilong_qa18__ff842f79__s000__tb262144__v002",
"uid": "babilong_qa18__ff842f79__s000__tb262144__v002",
"axis": "reasoning",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"babilong_qa18__ff842f79__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1864.0918329487351, \"nll_base\": 7.530529260635376, \"eval_ref\": \"babilong_qa18__ff842f79__s000__tb65536__v000\", \"oracle_delta\": 0.81562232971191... | {
"index": "babilong_qa18__ff842f79__s000__tb65536__v000",
"uid": "babilong_qa18__ff842f79__s000__tb65536__v000",
"axis": "reasoning",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"babilong_qa19__4fe9c6c8__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1433.5057464366146, \"nll_base\": 7.267878293991089, \"eval_ref\": \"babilong_qa19__4fe9c6c8__s000__tb131072__v001\", \"oracle_delta\": 0.21663308143... | {
"index": "babilong_qa19__4fe9c6c8__s000__tb131072__v001",
"uid": "babilong_qa19__4fe9c6c8__s000__tb131072__v001",
"axis": "reasoning",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"babilong_qa19__4fe9c6c8__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1433.5057464366146, \"nll_base\": 7.267878293991089, \"eval_ref\": \"babilong_qa19__4fe9c6c8__s000__tb65536__v000\", \"oracle_delta\": 0.05436635017395... | {
"index": "babilong_qa19__4fe9c6c8__s000__tb65536__v000",
"uid": "babilong_qa19__4fe9c6c8__s000__tb65536__v000",
"axis": "reasoning",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"babilong_qa1__7fd5d334__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1511.3957484413688, \"nll_base\": 7.320788839588994, \"eval_ref\": \"babilong_qa1__7fd5d334__s000__tb131072__v001\", \"oracle_delta\": 0.4176081781801... | {
"index": "babilong_qa1__7fd5d334__s000__tb131072__v001",
"uid": "babilong_qa1__7fd5d334__s000__tb131072__v001",
"axis": "reasoning",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.