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_philosophy_continuation__8f294686__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 8.028391031915913, \"nll_base\": 2.082984138257576, \"eval_ref\": \"mmlu_philosophy_continuation__8f294686__s000__tb131072__v000\", \"... | {
"index": "mmlu_philosophy_continuation__8f294686__s000__tb131072__v000",
"uid": "mmlu_philosophy_continuation__8f294686__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_philosophy_generative__a5009ed1__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1313.9332658974374, \"nll_base\": 7.180780410766602, \"eval_ref\": \"mmlu_philosophy_generative__a5009ed1__s000__tb262144__v000\", \"ora... | {
"index": "mmlu_philosophy_generative__a5009ed1__s000__tb262144__v000",
"uid": "mmlu_philosophy_generative__a5009ed1__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_philosophy_generative__a5009ed1__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1313.9332658974374, \"nll_base\": 7.180780410766602, \"eval_ref\": \"mmlu_philosophy_generative__a5009ed1__s000__tb65536__v001\", \"oracle... | {
"index": "mmlu_philosophy_generative__a5009ed1__s000__tb65536__v001",
"uid": "mmlu_philosophy_generative__a5009ed1__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_philosophy_generative_spanish__ccb45599__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2574.8157778984937, \"nll_base\": 7.8535332679748535, \"eval_ref\": \"mmlu_philosophy_generative_spanish__ccb45599__s000__tb1310... | {
"index": "mmlu_philosophy_generative_spanish__ccb45599__s000__tb131072__v000",
"uid": "mmlu_philosophy_generative_spanish__ccb45599__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_philosophy_generative_spanish__ccb45599__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2574.8157778984937, \"nll_base\": 7.8535332679748535, \"eval_ref\": \"mmlu_philosophy_generative_spanish__ccb45599__s000__tb2621... | {
"index": "mmlu_philosophy_generative_spanish__ccb45599__s000__tb262144__v001",
"uid": "mmlu_philosophy_generative_spanish__ccb45599__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_prehistory__528327ed__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1462.671585369278, \"nll_base\": 7.288019895553589, \"eval_ref\": \"mmlu_prehistory__528327ed__s000__tb262144__v000\", \"oracle_delta\": 1.45029234... | {
"index": "mmlu_prehistory__528327ed__s000__tb262144__v000",
"uid": "mmlu_prehistory__528327ed__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_prehistory_continuation__1bcbba34__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 5.7328177567547325, \"nll_base\": 1.7462071649602688, \"eval_ref\": \"mmlu_prehistory_continuation__1bcbba34__s000__tb65536__v000\", \"o... | {
"index": "mmlu_prehistory_continuation__1bcbba34__s000__tb65536__v000",
"uid": "mmlu_prehistory_continuation__1bcbba34__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_prehistory_generative__d8d0716f__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1822.6362774036415, \"nll_base\": 7.508039236068726, \"eval_ref\": \"mmlu_prehistory_generative__d8d0716f__s000__tb131072__v000\", \"ora... | {
"index": "mmlu_prehistory_generative__d8d0716f__s000__tb131072__v000",
"uid": "mmlu_prehistory_generative__d8d0716f__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_pro_llama_math__23eecdf2__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 8635.18608263135, \"nll_base\": 9.063600540161133, \"eval_ref\": \"mmlu_pro_llama_math__23eecdf2__s000__tb262144__v000\", \"oracle_delta\": 0.3... | {
"index": "mmlu_pro_llama_math__23eecdf2__s000__tb262144__v000",
"uid": "mmlu_pro_llama_math__23eecdf2__s000__tb262144__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_pro_plus_biology__b93b3bdb__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2988.8559549136203, \"nll_base\": 8.00264596939087, \"eval_ref\": \"mmlu_pro_plus_biology__b93b3bdb__s000__tb262144__v000\", \"oracle_delta\"... | {
"index": "mmlu_pro_plus_biology__b93b3bdb__s000__tb262144__v000",
"uid": "mmlu_pro_plus_biology__b93b3bdb__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_pro_plus_computer_science__bfdbf472__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 6236.333668713335, \"nll_base\": 8.738147735595703, \"eval_ref\": \"mmlu_pro_plus_computer_science__bfdbf472__s000__tb131072__v000\"... | {
"index": "mmlu_pro_plus_computer_science__bfdbf472__s000__tb131072__v000",
"uid": "mmlu_pro_plus_computer_science__bfdbf472__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_pro_plus_economics__6c3c801e__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4470.814517050708, \"nll_base\": 8.405325889587402, \"eval_ref\": \"mmlu_pro_plus_economics__6c3c801e__s000__tb131072__v000\", \"oracle_del... | {
"index": "mmlu_pro_plus_economics__6c3c801e__s000__tb131072__v000",
"uid": "mmlu_pro_plus_economics__6c3c801e__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_pro_plus_economics__6c3c801e__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 4470.814517050708, \"nll_base\": 8.405325889587402, \"eval_ref\": \"mmlu_pro_plus_economics__6c3c801e__s000__tb262144__v001\", \"oracle_del... | {
"index": "mmlu_pro_plus_economics__6c3c801e__s000__tb262144__v001",
"uid": "mmlu_pro_plus_economics__6c3c801e__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_pro_plus_history__42ddbb22__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 5083.689191119093, \"nll_base\": 8.533792495727539, \"eval_ref\": \"mmlu_pro_plus_history__42ddbb22__s000__tb131072__v001\", \"oracle_delta\"... | {
"index": "mmlu_pro_plus_history__42ddbb22__s000__tb131072__v001",
"uid": "mmlu_pro_plus_history__42ddbb22__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_pro_plus_history__42ddbb22__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 5083.689191119093, \"nll_base\": 8.533792495727539, \"eval_ref\": \"mmlu_pro_plus_history__42ddbb22__s000__tb65536__v000\", \"oracle_delta\": 0... | {
"index": "mmlu_pro_plus_history__42ddbb22__s000__tb65536__v000",
"uid": "mmlu_pro_plus_history__42ddbb22__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_professional_accounting_continuation__63f52efd__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 5.661711259295304, \"nll_base\": 1.7337261891504476, \"eval_ref\": \"mmlu_professional_accounting_continuation__63f52efd__s... | {
"index": "mmlu_professional_accounting_continuation__63f52efd__s000__tb65536__v000",
"uid": "mmlu_professional_accounting_continuation__63f52efd__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_professional_accounting_generative__978eb9ca__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 805.1204378868163, \"nll_base\": 6.6909918785095215, \"eval_ref\": \"mmlu_professional_accounting_generative__978eb9ca__s00... | {
"index": "mmlu_professional_accounting_generative__978eb9ca__s000__tb131072__v000",
"uid": "mmlu_professional_accounting_generative__978eb9ca__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_professional_accounting_generative_spanish__40077ee4__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2534.9960773352705, \"nll_base\": 7.837947368621826, \"eval_ref\": \"mmlu_professional_accounting_generative_spanish_... | {
"index": "mmlu_professional_accounting_generative_spanish__40077ee4__s000__tb65536__v000",
"uid": "mmlu_professional_accounting_generative_spanish__40077ee4__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_professional_law__548be44d__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 708.5419700497318, \"nll_base\": 6.563209295272827, \"eval_ref\": \"mmlu_professional_law__548be44d__s000__tb65536__v000\", \"oracle_delta\": 0... | {
"index": "mmlu_professional_law__548be44d__s000__tb65536__v000",
"uid": "mmlu_professional_law__548be44d__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_professional_law_continuation__ffff40a5__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2.116625281894577, \"nll_base\": 0.7498229721845207, \"eval_ref\": \"mmlu_professional_law_continuation__ffff40a5__s000__tb65536__... | {
"index": "mmlu_professional_law_continuation__ffff40a5__s000__tb65536__v001",
"uid": "mmlu_professional_law_continuation__ffff40a5__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_professional_law_generative__c276e003__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 620.265670010554, \"nll_base\": 6.430147886276245, \"eval_ref\": \"mmlu_professional_law_generative__c276e003__s000__tb131072__v00... | {
"index": "mmlu_professional_law_generative__c276e003__s000__tb131072__v001",
"uid": "mmlu_professional_law_generative__c276e003__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_professional_law_generative__c276e003__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 620.265670010554, \"nll_base\": 6.430147886276245, \"eval_ref\": \"mmlu_professional_law_generative__c276e003__s000__tb65536__v000\"... | {
"index": "mmlu_professional_law_generative__c276e003__s000__tb65536__v000",
"uid": "mmlu_professional_law_generative__c276e003__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_professional_law_generative_spanish__01ef41a3__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1292.7135381429184, \"nll_base\": 7.164498805999756, \"eval_ref\": \"mmlu_professional_law_generative_spanish__01ef41a3__s00... | {
"index": "mmlu_professional_law_generative_spanish__01ef41a3__s000__tb65536__v000",
"uid": "mmlu_professional_law_generative_spanish__01ef41a3__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_professional_medicine__3658baa5__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1310.817462241634, \"nll_base\": 7.178406238555908, \"eval_ref\": \"mmlu_professional_medicine__3658baa5__s000__tb131072__v000\", \"orac... | {
"index": "mmlu_professional_medicine__3658baa5__s000__tb131072__v000",
"uid": "mmlu_professional_medicine__3658baa5__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_professional_medicine_continuation__8e9be255__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 15.072503162941064, \"nll_base\": 2.7128721012298715, \"eval_ref\": \"mmlu_professional_medicine_continuation__8e9be255__s0... | {
"index": "mmlu_professional_medicine_continuation__8e9be255__s000__tb131072__v001",
"uid": "mmlu_professional_medicine_continuation__8e9be255__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_professional_medicine_continuation__8e9be255__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 15.072503162941064, \"nll_base\": 2.7128721012298715, \"eval_ref\": \"mmlu_professional_medicine_continuation__8e9be255__s000... | {
"index": "mmlu_professional_medicine_continuation__8e9be255__s000__tb65536__v000",
"uid": "mmlu_professional_medicine_continuation__8e9be255__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_professional_medicine_generative__626ad258__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1446.5983995858855, \"nll_base\": 7.276970148086548, \"eval_ref\": \"mmlu_professional_medicine_generative__626ad258__s000__t... | {
"index": "mmlu_professional_medicine_generative__626ad258__s000__tb131072__v000",
"uid": "mmlu_professional_medicine_generative__626ad258__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_professional_medicine_generative_spanish__4789f93d__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2824.9478789508744, \"nll_base\": 7.946245193481445, \"eval_ref\": \"mmlu_professional_medicine_generative_spanish__4... | {
"index": "mmlu_professional_medicine_generative_spanish__4789f93d__s000__tb131072__v000",
"uid": "mmlu_professional_medicine_generative_spanish__4789f93d__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_professional_medicine_generative_spanish__4789f93d__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2824.9478789508744, \"nll_base\": 7.946245193481445, \"eval_ref\": \"mmlu_professional_medicine_generative_spanish__4... | {
"index": "mmlu_professional_medicine_generative_spanish__4789f93d__s000__tb262144__v001",
"uid": "mmlu_professional_medicine_generative_spanish__4789f93d__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_professional_psychology_generative_spanish__37070d10__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1468.9450182586118, \"nll_base\": 7.292299747467041, \"eval_ref\": \"mmlu_professional_psychology_generative_spanis... | {
"index": "mmlu_professional_psychology_generative_spanish__37070d10__s000__tb262144__v000",
"uid": "mmlu_professional_psychology_generative_spanish__37070d10__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_professional_psychology_generative_spanish__37070d10__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1468.9450182586118, \"nll_base\": 7.292299747467041, \"eval_ref\": \"mmlu_professional_psychology_generative_spanish_... | {
"index": "mmlu_professional_psychology_generative_spanish__37070d10__s000__tb65536__v001",
"uid": "mmlu_professional_psychology_generative_spanish__37070d10__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_prox_af_business__f9b5e597__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3059.287481965134, \"nll_base\": 8.02593731880188, \"eval_ref\": \"mmlu_prox_af_business__f9b5e597__s000__tb65536__v000\", \"oracle_delta\": 0.... | {
"index": "mmlu_prox_af_business__f9b5e597__s000__tb65536__v000",
"uid": "mmlu_prox_af_business__f9b5e597__s000__tb65536__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_af_computer_science__dfac89cc__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2605.271862938587, \"nll_base\": 7.865292310714722, \"eval_ref\": \"mmlu_prox_af_computer_science__dfac89cc__s000__tb131072__v000\", ... | {
"index": "mmlu_prox_af_computer_science__dfac89cc__s000__tb131072__v000",
"uid": "mmlu_prox_af_computer_science__dfac89cc__s000__tb131072__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_af_law__b7e6f83e__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1623.5903769271044, \"nll_base\": 7.392395257949829, \"eval_ref\": \"mmlu_prox_af_law__b7e6f83e__s000__tb65536__v000\", \"oracle_delta\": 0.09174919... | {
"index": "mmlu_prox_af_law__b7e6f83e__s000__tb65536__v000",
"uid": "mmlu_prox_af_law__b7e6f83e__s000__tb65536__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_af_math__800893d7__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2786.4408781952434, \"nll_base\": 7.932520389556885, \"eval_ref\": \"mmlu_prox_af_math__800893d7__s000__tb262144__v000\", \"oracle_delta\": 1.683... | {
"index": "mmlu_prox_af_math__800893d7__s000__tb262144__v000",
"uid": "mmlu_prox_af_math__800893d7__s000__tb262144__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_af_other__08842fee__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3182.0006624965, \"nll_base\": 8.065265417098999, \"eval_ref\": \"mmlu_prox_af_other__08842fee__s000__tb65536__v000\", \"oracle_delta\": 0.2390697... | {
"index": "mmlu_prox_af_other__08842fee__s000__tb65536__v000",
"uid": "mmlu_prox_af_other__08842fee__s000__tb65536__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_af_philosophy__e4ece627__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2319.7388202135107, \"nll_base\": 7.749209880828857, \"eval_ref\": \"mmlu_prox_af_philosophy__e4ece627__s000__tb262144__v000\", \"oracle_de... | {
"index": "mmlu_prox_af_philosophy__e4ece627__s000__tb262144__v000",
"uid": "mmlu_prox_af_philosophy__e4ece627__s000__tb262144__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_af_psychology__1c8f075d__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2116.163034574389, \"nll_base\": 7.657359838485718, \"eval_ref\": \"mmlu_prox_af_psychology__1c8f075d__s000__tb65536__v000\", \"oracle_delta\... | {
"index": "mmlu_prox_af_psychology__1c8f075d__s000__tb65536__v000",
"uid": "mmlu_prox_af_psychology__1c8f075d__s000__tb65536__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_ar_business__873bc99d__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4595.0370839378775, \"nll_base\": 8.432732105255127, \"eval_ref\": \"mmlu_prox_ar_business__873bc99d__s000__tb131072__v000\", \"oracle_delta\... | {
"index": "mmlu_prox_ar_business__873bc99d__s000__tb131072__v000",
"uid": "mmlu_prox_ar_business__873bc99d__s000__tb131072__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_ar_chemistry__a89223cb__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4673.106060504942, \"nll_base\": 8.449579238891602, \"eval_ref\": \"mmlu_prox_ar_chemistry__a89223cb__s000__tb65536__v000\", \"oracle_delta\":... | {
"index": "mmlu_prox_ar_chemistry__a89223cb__s000__tb65536__v000",
"uid": "mmlu_prox_ar_chemistry__a89223cb__s000__tb65536__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_ar_computer_science__b5f6ee47__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4513.791079377754, \"nll_base\": 8.414892673492432, \"eval_ref\": \"mmlu_prox_ar_computer_science__b5f6ee47__s000__tb131072__v000\", ... | {
"index": "mmlu_prox_ar_computer_science__b5f6ee47__s000__tb131072__v000",
"uid": "mmlu_prox_ar_computer_science__b5f6ee47__s000__tb131072__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_ar_economics__8039af0a__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3112.600546349767, \"nll_base\": 8.043213844299316, \"eval_ref\": \"mmlu_prox_ar_economics__8039af0a__s000__tb65536__v000\", \"oracle_delta\":... | {
"index": "mmlu_prox_ar_economics__8039af0a__s000__tb65536__v000",
"uid": "mmlu_prox_ar_economics__8039af0a__s000__tb65536__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_ar_health__671e426a__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4407.900651745253, \"nll_base\": 8.391153812408447, \"eval_ref\": \"mmlu_prox_ar_health__671e426a__s000__tb131072__v000\", \"oracle_delta\": 0.... | {
"index": "mmlu_prox_ar_health__671e426a__s000__tb131072__v000",
"uid": "mmlu_prox_ar_health__671e426a__s000__tb131072__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_ar_law__d264ee98__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2320.9896459533607, \"nll_base\": 7.749748945236206, \"eval_ref\": \"mmlu_prox_ar_law__d264ee98__s000__tb131072__v000\", \"oracle_delta\": 0.10840... | {
"index": "mmlu_prox_ar_law__d264ee98__s000__tb131072__v000",
"uid": "mmlu_prox_ar_law__d264ee98__s000__tb131072__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_ar_philosophy__e27088c3__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4379.021655877906, \"nll_base\": 8.384580612182617, \"eval_ref\": \"mmlu_prox_ar_philosophy__e27088c3__s000__tb65536__v000\", \"oracle_delta\... | {
"index": "mmlu_prox_ar_philosophy__e27088c3__s000__tb65536__v000",
"uid": "mmlu_prox_ar_philosophy__e27088c3__s000__tb65536__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_bn_math__b830ab32__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 9994.890890276354, \"nll_base\": 9.209829330444336, \"eval_ref\": \"mmlu_prox_bn_math__b830ab32__s000__tb65536__v000\", \"oracle_delta\": 0.1738395... | {
"index": "mmlu_prox_bn_math__b830ab32__s000__tb65536__v000",
"uid": "mmlu_prox_bn_math__b830ab32__s000__tb65536__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_cs_biology__58306de6__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3917.2352247102126, \"nll_base\": 8.273141384124756, \"eval_ref\": \"mmlu_prox_cs_biology__58306de6__s000__tb65536__v000\", \"oracle_delta\": 0.... | {
"index": "mmlu_prox_cs_biology__58306de6__s000__tb65536__v000",
"uid": "mmlu_prox_cs_biology__58306de6__s000__tb65536__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_cs_business__313d7af4__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 5640.18119439644, \"nll_base\": 8.63767147064209, \"eval_ref\": \"mmlu_prox_cs_business__313d7af4__s000__tb65536__v000\", \"oracle_delta\": 0.1... | {
"index": "mmlu_prox_cs_business__313d7af4__s000__tb65536__v000",
"uid": "mmlu_prox_cs_business__313d7af4__s000__tb65536__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_cs_chemistry__3a5db406__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3450.074368315018, \"nll_base\": 8.146151065826416, \"eval_ref\": \"mmlu_prox_cs_chemistry__3a5db406__s000__tb65536__v000\", \"oracle_delta\":... | {
"index": "mmlu_prox_cs_chemistry__3a5db406__s000__tb65536__v000",
"uid": "mmlu_prox_cs_chemistry__3a5db406__s000__tb65536__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_cs_computer_science__18f4112b__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3141.270042242392, \"nll_base\": 8.052382469177246, \"eval_ref\": \"mmlu_prox_cs_computer_science__18f4112b__s000__tb131072__v000\", ... | {
"index": "mmlu_prox_cs_computer_science__18f4112b__s000__tb131072__v000",
"uid": "mmlu_prox_cs_computer_science__18f4112b__s000__tb131072__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_cs_economics__f3a5b456__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 5002.278046207764, \"nll_base\": 8.517648696899414, \"eval_ref\": \"mmlu_prox_cs_economics__f3a5b456__s000__tb65536__v000\", \"oracle_delta\":... | {
"index": "mmlu_prox_cs_economics__f3a5b456__s000__tb65536__v000",
"uid": "mmlu_prox_cs_economics__f3a5b456__s000__tb65536__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_cs_engineering__8737bfe2__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3420.4812130417645, \"nll_base\": 8.137536525726318, \"eval_ref\": \"mmlu_prox_cs_engineering__8737bfe2__s000__tb131072__v000\", \"oracle_... | {
"index": "mmlu_prox_cs_engineering__8737bfe2__s000__tb131072__v000",
"uid": "mmlu_prox_cs_engineering__8737bfe2__s000__tb131072__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_cs_history__abf6c1fd__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3224.3731348381075, \"nll_base\": 8.07849383354187, \"eval_ref\": \"mmlu_prox_cs_history__abf6c1fd__s000__tb131072__v000\", \"oracle_delta\": ... | {
"index": "mmlu_prox_cs_history__abf6c1fd__s000__tb131072__v000",
"uid": "mmlu_prox_cs_history__abf6c1fd__s000__tb131072__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_cs_law__23abcb90__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4421.857973610306, \"nll_base\": 8.394315242767334, \"eval_ref\": \"mmlu_prox_cs_law__23abcb90__s000__tb65536__v000\", \"oracle_delta\": 0.159305095... | {
"index": "mmlu_prox_cs_law__23abcb90__s000__tb65536__v000",
"uid": "mmlu_prox_cs_law__23abcb90__s000__tb65536__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_cs_math__3261a07f__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 5609.711664675771, \"nll_base\": 8.632254600524902, \"eval_ref\": \"mmlu_prox_cs_math__3261a07f__s000__tb65536__v000\", \"oracle_delta\": 0.1748490... | {
"index": "mmlu_prox_cs_math__3261a07f__s000__tb65536__v000",
"uid": "mmlu_prox_cs_math__3261a07f__s000__tb65536__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_cs_other__c9ad4b01__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3799.2735448442572, \"nll_base\": 8.242565155029297, \"eval_ref\": \"mmlu_prox_cs_other__c9ad4b01__s000__tb131072__v000\", \"oracle_delta\": 0.4... | {
"index": "mmlu_prox_cs_other__c9ad4b01__s000__tb131072__v000",
"uid": "mmlu_prox_cs_other__c9ad4b01__s000__tb131072__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_cs_physics__6d9877fe__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 4806.48445959435, \"nll_base\": 8.477721214294434, \"eval_ref\": \"mmlu_prox_cs_physics__6d9877fe__s000__tb262144__v000\", \"oracle_delta\": 1... | {
"index": "mmlu_prox_cs_physics__6d9877fe__s000__tb262144__v000",
"uid": "mmlu_prox_cs_physics__6d9877fe__s000__tb262144__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_cs_psychology__efcff8cb__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4749.891506931456, \"nll_base\": 8.465877056121826, \"eval_ref\": \"mmlu_prox_cs_psychology__efcff8cb__s000__tb65536__v000\", \"oracle_delta\... | {
"index": "mmlu_prox_cs_psychology__efcff8cb__s000__tb65536__v000",
"uid": "mmlu_prox_cs_psychology__efcff8cb__s000__tb65536__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_de_biology__0a32c211__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2802.595762775365, \"nll_base\": 7.93830132484436, \"eval_ref\": \"mmlu_prox_de_biology__0a32c211__s000__tb65536__v000\", \"oracle_delta\": 0.00... | {
"index": "mmlu_prox_de_biology__0a32c211__s000__tb65536__v000",
"uid": "mmlu_prox_de_biology__0a32c211__s000__tb65536__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_de_computer_science__7402683c__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2664.5099103750276, \"nll_base\": 7.887775421142578, \"eval_ref\": \"mmlu_prox_de_computer_science__7402683c__s000__tb65536__v000\", \"... | {
"index": "mmlu_prox_de_computer_science__7402683c__s000__tb65536__v000",
"uid": "mmlu_prox_de_computer_science__7402683c__s000__tb65536__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_de_economics__bc313756__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3273.712381934057, \"nll_base\": 8.093679904937744, \"eval_ref\": \"mmlu_prox_de_economics__bc313756__s000__tb262144__v000\", \"oracle_delta... | {
"index": "mmlu_prox_de_economics__bc313756__s000__tb262144__v000",
"uid": "mmlu_prox_de_economics__bc313756__s000__tb262144__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_de_engineering__f1a6d46c__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2995.204817668059, \"nll_base\": 8.004767894744873, \"eval_ref\": \"mmlu_prox_de_engineering__f1a6d46c__s000__tb262144__v000\", \"oracle_d... | {
"index": "mmlu_prox_de_engineering__f1a6d46c__s000__tb262144__v000",
"uid": "mmlu_prox_de_engineering__f1a6d46c__s000__tb262144__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_de_law__58c848e2__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2342.322548517989, \"nll_base\": 7.7588982582092285, \"eval_ref\": \"mmlu_prox_de_law__58c848e2__s000__tb131072__v000\", \"oracle_delta\": 0.14415... | {
"index": "mmlu_prox_de_law__58c848e2__s000__tb131072__v000",
"uid": "mmlu_prox_de_law__58c848e2__s000__tb131072__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_de_other__eb6589a9__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3258.3288085005647, \"nll_base\": 8.088969707489014, \"eval_ref\": \"mmlu_prox_de_other__eb6589a9__s000__tb65536__v000\", \"oracle_delta\": 0.1930... | {
"index": "mmlu_prox_de_other__eb6589a9__s000__tb65536__v000",
"uid": "mmlu_prox_de_other__eb6589a9__s000__tb65536__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_de_psychology__b3f87886__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3459.7629188958886, \"nll_base\": 8.148955345153809, \"eval_ref\": \"mmlu_prox_de_psychology__b3f87886__s000__tb131072__v000\", \"oracle_de... | {
"index": "mmlu_prox_de_psychology__b3f87886__s000__tb131072__v000",
"uid": "mmlu_prox_de_psychology__b3f87886__s000__tb131072__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_en_biology__ec74355a__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3518.1008571689476, \"nll_base\": 8.165676593780518, \"eval_ref\": \"mmlu_prox_en_biology__ec74355a__s000__tb262144__v000\", \"oracle_delta\":... | {
"index": "mmlu_prox_en_biology__ec74355a__s000__tb262144__v000",
"uid": "mmlu_prox_en_biology__ec74355a__s000__tb262144__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_en_economics__683bfee7__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2526.048798202395, \"nll_base\": 7.83441162109375, \"eval_ref\": \"mmlu_prox_en_economics__683bfee7__s000__tb131072__v000\", \"oracle_delta\... | {
"index": "mmlu_prox_en_economics__683bfee7__s000__tb131072__v000",
"uid": "mmlu_prox_en_economics__683bfee7__s000__tb131072__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_es_business__3ba60a24__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 5683.965041345758, \"nll_base\": 8.64540433883667, \"eval_ref\": \"mmlu_prox_es_business__3ba60a24__s000__tb65536__v000\", \"oracle_delta\": 0.... | {
"index": "mmlu_prox_es_business__3ba60a24__s000__tb65536__v000",
"uid": "mmlu_prox_es_business__3ba60a24__s000__tb65536__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_es_economics__946b35b1__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2178.4599740383433, \"nll_base\": 7.686373472213745, \"eval_ref\": \"mmlu_prox_es_economics__946b35b1__s000__tb131072__v000\", \"oracle_delt... | {
"index": "mmlu_prox_es_economics__946b35b1__s000__tb131072__v000",
"uid": "mmlu_prox_es_economics__946b35b1__s000__tb131072__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_es_engineering__8a709d67__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2579.6872521473083, \"nll_base\": 7.855423450469971, \"eval_ref\": \"mmlu_prox_es_engineering__8a709d67__s000__tb262144__v000\", \"oracle_... | {
"index": "mmlu_prox_es_engineering__8a709d67__s000__tb262144__v000",
"uid": "mmlu_prox_es_engineering__8a709d67__s000__tb262144__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_es_history__6dd99088__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2886.8858939328816, \"nll_base\": 7.967933654785156, \"eval_ref\": \"mmlu_prox_es_history__6dd99088__s000__tb262144__v000\", \"oracle_delta\":... | {
"index": "mmlu_prox_es_history__6dd99088__s000__tb262144__v000",
"uid": "mmlu_prox_es_history__6dd99088__s000__tb262144__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_es_math__ab611024__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 5441.105299337839, \"nll_base\": 8.60173749923706, \"eval_ref\": \"mmlu_prox_es_math__ab611024__s000__tb65536__v000\", \"oracle_delta\": 0.16992759... | {
"index": "mmlu_prox_es_math__ab611024__s000__tb65536__v000",
"uid": "mmlu_prox_es_math__ab611024__s000__tb65536__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_es_other__4e99656e__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2961.8973782568896, \"nll_base\": 7.9935853481292725, \"eval_ref\": \"mmlu_prox_es_other__4e99656e__s000__tb262144__v000\", \"oracle_delta\": 2.... | {
"index": "mmlu_prox_es_other__4e99656e__s000__tb262144__v000",
"uid": "mmlu_prox_es_other__4e99656e__s000__tb262144__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_es_philosophy__d5dda7b9__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2431.577840877198, \"nll_base\": 7.796295642852783, \"eval_ref\": \"mmlu_prox_es_philosophy__d5dda7b9__s000__tb131072__v000\", \"oracle_del... | {
"index": "mmlu_prox_es_philosophy__d5dda7b9__s000__tb131072__v000",
"uid": "mmlu_prox_es_philosophy__d5dda7b9__s000__tb131072__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_es_physics__9313e43a__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2846.325061084189, \"nll_base\": 7.953783988952637, \"eval_ref\": \"mmlu_prox_es_physics__9313e43a__s000__tb131072__v000\", \"oracle_delta\": ... | {
"index": "mmlu_prox_es_physics__9313e43a__s000__tb131072__v000",
"uid": "mmlu_prox_es_physics__9313e43a__s000__tb131072__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_es_psychology__c87979ac__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2403.608958860481, \"nll_base\": 7.784726619720459, \"eval_ref\": \"mmlu_prox_es_psychology__c87979ac__s000__tb65536__v000\", \"oracle_delta\... | {
"index": "mmlu_prox_es_psychology__c87979ac__s000__tb65536__v000",
"uid": "mmlu_prox_es_psychology__c87979ac__s000__tb65536__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_fr_biology__521acfcf__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 7616.267406570949, \"nll_base\": 8.938041687011719, \"eval_ref\": \"mmlu_prox_fr_biology__521acfcf__s000__tb262144__v000\", \"oracle_delta\": ... | {
"index": "mmlu_prox_fr_biology__521acfcf__s000__tb262144__v000",
"uid": "mmlu_prox_fr_biology__521acfcf__s000__tb262144__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_fr_business__62e1be3e__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 5177.602139035392, \"nll_base\": 8.55209732055664, \"eval_ref\": \"mmlu_prox_fr_business__62e1be3e__s000__tb262144__v000\", \"oracle_delta\":... | {
"index": "mmlu_prox_fr_business__62e1be3e__s000__tb262144__v000",
"uid": "mmlu_prox_fr_business__62e1be3e__s000__tb262144__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_fr_chemistry__319f279c__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3816.6887266795934, \"nll_base\": 8.247138500213623, \"eval_ref\": \"mmlu_prox_fr_chemistry__319f279c__s000__tb131072__v000\", \"oracle_delt... | {
"index": "mmlu_prox_fr_chemistry__319f279c__s000__tb131072__v000",
"uid": "mmlu_prox_fr_chemistry__319f279c__s000__tb131072__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_fr_computer_science__aa50b607__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 6406.69954317277, \"nll_base\": 8.76509952545166, \"eval_ref\": \"mmlu_prox_fr_computer_science__aa50b607__s000__tb131072__v000\", \"... | {
"index": "mmlu_prox_fr_computer_science__aa50b607__s000__tb131072__v000",
"uid": "mmlu_prox_fr_computer_science__aa50b607__s000__tb131072__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_fr_economics__67954c3e__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 5504.904698743157, \"nll_base\": 8.613394737243652, \"eval_ref\": \"mmlu_prox_fr_economics__67954c3e__s000__tb131072__v000\", \"oracle_delta... | {
"index": "mmlu_prox_fr_economics__67954c3e__s000__tb131072__v000",
"uid": "mmlu_prox_fr_economics__67954c3e__s000__tb131072__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_fr_health__74ff237e__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 6222.209669071918, \"nll_base\": 8.735880374908447, \"eval_ref\": \"mmlu_prox_fr_health__74ff237e__s000__tb65536__v000\", \"oracle_delta\": 0.196... | {
"index": "mmlu_prox_fr_health__74ff237e__s000__tb65536__v000",
"uid": "mmlu_prox_fr_health__74ff237e__s000__tb65536__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_fr_history__45ee3627__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4374.530411943998, \"nll_base\": 8.383554458618164, \"eval_ref\": \"mmlu_prox_fr_history__45ee3627__s000__tb131072__v000\", \"oracle_delta\": ... | {
"index": "mmlu_prox_fr_history__45ee3627__s000__tb131072__v000",
"uid": "mmlu_prox_fr_history__45ee3627__s000__tb131072__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_fr_law__30a395a4__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 4294.024757209067, \"nll_base\": 8.36497974395752, \"eval_ref\": \"mmlu_prox_fr_law__30a395a4__s000__tb262144__v000\", \"oracle_delta\": 2.0811901... | {
"index": "mmlu_prox_fr_law__30a395a4__s000__tb262144__v000",
"uid": "mmlu_prox_fr_law__30a395a4__s000__tb262144__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_fr_physics__8b8de36f__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 5397.948854824503, \"nll_base\": 8.593774318695068, \"eval_ref\": \"mmlu_prox_fr_physics__8b8de36f__s000__tb65536__v000\", \"oracle_delta\": -0.... | {
"index": "mmlu_prox_fr_physics__8b8de36f__s000__tb65536__v000",
"uid": "mmlu_prox_fr_physics__8b8de36f__s000__tb65536__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_hi_biology__3f8eb166__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3566.650946392786, \"nll_base\": 8.17938232421875, \"eval_ref\": \"mmlu_prox_hi_biology__3f8eb166__s000__tb131072__v000\", \"oracle_delta\": 0... | {
"index": "mmlu_prox_hi_biology__3f8eb166__s000__tb131072__v000",
"uid": "mmlu_prox_hi_biology__3f8eb166__s000__tb131072__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_hi_business__790fb926__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 6737.074438529424, \"nll_base\": 8.815381050109863, \"eval_ref\": \"mmlu_prox_hi_business__790fb926__s000__tb65536__v000\", \"oracle_delta\": 0... | {
"index": "mmlu_prox_hi_business__790fb926__s000__tb65536__v000",
"uid": "mmlu_prox_hi_business__790fb926__s000__tb65536__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_hi_health__3d33148d__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 5170.75061577885, \"nll_base\": 8.55077314376831, \"eval_ref\": \"mmlu_prox_hi_health__3d33148d__s000__tb131072__v000\", \"oracle_delta\": 0.52... | {
"index": "mmlu_prox_hi_health__3d33148d__s000__tb131072__v000",
"uid": "mmlu_prox_hi_health__3d33148d__s000__tb131072__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_hi_history__99629add__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2049.691659152145, \"nll_base\": 7.625444650650024, \"eval_ref\": \"mmlu_prox_hi_history__99629add__s000__tb65536__v000\", \"oracle_delta\": 0.2... | {
"index": "mmlu_prox_hi_history__99629add__s000__tb65536__v000",
"uid": "mmlu_prox_hi_history__99629add__s000__tb65536__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_hi_other__fd72a297__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 6449.698065262637, \"nll_base\": 8.771788597106934, \"eval_ref\": \"mmlu_prox_hi_other__fd72a297__s000__tb131072__v000\", \"oracle_delta\": 0.50... | {
"index": "mmlu_prox_hi_other__fd72a297__s000__tb131072__v000",
"uid": "mmlu_prox_hi_other__fd72a297__s000__tb131072__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_hu_business__018dbffb__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 5380.443603685957, \"nll_base\": 8.590526103973389, \"eval_ref\": \"mmlu_prox_hu_business__018dbffb__s000__tb65536__v000\", \"oracle_delta\": 0... | {
"index": "mmlu_prox_hu_business__018dbffb__s000__tb65536__v000",
"uid": "mmlu_prox_hu_business__018dbffb__s000__tb65536__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_hu_computer_science__fa1ff052__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2938.264104926442, \"nll_base\": 7.985574245452881, \"eval_ref\": \"mmlu_prox_hu_computer_science__fa1ff052__s000__tb131072__v000\", ... | {
"index": "mmlu_prox_hu_computer_science__fa1ff052__s000__tb131072__v000",
"uid": "mmlu_prox_hu_computer_science__fa1ff052__s000__tb131072__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_hu_economics__abe7cf0b__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 4040.221261247666, \"nll_base\": 8.304054737091064, \"eval_ref\": \"mmlu_prox_hu_economics__abe7cf0b__s000__tb262144__v000\", \"oracle_delta... | {
"index": "mmlu_prox_hu_economics__abe7cf0b__s000__tb262144__v000",
"uid": "mmlu_prox_hu_economics__abe7cf0b__s000__tb262144__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_hu_engineering__c3608f64__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4101.347331055003, \"nll_base\": 8.319070816040039, \"eval_ref\": \"mmlu_prox_hu_engineering__c3608f64__s000__tb65536__v000\", \"oracle_delt... | {
"index": "mmlu_prox_hu_engineering__c3608f64__s000__tb65536__v000",
"uid": "mmlu_prox_hu_engineering__c3608f64__s000__tb65536__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_hu_health__45bf6cbf__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3472.3071585257762, \"nll_base\": 8.15257453918457, \"eval_ref\": \"mmlu_prox_hu_health__45bf6cbf__s000__tb131072__v000\", \"oracle_delta\": 0.... | {
"index": "mmlu_prox_hu_health__45bf6cbf__s000__tb131072__v000",
"uid": "mmlu_prox_hu_health__45bf6cbf__s000__tb131072__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_hu_math__65332ace__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 5305.050578458278, \"nll_base\": 8.576414585113525, \"eval_ref\": \"mmlu_prox_hu_math__65332ace__s000__tb131072__v000\", \"oracle_delta\": 0.9172... | {
"index": "mmlu_prox_hu_math__65332ace__s000__tb131072__v000",
"uid": "mmlu_prox_hu_math__65332ace__s000__tb131072__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_hu_other__9b2fdc8f__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3225.3903521192697, \"nll_base\": 8.078809261322021, \"eval_ref\": \"mmlu_prox_hu_other__9b2fdc8f__s000__tb131072__v000\", \"oracle_delta\": 0.8... | {
"index": "mmlu_prox_hu_other__9b2fdc8f__s000__tb131072__v000",
"uid": "mmlu_prox_hu_other__9b2fdc8f__s000__tb131072__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_hu_philosophy__6b88e4bb__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3634.072320869116, \"nll_base\": 8.198109149932861, \"eval_ref\": \"mmlu_prox_hu_philosophy__6b88e4bb__s000__tb131072__v000\", \"oracle_del... | {
"index": "mmlu_prox_hu_philosophy__6b88e4bb__s000__tb131072__v000",
"uid": "mmlu_prox_hu_philosophy__6b88e4bb__s000__tb131072__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_id_biology__56f96fc5__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 7611.540375555674, \"nll_base\": 8.937420845031738, \"eval_ref\": \"mmlu_prox_id_biology__56f96fc5__s000__tb131072__v000\", \"oracle_delta\": ... | {
"index": "mmlu_prox_id_biology__56f96fc5__s000__tb131072__v000",
"uid": "mmlu_prox_id_biology__56f96fc5__s000__tb131072__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_id_business__7796c69e__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 8397.880683984751, \"nll_base\": 9.0357346534729, \"eval_ref\": \"mmlu_prox_id_business__7796c69e__s000__tb262144__v000\", \"oracle_delta\": ... | {
"index": "mmlu_prox_id_business__7796c69e__s000__tb262144__v000",
"uid": "mmlu_prox_id_business__7796c69e__s000__tb262144__v000",
"axis": "instruction",
"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.