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_flan_cot_zeroshot_high_school_macroeconomics__12a100a6__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 82.81563869264343, \"nll_base\": 4.416616916656494, \"eval_ref\": \"mmlu_flan_cot_zeroshot_high_school_macroecono... | {
"index": "mmlu_flan_cot_zeroshot_high_school_macroeconomics__12a100a6__s000__tb131072__v000",
"uid": "mmlu_flan_cot_zeroshot_high_school_macroeconomics__12a100a6__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_flan_cot_zeroshot_high_school_mathematics__a72efc4d__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 70.17759852861596, \"nll_base\": 4.25102915082659, \"eval_ref\": \"mmlu_flan_cot_zeroshot_high_school_mathematics__a... | {
"index": "mmlu_flan_cot_zeroshot_high_school_mathematics__a72efc4d__s000__tb131072__v000",
"uid": "mmlu_flan_cot_zeroshot_high_school_mathematics__a72efc4d__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_flan_cot_zeroshot_high_school_mathematics__a72efc4d__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 70.17759852861596, \"nll_base\": 4.25102915082659, \"eval_ref\": \"mmlu_flan_cot_zeroshot_high_school_mathematics__a... | {
"index": "mmlu_flan_cot_zeroshot_high_school_mathematics__a72efc4d__s000__tb262144__v001",
"uid": "mmlu_flan_cot_zeroshot_high_school_mathematics__a72efc4d__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_flan_cot_zeroshot_high_school_microeconomics__168bc93a__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 90.59835263708449, \"nll_base\": 4.506436030069987, \"eval_ref\": \"mmlu_flan_cot_zeroshot_high_school_microecono... | {
"index": "mmlu_flan_cot_zeroshot_high_school_microeconomics__168bc93a__s000__tb262144__v000",
"uid": "mmlu_flan_cot_zeroshot_high_school_microeconomics__168bc93a__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_flan_cot_zeroshot_high_school_physics__cd88b430__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 71.48042176718528, \"nll_base\": 4.269423590766059, \"eval_ref\": \"mmlu_flan_cot_zeroshot_high_school_physics__cd88b430... | {
"index": "mmlu_flan_cot_zeroshot_high_school_physics__cd88b430__s000__tb131072__v000",
"uid": "mmlu_flan_cot_zeroshot_high_school_physics__cd88b430__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_flan_cot_zeroshot_high_school_statistics__ac44a05b__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 75.13733909981643, \"nll_base\": 4.319317626953125, \"eval_ref\": \"mmlu_flan_cot_zeroshot_high_school_statistics__ac44... | {
"index": "mmlu_flan_cot_zeroshot_high_school_statistics__ac44a05b__s000__tb65536__v000",
"uid": "mmlu_flan_cot_zeroshot_high_school_statistics__ac44a05b__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_flan_cot_zeroshot_high_school_us_history__535e7d8c__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 82.30990052168144, \"nll_base\": 4.410491398402622, \"eval_ref\": \"mmlu_flan_cot_zeroshot_high_school_us_history__53... | {
"index": "mmlu_flan_cot_zeroshot_high_school_us_history__535e7d8c__s000__tb262144__v000",
"uid": "mmlu_flan_cot_zeroshot_high_school_us_history__535e7d8c__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_flan_cot_zeroshot_high_school_us_history__535e7d8c__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 82.30990052168144, \"nll_base\": 4.410491398402622, \"eval_ref\": \"mmlu_flan_cot_zeroshot_high_school_us_history__535e... | {
"index": "mmlu_flan_cot_zeroshot_high_school_us_history__535e7d8c__s000__tb65536__v001",
"uid": "mmlu_flan_cot_zeroshot_high_school_us_history__535e7d8c__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_flan_cot_zeroshot_high_school_world_history__5da849f1__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 81.26424571867285, \"nll_base\": 4.397706137763129, \"eval_ref\": \"mmlu_flan_cot_zeroshot_high_school_world_histo... | {
"index": "mmlu_flan_cot_zeroshot_high_school_world_history__5da849f1__s000__tb131072__v001",
"uid": "mmlu_flan_cot_zeroshot_high_school_world_history__5da849f1__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_flan_cot_zeroshot_human_aging__7686c8d6__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 80.90264028710997, \"nll_base\": 4.393246459960937, \"eval_ref\": \"mmlu_flan_cot_zeroshot_human_aging__7686c8d6__s000__tb65536__v... | {
"index": "mmlu_flan_cot_zeroshot_human_aging__7686c8d6__s000__tb65536__v001",
"uid": "mmlu_flan_cot_zeroshot_human_aging__7686c8d6__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_flan_cot_zeroshot_human_sexuality__05e2af00__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 61.789209900235406, \"nll_base\": 4.1237287521362305, \"eval_ref\": \"mmlu_flan_cot_zeroshot_human_sexuality__05e2af00__s000... | {
"index": "mmlu_flan_cot_zeroshot_human_sexuality__05e2af00__s000__tb131072__v000",
"uid": "mmlu_flan_cot_zeroshot_human_sexuality__05e2af00__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_flan_cot_zeroshot_international_law__fa6409fd__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 86.08533315514997, \"nll_base\": 4.455339050292968, \"eval_ref\": \"mmlu_flan_cot_zeroshot_international_law__fa6409fd__s0... | {
"index": "mmlu_flan_cot_zeroshot_international_law__fa6409fd__s000__tb131072__v000",
"uid": "mmlu_flan_cot_zeroshot_international_law__fa6409fd__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_flan_cot_zeroshot_jurisprudence__14c0a91c__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 95.47018967589334, \"nll_base\": 4.55881404876709, \"eval_ref\": \"mmlu_flan_cot_zeroshot_jurisprudence__14c0a91c__s000__tb65536... | {
"index": "mmlu_flan_cot_zeroshot_jurisprudence__14c0a91c__s000__tb65536__v000",
"uid": "mmlu_flan_cot_zeroshot_jurisprudence__14c0a91c__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_flan_cot_zeroshot_logical_fallacies__fb46c696__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 79.859440605719, \"nll_base\": 4.380268096923828, \"eval_ref\": \"mmlu_flan_cot_zeroshot_logical_fallacies__fb46c696__s000... | {
"index": "mmlu_flan_cot_zeroshot_logical_fallacies__fb46c696__s000__tb131072__v001",
"uid": "mmlu_flan_cot_zeroshot_logical_fallacies__fb46c696__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_flan_cot_zeroshot_logical_fallacies__fb46c696__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 79.859440605719, \"nll_base\": 4.380268096923828, \"eval_ref\": \"mmlu_flan_cot_zeroshot_logical_fallacies__fb46c696__s000__... | {
"index": "mmlu_flan_cot_zeroshot_logical_fallacies__fb46c696__s000__tb65536__v000",
"uid": "mmlu_flan_cot_zeroshot_logical_fallacies__fb46c696__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_flan_cot_zeroshot_management__0feb988c__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 90.41436090219516, \"nll_base\": 4.504403114318848, \"eval_ref\": \"mmlu_flan_cot_zeroshot_management__0feb988c__s000__tb262144__... | {
"index": "mmlu_flan_cot_zeroshot_management__0feb988c__s000__tb262144__v000",
"uid": "mmlu_flan_cot_zeroshot_management__0feb988c__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_flan_cot_zeroshot_medical_genetics__6383b0b8__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 82.83152818863711, \"nll_base\": 4.4168087641398115, \"eval_ref\": \"mmlu_flan_cot_zeroshot_medical_genetics__6383b0b8__s00... | {
"index": "mmlu_flan_cot_zeroshot_medical_genetics__6383b0b8__s000__tb131072__v000",
"uid": "mmlu_flan_cot_zeroshot_medical_genetics__6383b0b8__s000__tb131072__v000",
"axis": "domain",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_flan_cot_zeroshot_medical_genetics__6383b0b8__s000__tb65536__v002\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 82.83152818863711, \"nll_base\": 4.4168087641398115, \"eval_ref\": \"mmlu_flan_cot_zeroshot_medical_genetics__6383b0b8__s000_... | {
"index": "mmlu_flan_cot_zeroshot_medical_genetics__6383b0b8__s000__tb65536__v002",
"uid": "mmlu_flan_cot_zeroshot_medical_genetics__6383b0b8__s000__tb65536__v002",
"axis": "domain",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_flan_cot_zeroshot_medical_genetics__6383b0b8__s001__tb131072__v003\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 65.61328082332106, \"nll_base\": 4.183778127034505, \"eval_ref\": \"mmlu_flan_cot_zeroshot_medical_genetics__6383b0b8__s001... | {
"index": "mmlu_flan_cot_zeroshot_medical_genetics__6383b0b8__s001__tb131072__v003",
"uid": "mmlu_flan_cot_zeroshot_medical_genetics__6383b0b8__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\": \"mmlu_flan_cot_zeroshot_miscellaneous__78f81577__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 82.48230850429597, \"nll_base\": 4.412583827972412, \"eval_ref\": \"mmlu_flan_cot_zeroshot_miscellaneous__78f81577__s000__tb6553... | {
"index": "mmlu_flan_cot_zeroshot_miscellaneous__78f81577__s000__tb65536__v000",
"uid": "mmlu_flan_cot_zeroshot_miscellaneous__78f81577__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_flan_cot_zeroshot_moral_disputes__3f8f0507__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 84.09660065036088, \"nll_base\": 4.431966145833333, \"eval_ref\": \"mmlu_flan_cot_zeroshot_moral_disputes__3f8f0507__s000__tb65... | {
"index": "mmlu_flan_cot_zeroshot_moral_disputes__3f8f0507__s000__tb65536__v001",
"uid": "mmlu_flan_cot_zeroshot_moral_disputes__3f8f0507__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_flan_cot_zeroshot_nutrition__32fd360b__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 91.31184790185165, \"nll_base\": 4.514280548095703, \"eval_ref\": \"mmlu_flan_cot_zeroshot_nutrition__32fd360b__s000__tb131072__v0... | {
"index": "mmlu_flan_cot_zeroshot_nutrition__32fd360b__s000__tb131072__v000",
"uid": "mmlu_flan_cot_zeroshot_nutrition__32fd360b__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_flan_cot_zeroshot_philosophy__ea2ea96f__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 80.21285234631475, \"nll_base\": 4.384683755727915, \"eval_ref\": \"mmlu_flan_cot_zeroshot_philosophy__ea2ea96f__s000__tb65536__v00... | {
"index": "mmlu_flan_cot_zeroshot_philosophy__ea2ea96f__s000__tb65536__v000",
"uid": "mmlu_flan_cot_zeroshot_philosophy__ea2ea96f__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_flan_cot_zeroshot_prehistory__9b568473__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 77.97933842787182, \"nll_base\": 4.356443899649161, \"eval_ref\": \"mmlu_flan_cot_zeroshot_prehistory__9b568473__s000__tb131072__... | {
"index": "mmlu_flan_cot_zeroshot_prehistory__9b568473__s000__tb131072__v000",
"uid": "mmlu_flan_cot_zeroshot_prehistory__9b568473__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_flan_cot_zeroshot_prehistory__9b568473__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 77.97933842787182, \"nll_base\": 4.356443899649161, \"eval_ref\": \"mmlu_flan_cot_zeroshot_prehistory__9b568473__s000__tb262144__... | {
"index": "mmlu_flan_cot_zeroshot_prehistory__9b568473__s000__tb262144__v001",
"uid": "mmlu_flan_cot_zeroshot_prehistory__9b568473__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_flan_cot_zeroshot_professional_accounting__dbfed76a__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 76.17955066430672, \"nll_base\": 4.333093062691066, \"eval_ref\": \"mmlu_flan_cot_zeroshot_professional_accounting__... | {
"index": "mmlu_flan_cot_zeroshot_professional_accounting__dbfed76a__s000__tb131072__v000",
"uid": "mmlu_flan_cot_zeroshot_professional_accounting__dbfed76a__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_flan_cot_zeroshot_professional_accounting__dbfed76a__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 76.17955066430672, \"nll_base\": 4.333093062691066, \"eval_ref\": \"mmlu_flan_cot_zeroshot_professional_accounting__... | {
"index": "mmlu_flan_cot_zeroshot_professional_accounting__dbfed76a__s000__tb262144__v001",
"uid": "mmlu_flan_cot_zeroshot_professional_accounting__dbfed76a__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_flan_cot_zeroshot_professional_law__ea95c443__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 80.78627436001288, \"nll_base\": 4.3918070793151855, \"eval_ref\": \"mmlu_flan_cot_zeroshot_professional_law__ea95c443__s000_... | {
"index": "mmlu_flan_cot_zeroshot_professional_law__ea95c443__s000__tb65536__v001",
"uid": "mmlu_flan_cot_zeroshot_professional_law__ea95c443__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_flan_cot_zeroshot_professional_medicine__a5304956__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 92.56206414503997, \"nll_base\": 4.527879383253015, \"eval_ref\": \"mmlu_flan_cot_zeroshot_professional_medicine__a530... | {
"index": "mmlu_flan_cot_zeroshot_professional_medicine__a5304956__s000__tb262144__v000",
"uid": "mmlu_flan_cot_zeroshot_professional_medicine__a5304956__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_flan_cot_zeroshot_professional_medicine__a5304956__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 92.56206414503997, \"nll_base\": 4.527879383253015, \"eval_ref\": \"mmlu_flan_cot_zeroshot_professional_medicine__a53049... | {
"index": "mmlu_flan_cot_zeroshot_professional_medicine__a5304956__s000__tb65536__v001",
"uid": "mmlu_flan_cot_zeroshot_professional_medicine__a5304956__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_flan_cot_zeroshot_sociology__6bd9bd63__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 79.95426152627016, \"nll_base\": 4.381454740251813, \"eval_ref\": \"mmlu_flan_cot_zeroshot_sociology__6bd9bd63__s000__tb131072__v0... | {
"index": "mmlu_flan_cot_zeroshot_sociology__6bd9bd63__s000__tb131072__v001",
"uid": "mmlu_flan_cot_zeroshot_sociology__6bd9bd63__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_flan_cot_zeroshot_us_foreign_policy__d0b8dced__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 79.38445324990424, \"nll_base\": 4.3743025461832685, \"eval_ref\": \"mmlu_flan_cot_zeroshot_us_foreign_policy__d0b8dced__s00... | {
"index": "mmlu_flan_cot_zeroshot_us_foreign_policy__d0b8dced__s000__tb65536__v000",
"uid": "mmlu_flan_cot_zeroshot_us_foreign_policy__d0b8dced__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_flan_cot_zeroshot_virology__06f58552__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 83.00030006428904, \"nll_base\": 4.418844223022461, \"eval_ref\": \"mmlu_flan_cot_zeroshot_virology__06f58552__s000__tb131072__v000... | {
"index": "mmlu_flan_cot_zeroshot_virology__06f58552__s000__tb131072__v000",
"uid": "mmlu_flan_cot_zeroshot_virology__06f58552__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_flan_cot_zeroshot_world_religions__4910c0bb__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 79.16757215281446, \"nll_base\": 4.3715667724609375, \"eval_ref\": \"mmlu_flan_cot_zeroshot_world_religions__4910c0bb__s000_... | {
"index": "mmlu_flan_cot_zeroshot_world_religions__4910c0bb__s000__tb131072__v000",
"uid": "mmlu_flan_cot_zeroshot_world_religions__4910c0bb__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_flan_n_shot_generative_astronomy__d98a3969__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 64.87539459413189, \"nll_base\": 4.172468423843384, \"eval_ref\": \"mmlu_flan_n_shot_generative_astronomy__d98a3969__s000__tb65... | {
"index": "mmlu_flan_n_shot_generative_astronomy__d98a3969__s000__tb65536__v000",
"uid": "mmlu_flan_n_shot_generative_astronomy__d98a3969__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_flan_n_shot_generative_clinical_knowledge__9c6c359a__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 77.10797694339796, \"nll_base\": 4.3452067375183105, \"eval_ref\": \"mmlu_flan_n_shot_generative_clinical_knowledge_... | {
"index": "mmlu_flan_n_shot_generative_clinical_knowledge__9c6c359a__s000__tb131072__v000",
"uid": "mmlu_flan_n_shot_generative_clinical_knowledge__9c6c359a__s000__tb131072__v000",
"axis": "domain",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_flan_n_shot_generative_clinical_knowledge__9c6c359a__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 77.10797694339796, \"nll_base\": 4.3452067375183105, \"eval_ref\": \"mmlu_flan_n_shot_generative_clinical_knowledge_... | {
"index": "mmlu_flan_n_shot_generative_clinical_knowledge__9c6c359a__s000__tb262144__v001",
"uid": "mmlu_flan_n_shot_generative_clinical_knowledge__9c6c359a__s000__tb262144__v001",
"axis": "domain",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_flan_n_shot_generative_clinical_knowledge__9c6c359a__s000__tb65536__v002\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 77.10797694339796, \"nll_base\": 4.3452067375183105, \"eval_ref\": \"mmlu_flan_n_shot_generative_clinical_knowledge__9... | {
"index": "mmlu_flan_n_shot_generative_clinical_knowledge__9c6c359a__s000__tb65536__v002",
"uid": "mmlu_flan_n_shot_generative_clinical_knowledge__9c6c359a__s000__tb65536__v002",
"axis": "domain",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_flan_n_shot_generative_clinical_knowledge__9c6c359a__s001__tb131072__v003\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 76.58235876567285, \"nll_base\": 4.338366746902466, \"eval_ref\": \"mmlu_flan_n_shot_generative_clinical_knowledge__... | {
"index": "mmlu_flan_n_shot_generative_clinical_knowledge__9c6c359a__s001__tb131072__v003",
"uid": "mmlu_flan_n_shot_generative_clinical_knowledge__9c6c359a__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\": \"mmlu_flan_n_shot_generative_college_biology__4979ba5a__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 63.768942996487176, \"nll_base\": 4.155266284942627, \"eval_ref\": \"mmlu_flan_n_shot_generative_college_biology__4979ba5... | {
"index": "mmlu_flan_n_shot_generative_college_biology__4979ba5a__s000__tb65536__v001",
"uid": "mmlu_flan_n_shot_generative_college_biology__4979ba5a__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_flan_n_shot_generative_college_medicine__6a35a87f__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 72.6901067272801, \"nll_base\": 4.286205291748047, \"eval_ref\": \"mmlu_flan_n_shot_generative_college_medicine__6a35a... | {
"index": "mmlu_flan_n_shot_generative_college_medicine__6a35a87f__s000__tb131072__v000",
"uid": "mmlu_flan_n_shot_generative_college_medicine__6a35a87f__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_flan_n_shot_generative_college_physics__27fa3989__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 71.20832538956654, \"nll_base\": 4.2656097412109375, \"eval_ref\": \"mmlu_flan_n_shot_generative_college_physics__27fa398... | {
"index": "mmlu_flan_n_shot_generative_college_physics__27fa3989__s000__tb65536__v001",
"uid": "mmlu_flan_n_shot_generative_college_physics__27fa3989__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_flan_n_shot_generative_computer_security__49cd6902__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 87.19814350594825, \"nll_base\": 4.4681830406188965, \"eval_ref\": \"mmlu_flan_n_shot_generative_computer_security__4... | {
"index": "mmlu_flan_n_shot_generative_computer_security__49cd6902__s000__tb131072__v001",
"uid": "mmlu_flan_n_shot_generative_computer_security__49cd6902__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_flan_n_shot_generative_computer_security__49cd6902__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 87.19814350594825, \"nll_base\": 4.4681830406188965, \"eval_ref\": \"mmlu_flan_n_shot_generative_computer_security__49c... | {
"index": "mmlu_flan_n_shot_generative_computer_security__49cd6902__s000__tb65536__v000",
"uid": "mmlu_flan_n_shot_generative_computer_security__49cd6902__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_flan_n_shot_generative_econometrics__dd53b33d__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 79.34126159170908, \"nll_base\": 4.373758316040039, \"eval_ref\": \"mmlu_flan_n_shot_generative_econometrics__dd53b33d__s000... | {
"index": "mmlu_flan_n_shot_generative_econometrics__dd53b33d__s000__tb65536__v000",
"uid": "mmlu_flan_n_shot_generative_econometrics__dd53b33d__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_flan_n_shot_generative_electrical_engineering__d451c640__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 65.06482266028112, \"nll_base\": 4.175384044647217, \"eval_ref\": \"mmlu_flan_n_shot_generative_electrical_engin... | {
"index": "mmlu_flan_n_shot_generative_electrical_engineering__d451c640__s000__tb131072__v001",
"uid": "mmlu_flan_n_shot_generative_electrical_engineering__d451c640__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_flan_n_shot_generative_electrical_engineering__d451c640__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 65.06482266028112, \"nll_base\": 4.175384044647217, \"eval_ref\": \"mmlu_flan_n_shot_generative_electrical_enginee... | {
"index": "mmlu_flan_n_shot_generative_electrical_engineering__d451c640__s000__tb65536__v000",
"uid": "mmlu_flan_n_shot_generative_electrical_engineering__d451c640__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_flan_n_shot_generative_elementary_mathematics__4c6e1e65__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 77.91979191700644, \"nll_base\": 4.355679988861084, \"eval_ref\": \"mmlu_flan_n_shot_generative_elementary_mathe... | {
"index": "mmlu_flan_n_shot_generative_elementary_mathematics__4c6e1e65__s000__tb131072__v000",
"uid": "mmlu_flan_n_shot_generative_elementary_mathematics__4c6e1e65__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_flan_n_shot_generative_formal_logic__9a505bed__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 75.19970258224856, \"nll_base\": 4.320147275924683, \"eval_ref\": \"mmlu_flan_n_shot_generative_formal_logic__9a505bed__s0... | {
"index": "mmlu_flan_n_shot_generative_formal_logic__9a505bed__s000__tb131072__v001",
"uid": "mmlu_flan_n_shot_generative_formal_logic__9a505bed__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_flan_n_shot_generative_formal_logic__9a505bed__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 75.19970258224856, \"nll_base\": 4.320147275924683, \"eval_ref\": \"mmlu_flan_n_shot_generative_formal_logic__9a505bed__s000... | {
"index": "mmlu_flan_n_shot_generative_formal_logic__9a505bed__s000__tb65536__v000",
"uid": "mmlu_flan_n_shot_generative_formal_logic__9a505bed__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_flan_n_shot_generative_high_school_biology__5b8e1d1c__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 51.92885040952416, \"nll_base\": 3.949874520301819, \"eval_ref\": \"mmlu_flan_n_shot_generative_high_school_biology__... | {
"index": "mmlu_flan_n_shot_generative_high_school_biology__5b8e1d1c__s000__tb65536__v001",
"uid": "mmlu_flan_n_shot_generative_high_school_biology__5b8e1d1c__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_flan_n_shot_generative_high_school_computer_science__5e53ea44__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 72.7159166788928, \"nll_base\": 4.286560297012329, \"eval_ref\": \"mmlu_flan_n_shot_generative_high_school... | {
"index": "mmlu_flan_n_shot_generative_high_school_computer_science__5e53ea44__s000__tb131072__v000",
"uid": "mmlu_flan_n_shot_generative_high_school_computer_science__5e53ea44__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_flan_n_shot_generative_high_school_european_history__a85c6b9e__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 58.57409141774171, \"nll_base\": 4.0702924728393555, \"eval_ref\": \"mmlu_flan_n_shot_generative_high_school... | {
"index": "mmlu_flan_n_shot_generative_high_school_european_history__a85c6b9e__s000__tb65536__v000",
"uid": "mmlu_flan_n_shot_generative_high_school_european_history__a85c6b9e__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_flan_n_shot_generative_high_school_geography__ae7a9033__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 66.56586317365002, \"nll_base\": 4.19819188117981, \"eval_ref\": \"mmlu_flan_n_shot_generative_high_school_geogra... | {
"index": "mmlu_flan_n_shot_generative_high_school_geography__ae7a9033__s000__tb131072__v000",
"uid": "mmlu_flan_n_shot_generative_high_school_geography__ae7a9033__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_flan_n_shot_generative_high_school_government_and_politics__ab44292d__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 65.14038214116553, \"nll_base\": 4.176544666290283, \"eval_ref\": \"mmlu_flan_n_shot_generative_hig... | {
"index": "mmlu_flan_n_shot_generative_high_school_government_and_politics__ab44292d__s000__tb131072__v001",
"uid": "mmlu_flan_n_shot_generative_high_school_government_and_politics__ab44292d__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_flan_n_shot_generative_high_school_government_and_politics__ab44292d__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 65.14038214116553, \"nll_base\": 4.176544666290283, \"eval_ref\": \"mmlu_flan_n_shot_generative_high_... | {
"index": "mmlu_flan_n_shot_generative_high_school_government_and_politics__ab44292d__s000__tb65536__v000",
"uid": "mmlu_flan_n_shot_generative_high_school_government_and_politics__ab44292d__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_flan_n_shot_generative_high_school_mathematics__119ef873__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 72.26495108579114, \"nll_base\": 4.280339241027832, \"eval_ref\": \"mmlu_flan_n_shot_generative_high_school_mat... | {
"index": "mmlu_flan_n_shot_generative_high_school_mathematics__119ef873__s000__tb131072__v001",
"uid": "mmlu_flan_n_shot_generative_high_school_mathematics__119ef873__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_flan_n_shot_generative_high_school_microeconomics__37736bd4__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 66.76245559753302, \"nll_base\": 4.201140880584717, \"eval_ref\": \"mmlu_flan_n_shot_generative_high_school_... | {
"index": "mmlu_flan_n_shot_generative_high_school_microeconomics__37736bd4__s000__tb131072__v001",
"uid": "mmlu_flan_n_shot_generative_high_school_microeconomics__37736bd4__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_flan_n_shot_generative_high_school_microeconomics__37736bd4__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 66.76245559753302, \"nll_base\": 4.201140880584717, \"eval_ref\": \"mmlu_flan_n_shot_generative_high_school_mi... | {
"index": "mmlu_flan_n_shot_generative_high_school_microeconomics__37736bd4__s000__tb65536__v000",
"uid": "mmlu_flan_n_shot_generative_high_school_microeconomics__37736bd4__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_flan_n_shot_generative_high_school_physics__bc2b24c1__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 62.75916592791035, \"nll_base\": 4.1393046379089355, \"eval_ref\": \"mmlu_flan_n_shot_generative_high_school_physic... | {
"index": "mmlu_flan_n_shot_generative_high_school_physics__bc2b24c1__s000__tb131072__v000",
"uid": "mmlu_flan_n_shot_generative_high_school_physics__bc2b24c1__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_flan_n_shot_generative_high_school_psychology__3f4d94e0__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 71.5788156947741, \"nll_base\": 4.270799160003662, \"eval_ref\": \"mmlu_flan_n_shot_generative_high_school_psychol... | {
"index": "mmlu_flan_n_shot_generative_high_school_psychology__3f4d94e0__s000__tb65536__v001",
"uid": "mmlu_flan_n_shot_generative_high_school_psychology__3f4d94e0__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_flan_n_shot_generative_high_school_statistics__9f803d00__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 64.68773641044938, \"nll_base\": 4.1695716381073, \"eval_ref\": \"mmlu_flan_n_shot_generative_high_school_statis... | {
"index": "mmlu_flan_n_shot_generative_high_school_statistics__9f803d00__s000__tb131072__v000",
"uid": "mmlu_flan_n_shot_generative_high_school_statistics__9f803d00__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_flan_n_shot_generative_high_school_us_history__19a89ec9__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 50.439787647912645, \"nll_base\": 3.920780301094055, \"eval_ref\": \"mmlu_flan_n_shot_generative_high_school_us_... | {
"index": "mmlu_flan_n_shot_generative_high_school_us_history__19a89ec9__s000__tb262144__v000",
"uid": "mmlu_flan_n_shot_generative_high_school_us_history__19a89ec9__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_flan_n_shot_generative_high_school_world_history__60c2d5ba__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 52.491582752330764, \"nll_base\": 3.9606528282165527, \"eval_ref\": \"mmlu_flan_n_shot_generative_high_school... | {
"index": "mmlu_flan_n_shot_generative_high_school_world_history__60c2d5ba__s000__tb131072__v000",
"uid": "mmlu_flan_n_shot_generative_high_school_world_history__60c2d5ba__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_flan_n_shot_generative_human_sexuality__81b767ad__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 84.09875941541061, \"nll_base\": 4.431991815567017, \"eval_ref\": \"mmlu_flan_n_shot_generative_human_sexuality__81b767... | {
"index": "mmlu_flan_n_shot_generative_human_sexuality__81b767ad__s000__tb262144__v000",
"uid": "mmlu_flan_n_shot_generative_human_sexuality__81b767ad__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_flan_n_shot_generative_human_sexuality__81b767ad__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 84.09875941541061, \"nll_base\": 4.431991815567017, \"eval_ref\": \"mmlu_flan_n_shot_generative_human_sexuality__81b767ad... | {
"index": "mmlu_flan_n_shot_generative_human_sexuality__81b767ad__s000__tb65536__v001",
"uid": "mmlu_flan_n_shot_generative_human_sexuality__81b767ad__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_flan_n_shot_generative_international_law__fe87691f__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 75.15209821575827, \"nll_base\": 4.319514036178589, \"eval_ref\": \"mmlu_flan_n_shot_generative_international_law__fe87... | {
"index": "mmlu_flan_n_shot_generative_international_law__fe87691f__s000__tb65536__v000",
"uid": "mmlu_flan_n_shot_generative_international_law__fe87691f__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_flan_n_shot_generative_jurisprudence__fd847797__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 73.95528087190641, \"nll_base\": 4.303460597991943, \"eval_ref\": \"mmlu_flan_n_shot_generative_jurisprudence__fd847797__... | {
"index": "mmlu_flan_n_shot_generative_jurisprudence__fd847797__s000__tb131072__v000",
"uid": "mmlu_flan_n_shot_generative_jurisprudence__fd847797__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_flan_n_shot_generative_jurisprudence__fd847797__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 73.95528087190641, \"nll_base\": 4.303460597991943, \"eval_ref\": \"mmlu_flan_n_shot_generative_jurisprudence__fd847797__... | {
"index": "mmlu_flan_n_shot_generative_jurisprudence__fd847797__s000__tb262144__v001",
"uid": "mmlu_flan_n_shot_generative_jurisprudence__fd847797__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_flan_n_shot_generative_logical_fallacies__d2b20ce4__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 64.82432555460686, \"nll_base\": 4.171680927276611, \"eval_ref\": \"mmlu_flan_n_shot_generative_logical_fallacies__d2b2... | {
"index": "mmlu_flan_n_shot_generative_logical_fallacies__d2b20ce4__s000__tb65536__v001",
"uid": "mmlu_flan_n_shot_generative_logical_fallacies__d2b20ce4__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_flan_n_shot_generative_management__06de4edb__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 78.37150834286317, \"nll_base\": 4.361460447311401, \"eval_ref\": \"mmlu_flan_n_shot_generative_management__06de4edb__s000__... | {
"index": "mmlu_flan_n_shot_generative_management__06de4edb__s000__tb262144__v000",
"uid": "mmlu_flan_n_shot_generative_management__06de4edb__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_flan_n_shot_generative_management__06de4edb__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 78.37150834286317, \"nll_base\": 4.361460447311401, \"eval_ref\": \"mmlu_flan_n_shot_generative_management__06de4edb__s000__tb... | {
"index": "mmlu_flan_n_shot_generative_management__06de4edb__s000__tb65536__v001",
"uid": "mmlu_flan_n_shot_generative_management__06de4edb__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_flan_n_shot_generative_medical_genetics__2db4c610__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 73.38276552860509, \"nll_base\": 4.295689105987549, \"eval_ref\": \"mmlu_flan_n_shot_generative_medical_genetics__2db4... | {
"index": "mmlu_flan_n_shot_generative_medical_genetics__2db4c610__s000__tb131072__v001",
"uid": "mmlu_flan_n_shot_generative_medical_genetics__2db4c610__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_flan_n_shot_generative_medical_genetics__2db4c610__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 73.38276552860509, \"nll_base\": 4.295689105987549, \"eval_ref\": \"mmlu_flan_n_shot_generative_medical_genetics__2db4c6... | {
"index": "mmlu_flan_n_shot_generative_medical_genetics__2db4c610__s000__tb65536__v000",
"uid": "mmlu_flan_n_shot_generative_medical_genetics__2db4c610__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_flan_n_shot_generative_moral_disputes__ac2435d2__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 62.58228424477982, \"nll_base\": 4.136482238769531, \"eval_ref\": \"mmlu_flan_n_shot_generative_moral_disputes__ac2435d2... | {
"index": "mmlu_flan_n_shot_generative_moral_disputes__ac2435d2__s000__tb131072__v000",
"uid": "mmlu_flan_n_shot_generative_moral_disputes__ac2435d2__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_flan_n_shot_generative_nutrition__748cb8e2__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 79.14624625300884, \"nll_base\": 4.371297359466553, \"eval_ref\": \"mmlu_flan_n_shot_generative_nutrition__748cb8e2__s000__tb65... | {
"index": "mmlu_flan_n_shot_generative_nutrition__748cb8e2__s000__tb65536__v000",
"uid": "mmlu_flan_n_shot_generative_nutrition__748cb8e2__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_flan_n_shot_generative_prehistory__4f2bc573__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 75.47962792686525, \"nll_base\": 4.323862791061401, \"eval_ref\": \"mmlu_flan_n_shot_generative_prehistory__4f2bc573__s000__... | {
"index": "mmlu_flan_n_shot_generative_prehistory__4f2bc573__s000__tb131072__v000",
"uid": "mmlu_flan_n_shot_generative_prehistory__4f2bc573__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_flan_n_shot_generative_professional_accounting__5c61103a__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 64.61670736522666, \"nll_base\": 4.1684730052948, \"eval_ref\": \"mmlu_flan_n_shot_generative_professional_acco... | {
"index": "mmlu_flan_n_shot_generative_professional_accounting__5c61103a__s000__tb262144__v000",
"uid": "mmlu_flan_n_shot_generative_professional_accounting__5c61103a__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_flan_n_shot_generative_professional_law__91ea12d6__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 54.66991549416714, \"nll_base\": 4.00131356716156, \"eval_ref\": \"mmlu_flan_n_shot_generative_professional_law__91ea1... | {
"index": "mmlu_flan_n_shot_generative_professional_law__91ea12d6__s000__tb131072__v001",
"uid": "mmlu_flan_n_shot_generative_professional_law__91ea12d6__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_flan_n_shot_generative_professional_law__91ea12d6__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 54.66991549416714, \"nll_base\": 4.00131356716156, \"eval_ref\": \"mmlu_flan_n_shot_generative_professional_law__91ea12d... | {
"index": "mmlu_flan_n_shot_generative_professional_law__91ea12d6__s000__tb65536__v000",
"uid": "mmlu_flan_n_shot_generative_professional_law__91ea12d6__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_flan_n_shot_generative_professional_medicine__1bd523dd__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 63.33200249017049, \"nll_base\": 4.148390769958496, \"eval_ref\": \"mmlu_flan_n_shot_generative_professional_medi... | {
"index": "mmlu_flan_n_shot_generative_professional_medicine__1bd523dd__s000__tb131072__v001",
"uid": "mmlu_flan_n_shot_generative_professional_medicine__1bd523dd__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_flan_n_shot_generative_professional_medicine__1bd523dd__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 63.33200249017049, \"nll_base\": 4.148390769958496, \"eval_ref\": \"mmlu_flan_n_shot_generative_professional_medici... | {
"index": "mmlu_flan_n_shot_generative_professional_medicine__1bd523dd__s000__tb65536__v000",
"uid": "mmlu_flan_n_shot_generative_professional_medicine__1bd523dd__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_flan_n_shot_generative_sociology__15bf1a8f__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 72.34916528182946, \"nll_base\": 4.281503915786743, \"eval_ref\": \"mmlu_flan_n_shot_generative_sociology__15bf1a8f__s000__tb... | {
"index": "mmlu_flan_n_shot_generative_sociology__15bf1a8f__s000__tb262144__v000",
"uid": "mmlu_flan_n_shot_generative_sociology__15bf1a8f__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_flan_n_shot_generative_sociology__15bf1a8f__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 72.34916528182946, \"nll_base\": 4.281503915786743, \"eval_ref\": \"mmlu_flan_n_shot_generative_sociology__15bf1a8f__s000__tb65... | {
"index": "mmlu_flan_n_shot_generative_sociology__15bf1a8f__s000__tb65536__v001",
"uid": "mmlu_flan_n_shot_generative_sociology__15bf1a8f__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_flan_n_shot_generative_us_foreign_policy__a87f674d__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 71.86916182660954, \"nll_base\": 4.2748472690582275, \"eval_ref\": \"mmlu_flan_n_shot_generative_us_foreign_policy__a... | {
"index": "mmlu_flan_n_shot_generative_us_foreign_policy__a87f674d__s000__tb131072__v000",
"uid": "mmlu_flan_n_shot_generative_us_foreign_policy__a87f674d__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_flan_n_shot_generative_virology__b3f91a5f__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 74.61002882766107, \"nll_base\": 4.312274932861328, \"eval_ref\": \"mmlu_flan_n_shot_generative_virology__b3f91a5f__s000__tb13... | {
"index": "mmlu_flan_n_shot_generative_virology__b3f91a5f__s000__tb131072__v000",
"uid": "mmlu_flan_n_shot_generative_virology__b3f91a5f__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_flan_n_shot_loglikelihood_anatomy__c5725167__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 65.41189461815718, \"nll_base\": 4.180704116821289, \"eval_ref\": \"mmlu_flan_n_shot_loglikelihood_anatomy__c5725167__s000__... | {
"index": "mmlu_flan_n_shot_loglikelihood_anatomy__c5725167__s000__tb131072__v000",
"uid": "mmlu_flan_n_shot_loglikelihood_anatomy__c5725167__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_flan_n_shot_loglikelihood_anatomy__c5725167__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 65.41189461815718, \"nll_base\": 4.180704116821289, \"eval_ref\": \"mmlu_flan_n_shot_loglikelihood_anatomy__c5725167__s000__... | {
"index": "mmlu_flan_n_shot_loglikelihood_anatomy__c5725167__s000__tb262144__v001",
"uid": "mmlu_flan_n_shot_loglikelihood_anatomy__c5725167__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_flan_n_shot_loglikelihood_astronomy__bf09b0e2__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 71.80145930393945, \"nll_base\": 4.273904800415039, \"eval_ref\": \"mmlu_flan_n_shot_loglikelihood_astronomy__bf09b0e2__s0... | {
"index": "mmlu_flan_n_shot_loglikelihood_astronomy__bf09b0e2__s000__tb131072__v000",
"uid": "mmlu_flan_n_shot_loglikelihood_astronomy__bf09b0e2__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_flan_n_shot_loglikelihood_business_ethics__4c30bbf9__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 73.15518368577031, \"nll_base\": 4.292582988739014, \"eval_ref\": \"mmlu_flan_n_shot_loglikelihood_business_ethics__... | {
"index": "mmlu_flan_n_shot_loglikelihood_business_ethics__4c30bbf9__s000__tb131072__v000",
"uid": "mmlu_flan_n_shot_loglikelihood_business_ethics__4c30bbf9__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_flan_n_shot_loglikelihood_clinical_knowledge__6060a91e__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 75.16492835286348, \"nll_base\": 4.319684743881226, \"eval_ref\": \"mmlu_flan_n_shot_loglikelihood_clinical_knowl... | {
"index": "mmlu_flan_n_shot_loglikelihood_clinical_knowledge__6060a91e__s000__tb131072__v000",
"uid": "mmlu_flan_n_shot_loglikelihood_clinical_knowledge__6060a91e__s000__tb131072__v000",
"axis": "domain",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_flan_n_shot_loglikelihood_clinical_knowledge__6060a91e__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 75.16492835286348, \"nll_base\": 4.319684743881226, \"eval_ref\": \"mmlu_flan_n_shot_loglikelihood_clinical_knowl... | {
"index": "mmlu_flan_n_shot_loglikelihood_clinical_knowledge__6060a91e__s000__tb262144__v001",
"uid": "mmlu_flan_n_shot_loglikelihood_clinical_knowledge__6060a91e__s000__tb262144__v001",
"axis": "domain",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_flan_n_shot_loglikelihood_clinical_knowledge__6060a91e__s000__tb65536__v002\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 75.16492835286348, \"nll_base\": 4.319684743881226, \"eval_ref\": \"mmlu_flan_n_shot_loglikelihood_clinical_knowled... | {
"index": "mmlu_flan_n_shot_loglikelihood_clinical_knowledge__6060a91e__s000__tb65536__v002",
"uid": "mmlu_flan_n_shot_loglikelihood_clinical_knowledge__6060a91e__s000__tb65536__v002",
"axis": "domain",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_flan_n_shot_loglikelihood_clinical_knowledge__6060a91e__s001__tb131072__v003\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 72.89432307185703, \"nll_base\": 4.289010763168335, \"eval_ref\": \"mmlu_flan_n_shot_loglikelihood_clinical_knowl... | {
"index": "mmlu_flan_n_shot_loglikelihood_clinical_knowledge__6060a91e__s001__tb131072__v003",
"uid": "mmlu_flan_n_shot_loglikelihood_clinical_knowledge__6060a91e__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\": \"mmlu_flan_n_shot_loglikelihood_college_biology__dcf79b29__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 61.71421219443975, \"nll_base\": 4.122514247894287, \"eval_ref\": \"mmlu_flan_n_shot_loglikelihood_college_biology__... | {
"index": "mmlu_flan_n_shot_loglikelihood_college_biology__dcf79b29__s000__tb131072__v000",
"uid": "mmlu_flan_n_shot_loglikelihood_college_biology__dcf79b29__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_flan_n_shot_loglikelihood_college_chemistry__c41b531f__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 70.64395346676295, \"nll_base\": 4.257652521133423, \"eval_ref\": \"mmlu_flan_n_shot_loglikelihood_college_chemist... | {
"index": "mmlu_flan_n_shot_loglikelihood_college_chemistry__c41b531f__s000__tb131072__v001",
"uid": "mmlu_flan_n_shot_loglikelihood_college_chemistry__c41b531f__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_flan_n_shot_loglikelihood_college_chemistry__c41b531f__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 70.64395346676295, \"nll_base\": 4.257652521133423, \"eval_ref\": \"mmlu_flan_n_shot_loglikelihood_college_chemistry... | {
"index": "mmlu_flan_n_shot_loglikelihood_college_chemistry__c41b531f__s000__tb65536__v000",
"uid": "mmlu_flan_n_shot_loglikelihood_college_chemistry__c41b531f__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_flan_n_shot_loglikelihood_college_medicine__3858a5f8__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 69.84625055887093, \"nll_base\": 4.246296405792236, \"eval_ref\": \"mmlu_flan_n_shot_loglikelihood_college_medicine... | {
"index": "mmlu_flan_n_shot_loglikelihood_college_medicine__3858a5f8__s000__tb131072__v000",
"uid": "mmlu_flan_n_shot_loglikelihood_college_medicine__3858a5f8__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_flan_n_shot_loglikelihood_college_physics__52ee623c__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 70.42716584627246, \"nll_base\": 4.254579067230225, \"eval_ref\": \"mmlu_flan_n_shot_loglikelihood_college_physics__... | {
"index": "mmlu_flan_n_shot_loglikelihood_college_physics__52ee623c__s000__tb131072__v001",
"uid": "mmlu_flan_n_shot_loglikelihood_college_physics__52ee623c__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_flan_n_shot_loglikelihood_computer_security__66c9d007__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 89.6696297019787, \"nll_base\": 4.496132135391235, \"eval_ref\": \"mmlu_flan_n_shot_loglikelihood_computer_securit... | {
"index": "mmlu_flan_n_shot_loglikelihood_computer_security__66c9d007__s000__tb262144__v000",
"uid": "mmlu_flan_n_shot_loglikelihood_computer_security__66c9d007__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.