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\": \"afrimgsm_translate_sot_prompt_3__58b0b777__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 196.89348656406892, \"nll_base\": 5.282662905179537, \"eval_ref\": \"afrimgsm_translate_sot_prompt_3__58b0b777__s000__tb131072__v00... | {
"index": "afrimgsm_translate_sot_prompt_3__58b0b777__s000__tb131072__v001",
"uid": "afrimgsm_translate_sot_prompt_3__58b0b777__s000__tb131072__v001",
"axis": "math",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_translate_sot_prompt_3__58b0b777__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 196.89348656406892, \"nll_base\": 5.282662905179537, \"eval_ref\": \"afrimgsm_translate_sot_prompt_3__58b0b777__s000__tb65536__v000\"... | {
"index": "afrimgsm_translate_sot_prompt_3__58b0b777__s000__tb65536__v000",
"uid": "afrimgsm_translate_sot_prompt_3__58b0b777__s000__tb65536__v000",
"axis": "math",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_translate_sot_prompt_4__472d5be6__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 134.84320233764123, \"nll_base\": 4.904112639250578, \"eval_ref\": \"afrimgsm_translate_sot_prompt_4__472d5be6__s000__tb262144__v00... | {
"index": "afrimgsm_translate_sot_prompt_4__472d5be6__s000__tb262144__v000",
"uid": "afrimgsm_translate_sot_prompt_4__472d5be6__s000__tb262144__v000",
"axis": "math",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_translate_sot_prompt_4__472d5be6__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 134.84320233764123, \"nll_base\": 4.904112639250578, \"eval_ref\": \"afrimgsm_translate_sot_prompt_4__472d5be6__s000__tb65536__v001\"... | {
"index": "afrimgsm_translate_sot_prompt_4__472d5be6__s000__tb65536__v001",
"uid": "afrimgsm_translate_sot_prompt_4__472d5be6__s000__tb65536__v001",
"axis": "math",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_translate_swa_prompt_1__fd9df364__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 173.78839975376042, \"nll_base\": 5.157838465803761, \"eval_ref\": \"afrimgsm_translate_swa_prompt_1__fd9df364__s000__tb131072__v00... | {
"index": "afrimgsm_translate_swa_prompt_1__fd9df364__s000__tb131072__v000",
"uid": "afrimgsm_translate_swa_prompt_1__fd9df364__s000__tb131072__v000",
"axis": "math",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_translate_swa_prompt_2__9a3d7e15__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 183.23943246372744, \"nll_base\": 5.210793671784578, \"eval_ref\": \"afrimgsm_translate_swa_prompt_2__9a3d7e15__s000__tb131072__v00... | {
"index": "afrimgsm_translate_swa_prompt_2__9a3d7e15__s000__tb131072__v000",
"uid": "afrimgsm_translate_swa_prompt_2__9a3d7e15__s000__tb131072__v000",
"axis": "math",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_translate_swa_prompt_3__6f7b804a__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 166.18589685823002, \"nll_base\": 5.1131070223721595, \"eval_ref\": \"afrimgsm_translate_swa_prompt_3__6f7b804a__s000__tb131072__v0... | {
"index": "afrimgsm_translate_swa_prompt_3__6f7b804a__s000__tb131072__v001",
"uid": "afrimgsm_translate_swa_prompt_3__6f7b804a__s000__tb131072__v001",
"axis": "math",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_translate_swa_prompt_3__6f7b804a__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 166.18589685823002, \"nll_base\": 5.1131070223721595, \"eval_ref\": \"afrimgsm_translate_swa_prompt_3__6f7b804a__s000__tb65536__v000\... | {
"index": "afrimgsm_translate_swa_prompt_3__6f7b804a__s000__tb65536__v000",
"uid": "afrimgsm_translate_swa_prompt_3__6f7b804a__s000__tb65536__v000",
"axis": "math",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_translate_swa_prompt_4__baf1f52f__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 120.58709167683145, \"nll_base\": 4.7923722443757235, \"eval_ref\": \"afrimgsm_translate_swa_prompt_4__baf1f52f__s000__tb131072__v0... | {
"index": "afrimgsm_translate_swa_prompt_4__baf1f52f__s000__tb131072__v000",
"uid": "afrimgsm_translate_swa_prompt_4__baf1f52f__s000__tb131072__v000",
"axis": "math",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_translate_swa_prompt_5__b3772601__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 196.50730812877575, \"nll_base\": 5.280699622112771, \"eval_ref\": \"afrimgsm_translate_swa_prompt_5__b3772601__s000__tb131072__v00... | {
"index": "afrimgsm_translate_swa_prompt_5__b3772601__s000__tb131072__v001",
"uid": "afrimgsm_translate_swa_prompt_5__b3772601__s000__tb131072__v001",
"axis": "math",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_translate_swa_prompt_5__b3772601__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 196.50730812877575, \"nll_base\": 5.280699622112771, \"eval_ref\": \"afrimgsm_translate_swa_prompt_5__b3772601__s000__tb65536__v000\"... | {
"index": "afrimgsm_translate_swa_prompt_5__b3772601__s000__tb65536__v000",
"uid": "afrimgsm_translate_swa_prompt_5__b3772601__s000__tb65536__v000",
"axis": "math",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_translate_twi_prompt_1__7207a4ed__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 322.0455774350214, \"nll_base\": 5.774693080357143, \"eval_ref\": \"afrimgsm_translate_twi_prompt_1__7207a4ed__s000__tb65536__v000\",... | {
"index": "afrimgsm_translate_twi_prompt_1__7207a4ed__s000__tb65536__v000",
"uid": "afrimgsm_translate_twi_prompt_1__7207a4ed__s000__tb65536__v000",
"axis": "math",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_translate_twi_prompt_2__3a6b1f22__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 332.96951082197245, \"nll_base\": 5.808050926695478, \"eval_ref\": \"afrimgsm_translate_twi_prompt_2__3a6b1f22__s000__tb65536__v001\"... | {
"index": "afrimgsm_translate_twi_prompt_2__3a6b1f22__s000__tb65536__v001",
"uid": "afrimgsm_translate_twi_prompt_2__3a6b1f22__s000__tb65536__v001",
"axis": "math",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_translate_twi_prompt_3__8202f6c3__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 253.63675477221923, \"nll_base\": 5.535903144128544, \"eval_ref\": \"afrimgsm_translate_twi_prompt_3__8202f6c3__s000__tb131072__v00... | {
"index": "afrimgsm_translate_twi_prompt_3__8202f6c3__s000__tb131072__v000",
"uid": "afrimgsm_translate_twi_prompt_3__8202f6c3__s000__tb131072__v000",
"axis": "math",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_translate_twi_prompt_4__11eca6bc__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 151.91890026669924, \"nll_base\": 5.0233468275803785, \"eval_ref\": \"afrimgsm_translate_twi_prompt_4__11eca6bc__s000__tb131072__v0... | {
"index": "afrimgsm_translate_twi_prompt_4__11eca6bc__s000__tb131072__v000",
"uid": "afrimgsm_translate_twi_prompt_4__11eca6bc__s000__tb131072__v000",
"axis": "math",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_translate_twi_prompt_4__11eca6bc__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 151.91890026669924, \"nll_base\": 5.0233468275803785, \"eval_ref\": \"afrimgsm_translate_twi_prompt_4__11eca6bc__s000__tb262144__v0... | {
"index": "afrimgsm_translate_twi_prompt_4__11eca6bc__s000__tb262144__v001",
"uid": "afrimgsm_translate_twi_prompt_4__11eca6bc__s000__tb262144__v001",
"axis": "math",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_translate_twi_prompt_5__a044806c__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 282.4165055497017, \"nll_base\": 5.643382951325061, \"eval_ref\": \"afrimgsm_translate_twi_prompt_5__a044806c__s000__tb131072__v000... | {
"index": "afrimgsm_translate_twi_prompt_5__a044806c__s000__tb131072__v000",
"uid": "afrimgsm_translate_twi_prompt_5__a044806c__s000__tb131072__v000",
"axis": "math",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_translate_wol_prompt_1__cec4ae08__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 233.85614011513363, \"nll_base\": 5.454706140466638, \"eval_ref\": \"afrimgsm_translate_wol_prompt_1__cec4ae08__s000__tb131072__v00... | {
"index": "afrimgsm_translate_wol_prompt_1__cec4ae08__s000__tb131072__v000",
"uid": "afrimgsm_translate_wol_prompt_1__cec4ae08__s000__tb131072__v000",
"axis": "math",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_translate_wol_prompt_1__cec4ae08__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 233.85614011513363, \"nll_base\": 5.454706140466638, \"eval_ref\": \"afrimgsm_translate_wol_prompt_1__cec4ae08__s000__tb262144__v00... | {
"index": "afrimgsm_translate_wol_prompt_1__cec4ae08__s000__tb262144__v001",
"uid": "afrimgsm_translate_wol_prompt_1__cec4ae08__s000__tb262144__v001",
"axis": "math",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_translate_wol_prompt_2__c3ee2eea__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 176.12232237306424, \"nll_base\": 5.171178767111449, \"eval_ref\": \"afrimgsm_translate_wol_prompt_2__c3ee2eea__s000__tb131072__v00... | {
"index": "afrimgsm_translate_wol_prompt_2__c3ee2eea__s000__tb131072__v000",
"uid": "afrimgsm_translate_wol_prompt_2__c3ee2eea__s000__tb131072__v000",
"axis": "math",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_translate_wol_prompt_2__c3ee2eea__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 176.12232237306424, \"nll_base\": 5.171178767111449, \"eval_ref\": \"afrimgsm_translate_wol_prompt_2__c3ee2eea__s000__tb262144__v00... | {
"index": "afrimgsm_translate_wol_prompt_2__c3ee2eea__s000__tb262144__v001",
"uid": "afrimgsm_translate_wol_prompt_2__c3ee2eea__s000__tb262144__v001",
"axis": "math",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_translate_wol_prompt_3__49f752d4__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 183.63612599922538, \"nll_base\": 5.2129562235324185, \"eval_ref\": \"afrimgsm_translate_wol_prompt_3__49f752d4__s000__tb65536__v001\... | {
"index": "afrimgsm_translate_wol_prompt_3__49f752d4__s000__tb65536__v001",
"uid": "afrimgsm_translate_wol_prompt_3__49f752d4__s000__tb65536__v001",
"axis": "math",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_translate_wol_prompt_4__eed9c80b__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 178.79525055557212, \"nll_base\": 5.186241299498315, \"eval_ref\": \"afrimgsm_translate_wol_prompt_4__eed9c80b__s000__tb65536__v001\"... | {
"index": "afrimgsm_translate_wol_prompt_4__eed9c80b__s000__tb65536__v001",
"uid": "afrimgsm_translate_wol_prompt_4__eed9c80b__s000__tb65536__v001",
"axis": "math",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_translate_wol_prompt_5__8100687c__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 217.67022301402753, \"nll_base\": 5.382981178977273, \"eval_ref\": \"afrimgsm_translate_wol_prompt_5__8100687c__s000__tb131072__v00... | {
"index": "afrimgsm_translate_wol_prompt_5__8100687c__s000__tb131072__v001",
"uid": "afrimgsm_translate_wol_prompt_5__8100687c__s000__tb131072__v001",
"axis": "math",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_translate_wol_prompt_5__8100687c__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 217.67022301402753, \"nll_base\": 5.382981178977273, \"eval_ref\": \"afrimgsm_translate_wol_prompt_5__8100687c__s000__tb65536__v000\"... | {
"index": "afrimgsm_translate_wol_prompt_5__8100687c__s000__tb65536__v000",
"uid": "afrimgsm_translate_wol_prompt_5__8100687c__s000__tb65536__v000",
"axis": "math",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_translate_xho_prompt_1__5e891bd6__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 289.7238776400721, \"nll_base\": 5.668928323332796, \"eval_ref\": \"afrimgsm_translate_xho_prompt_1__5e891bd6__s000__tb65536__v001\",... | {
"index": "afrimgsm_translate_xho_prompt_1__5e891bd6__s000__tb65536__v001",
"uid": "afrimgsm_translate_xho_prompt_1__5e891bd6__s000__tb65536__v001",
"axis": "math",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_translate_xho_prompt_2__8b465b9c__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 219.0589873674623, \"nll_base\": 5.389341042258523, \"eval_ref\": \"afrimgsm_translate_xho_prompt_2__8b465b9c__s000__tb131072__v000... | {
"index": "afrimgsm_translate_xho_prompt_2__8b465b9c__s000__tb131072__v000",
"uid": "afrimgsm_translate_xho_prompt_2__8b465b9c__s000__tb131072__v000",
"axis": "math",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_translate_xho_prompt_2__8b465b9c__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 219.0589873674623, \"nll_base\": 5.389341042258523, \"eval_ref\": \"afrimgsm_translate_xho_prompt_2__8b465b9c__s000__tb262144__v001... | {
"index": "afrimgsm_translate_xho_prompt_2__8b465b9c__s000__tb262144__v001",
"uid": "afrimgsm_translate_xho_prompt_2__8b465b9c__s000__tb262144__v001",
"axis": "math",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_translate_xho_prompt_3__62088321__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 159.71600546686454, \"nll_base\": 5.073397272283381, \"eval_ref\": \"afrimgsm_translate_xho_prompt_3__62088321__s000__tb131072__v00... | {
"index": "afrimgsm_translate_xho_prompt_3__62088321__s000__tb131072__v000",
"uid": "afrimgsm_translate_xho_prompt_3__62088321__s000__tb131072__v000",
"axis": "math",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_translate_xho_prompt_3__62088321__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 159.71600546686454, \"nll_base\": 5.073397272283381, \"eval_ref\": \"afrimgsm_translate_xho_prompt_3__62088321__s000__tb262144__v00... | {
"index": "afrimgsm_translate_xho_prompt_3__62088321__s000__tb262144__v001",
"uid": "afrimgsm_translate_xho_prompt_3__62088321__s000__tb262144__v001",
"axis": "math",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_translate_xho_prompt_4__57d84075__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 141.92675206852303, \"nll_base\": 4.955311094011579, \"eval_ref\": \"afrimgsm_translate_xho_prompt_4__57d84075__s000__tb131072__v00... | {
"index": "afrimgsm_translate_xho_prompt_4__57d84075__s000__tb131072__v001",
"uid": "afrimgsm_translate_xho_prompt_4__57d84075__s000__tb131072__v001",
"axis": "math",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_translate_xho_prompt_4__57d84075__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 141.92675206852303, \"nll_base\": 4.955311094011579, \"eval_ref\": \"afrimgsm_translate_xho_prompt_4__57d84075__s000__tb65536__v000\"... | {
"index": "afrimgsm_translate_xho_prompt_4__57d84075__s000__tb65536__v000",
"uid": "afrimgsm_translate_xho_prompt_4__57d84075__s000__tb65536__v000",
"axis": "math",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_translate_xho_prompt_5__b4de81ab__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 216.96633725677444, \"nll_base\": 5.3797422136579245, \"eval_ref\": \"afrimgsm_translate_xho_prompt_5__b4de81ab__s000__tb262144__v0... | {
"index": "afrimgsm_translate_xho_prompt_5__b4de81ab__s000__tb262144__v000",
"uid": "afrimgsm_translate_xho_prompt_5__b4de81ab__s000__tb262144__v000",
"axis": "math",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_translate_xho_prompt_5__b4de81ab__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 216.96633725677444, \"nll_base\": 5.3797422136579245, \"eval_ref\": \"afrimgsm_translate_xho_prompt_5__b4de81ab__s000__tb65536__v001\... | {
"index": "afrimgsm_translate_xho_prompt_5__b4de81ab__s000__tb65536__v001",
"uid": "afrimgsm_translate_xho_prompt_5__b4de81ab__s000__tb65536__v001",
"axis": "math",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_translate_yor_prompt_1__05596105__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 237.39959012830587, \"nll_base\": 5.469744755671575, \"eval_ref\": \"afrimgsm_translate_yor_prompt_1__05596105__s000__tb65536__v001\"... | {
"index": "afrimgsm_translate_yor_prompt_1__05596105__s000__tb65536__v001",
"uid": "afrimgsm_translate_yor_prompt_1__05596105__s000__tb65536__v001",
"axis": "math",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_translate_yor_prompt_2__db052ba4__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 155.70662579450155, \"nll_base\": 5.0479736328125, \"eval_ref\": \"afrimgsm_translate_yor_prompt_2__db052ba4__s000__tb131072__v000\... | {
"index": "afrimgsm_translate_yor_prompt_2__db052ba4__s000__tb131072__v000",
"uid": "afrimgsm_translate_yor_prompt_2__db052ba4__s000__tb131072__v000",
"axis": "math",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_translate_yor_prompt_2__db052ba4__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 155.70662579450155, \"nll_base\": 5.0479736328125, \"eval_ref\": \"afrimgsm_translate_yor_prompt_2__db052ba4__s000__tb262144__v001\... | {
"index": "afrimgsm_translate_yor_prompt_2__db052ba4__s000__tb262144__v001",
"uid": "afrimgsm_translate_yor_prompt_2__db052ba4__s000__tb262144__v001",
"axis": "math",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_translate_yor_prompt_3__18391233__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 205.34593990558312, \"nll_base\": 5.324696068624848, \"eval_ref\": \"afrimgsm_translate_yor_prompt_3__18391233__s000__tb262144__v00... | {
"index": "afrimgsm_translate_yor_prompt_3__18391233__s000__tb262144__v000",
"uid": "afrimgsm_translate_yor_prompt_3__18391233__s000__tb262144__v000",
"axis": "math",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_translate_yor_prompt_3__18391233__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 205.34593990558312, \"nll_base\": 5.324696068624848, \"eval_ref\": \"afrimgsm_translate_yor_prompt_3__18391233__s000__tb65536__v001\"... | {
"index": "afrimgsm_translate_yor_prompt_3__18391233__s000__tb65536__v001",
"uid": "afrimgsm_translate_yor_prompt_3__18391233__s000__tb65536__v001",
"axis": "math",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_translate_yor_prompt_4__01baec11__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 175.60058606478034, \"nll_base\": 5.168212018694196, \"eval_ref\": \"afrimgsm_translate_yor_prompt_4__01baec11__s000__tb65536__v001\"... | {
"index": "afrimgsm_translate_yor_prompt_4__01baec11__s000__tb65536__v001",
"uid": "afrimgsm_translate_yor_prompt_4__01baec11__s000__tb65536__v001",
"axis": "math",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_translate_yor_prompt_5__fe9fe78a__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 201.02557832274272, \"nll_base\": 5.303432155299831, \"eval_ref\": \"afrimgsm_translate_yor_prompt_5__fe9fe78a__s000__tb262144__v00... | {
"index": "afrimgsm_translate_yor_prompt_5__fe9fe78a__s000__tb262144__v000",
"uid": "afrimgsm_translate_yor_prompt_5__fe9fe78a__s000__tb262144__v000",
"axis": "math",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_translate_yor_prompt_5__fe9fe78a__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 201.02557832274272, \"nll_base\": 5.303432155299831, \"eval_ref\": \"afrimgsm_translate_yor_prompt_5__fe9fe78a__s000__tb65536__v001\"... | {
"index": "afrimgsm_translate_yor_prompt_5__fe9fe78a__s000__tb65536__v001",
"uid": "afrimgsm_translate_yor_prompt_5__fe9fe78a__s000__tb65536__v001",
"axis": "math",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_translate_zul_prompt_1__59a13fce__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 219.30167398176172, \"nll_base\": 5.390448288690476, \"eval_ref\": \"afrimgsm_translate_zul_prompt_1__59a13fce__s000__tb131072__v00... | {
"index": "afrimgsm_translate_zul_prompt_1__59a13fce__s000__tb131072__v001",
"uid": "afrimgsm_translate_zul_prompt_1__59a13fce__s000__tb131072__v001",
"axis": "math",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_translate_zul_prompt_1__59a13fce__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 219.30167398176172, \"nll_base\": 5.390448288690476, \"eval_ref\": \"afrimgsm_translate_zul_prompt_1__59a13fce__s000__tb65536__v000\"... | {
"index": "afrimgsm_translate_zul_prompt_1__59a13fce__s000__tb65536__v000",
"uid": "afrimgsm_translate_zul_prompt_1__59a13fce__s000__tb65536__v000",
"axis": "math",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_translate_zul_prompt_2__186eb183__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 189.52461827158314, \"nll_base\": 5.244518927808078, \"eval_ref\": \"afrimgsm_translate_zul_prompt_2__186eb183__s000__tb131072__v00... | {
"index": "afrimgsm_translate_zul_prompt_2__186eb183__s000__tb131072__v001",
"uid": "afrimgsm_translate_zul_prompt_2__186eb183__s000__tb131072__v001",
"axis": "math",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_translate_zul_prompt_2__186eb183__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 189.52461827158314, \"nll_base\": 5.244518927808078, \"eval_ref\": \"afrimgsm_translate_zul_prompt_2__186eb183__s000__tb65536__v000\"... | {
"index": "afrimgsm_translate_zul_prompt_2__186eb183__s000__tb65536__v000",
"uid": "afrimgsm_translate_zul_prompt_2__186eb183__s000__tb65536__v000",
"axis": "math",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_translate_zul_prompt_3__9299e79f__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 159.3675713917053, \"nll_base\": 5.071213303950795, \"eval_ref\": \"afrimgsm_translate_zul_prompt_3__9299e79f__s000__tb65536__v001\",... | {
"index": "afrimgsm_translate_zul_prompt_3__9299e79f__s000__tb65536__v001",
"uid": "afrimgsm_translate_zul_prompt_3__9299e79f__s000__tb65536__v001",
"axis": "math",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_translate_zul_prompt_4__1c1b942a__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 126.87517579247724, \"nll_base\": 4.843203735351563, \"eval_ref\": \"afrimgsm_translate_zul_prompt_4__1c1b942a__s000__tb131072__v00... | {
"index": "afrimgsm_translate_zul_prompt_4__1c1b942a__s000__tb131072__v001",
"uid": "afrimgsm_translate_zul_prompt_4__1c1b942a__s000__tb131072__v001",
"axis": "math",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_translate_zul_prompt_4__1c1b942a__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 126.87517579247724, \"nll_base\": 4.843203735351563, \"eval_ref\": \"afrimgsm_translate_zul_prompt_4__1c1b942a__s000__tb65536__v000\"... | {
"index": "afrimgsm_translate_zul_prompt_4__1c1b942a__s000__tb65536__v000",
"uid": "afrimgsm_translate_zul_prompt_4__1c1b942a__s000__tb65536__v000",
"axis": "math",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_translate_zul_prompt_5__7d67ca90__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 200.04684062763152, \"nll_base\": 5.298551542264921, \"eval_ref\": \"afrimgsm_translate_zul_prompt_5__7d67ca90__s000__tb131072__v00... | {
"index": "afrimgsm_translate_zul_prompt_5__7d67ca90__s000__tb131072__v000",
"uid": "afrimgsm_translate_zul_prompt_5__7d67ca90__s000__tb131072__v000",
"axis": "math",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_translate_zul_prompt_5__7d67ca90__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 200.04684062763152, \"nll_base\": 5.298551542264921, \"eval_ref\": \"afrimgsm_translate_zul_prompt_5__7d67ca90__s000__tb262144__v00... | {
"index": "afrimgsm_translate_zul_prompt_5__7d67ca90__s000__tb262144__v001",
"uid": "afrimgsm_translate_zul_prompt_5__7d67ca90__s000__tb262144__v001",
"axis": "math",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_twi_prompt_1__d3cd5d28__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 362.3928233530035, \"nll_base\": 5.892728770862926, \"eval_ref\": \"afrimgsm_twi_prompt_1__d3cd5d28__s000__tb262144__v000\", \"oracle_delta\"... | {
"index": "afrimgsm_twi_prompt_1__d3cd5d28__s000__tb262144__v000",
"uid": "afrimgsm_twi_prompt_1__d3cd5d28__s000__tb262144__v000",
"axis": "math",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_twi_prompt_1__d3cd5d28__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 362.3928233530035, \"nll_base\": 5.892728770862926, \"eval_ref\": \"afrimgsm_twi_prompt_1__d3cd5d28__s000__tb65536__v001\", \"oracle_delta\": 0... | {
"index": "afrimgsm_twi_prompt_1__d3cd5d28__s000__tb65536__v001",
"uid": "afrimgsm_twi_prompt_1__d3cd5d28__s000__tb65536__v001",
"axis": "math",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_twi_prompt_2__54c0b7f7__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 422.4998576515588, \"nll_base\": 6.046189109877784, \"eval_ref\": \"afrimgsm_twi_prompt_2__54c0b7f7__s000__tb131072__v001\", \"oracle_delta\"... | {
"index": "afrimgsm_twi_prompt_2__54c0b7f7__s000__tb131072__v001",
"uid": "afrimgsm_twi_prompt_2__54c0b7f7__s000__tb131072__v001",
"axis": "math",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_twi_prompt_2__54c0b7f7__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 422.4998576515588, \"nll_base\": 6.046189109877784, \"eval_ref\": \"afrimgsm_twi_prompt_2__54c0b7f7__s000__tb65536__v000\", \"oracle_delta\": 0... | {
"index": "afrimgsm_twi_prompt_2__54c0b7f7__s000__tb65536__v000",
"uid": "afrimgsm_twi_prompt_2__54c0b7f7__s000__tb65536__v000",
"axis": "math",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_wol_prompt_1__c51ebd65__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 326.49617713211615, \"nll_base\": 5.788418240017361, \"eval_ref\": \"afrimgsm_wol_prompt_1__c51ebd65__s000__tb131072__v000\", \"oracle_delta\... | {
"index": "afrimgsm_wol_prompt_1__c51ebd65__s000__tb131072__v000",
"uid": "afrimgsm_wol_prompt_1__c51ebd65__s000__tb131072__v000",
"axis": "math",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_wol_prompt_1__c51ebd65__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 326.49617713211615, \"nll_base\": 5.788418240017361, \"eval_ref\": \"afrimgsm_wol_prompt_1__c51ebd65__s000__tb262144__v001\", \"oracle_delta\... | {
"index": "afrimgsm_wol_prompt_1__c51ebd65__s000__tb262144__v001",
"uid": "afrimgsm_wol_prompt_1__c51ebd65__s000__tb262144__v001",
"axis": "math",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_wol_prompt_4__55249ddb__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 214.60237811182222, \"nll_base\": 5.3687869117373515, \"eval_ref\": \"afrimgsm_wol_prompt_4__55249ddb__s000__tb131072__v000\", \"oracle_delta... | {
"index": "afrimgsm_wol_prompt_4__55249ddb__s000__tb131072__v000",
"uid": "afrimgsm_wol_prompt_4__55249ddb__s000__tb131072__v000",
"axis": "math",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_wol_prompt_4__55249ddb__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 214.60237811182222, \"nll_base\": 5.3687869117373515, \"eval_ref\": \"afrimgsm_wol_prompt_4__55249ddb__s000__tb262144__v001\", \"oracle_delta... | {
"index": "afrimgsm_wol_prompt_4__55249ddb__s000__tb262144__v001",
"uid": "afrimgsm_wol_prompt_4__55249ddb__s000__tb262144__v001",
"axis": "math",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_wol_prompt_5__d184d469__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 379.9483796211272, \"nll_base\": 5.940035400390625, \"eval_ref\": \"afrimgsm_wol_prompt_5__d184d469__s000__tb131072__v000\", \"oracle_delta\"... | {
"index": "afrimgsm_wol_prompt_5__d184d469__s000__tb131072__v000",
"uid": "afrimgsm_wol_prompt_5__d184d469__s000__tb131072__v000",
"axis": "math",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_wol_prompt_5__d184d469__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 379.9483796211272, \"nll_base\": 5.940035400390625, \"eval_ref\": \"afrimgsm_wol_prompt_5__d184d469__s000__tb262144__v001\", \"oracle_delta\"... | {
"index": "afrimgsm_wol_prompt_5__d184d469__s000__tb262144__v001",
"uid": "afrimgsm_wol_prompt_5__d184d469__s000__tb262144__v001",
"axis": "math",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_yor_prompt_1__cbdef550__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 479.3448364086515, \"nll_base\": 6.172420247395833, \"eval_ref\": \"afrimgsm_yor_prompt_1__cbdef550__s000__tb131072__v000\", \"oracle_delta\"... | {
"index": "afrimgsm_yor_prompt_1__cbdef550__s000__tb131072__v000",
"uid": "afrimgsm_yor_prompt_1__cbdef550__s000__tb131072__v000",
"axis": "math",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_yor_prompt_1__cbdef550__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 479.3448364086515, \"nll_base\": 6.172420247395833, \"eval_ref\": \"afrimgsm_yor_prompt_1__cbdef550__s000__tb262144__v001\", \"oracle_delta\"... | {
"index": "afrimgsm_yor_prompt_1__cbdef550__s000__tb262144__v001",
"uid": "afrimgsm_yor_prompt_1__cbdef550__s000__tb262144__v001",
"axis": "math",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_yor_prompt_2__462cd6be__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 407.91552387046744, \"nll_base\": 6.011060103629399, \"eval_ref\": \"afrimgsm_yor_prompt_2__462cd6be__s000__tb131072__v000\", \"oracle_delta\... | {
"index": "afrimgsm_yor_prompt_2__462cd6be__s000__tb131072__v000",
"uid": "afrimgsm_yor_prompt_2__462cd6be__s000__tb131072__v000",
"axis": "math",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_yor_prompt_3__8042a13c__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 223.13560785330955, \"nll_base\": 5.407779693603516, \"eval_ref\": \"afrimgsm_yor_prompt_3__8042a13c__s000__tb131072__v000\", \"oracle_delta\... | {
"index": "afrimgsm_yor_prompt_3__8042a13c__s000__tb131072__v000",
"uid": "afrimgsm_yor_prompt_3__8042a13c__s000__tb131072__v000",
"axis": "math",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_yor_prompt_3__8042a13c__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 223.13560785330955, \"nll_base\": 5.407779693603516, \"eval_ref\": \"afrimgsm_yor_prompt_3__8042a13c__s000__tb262144__v001\", \"oracle_delta\... | {
"index": "afrimgsm_yor_prompt_3__8042a13c__s000__tb262144__v001",
"uid": "afrimgsm_yor_prompt_3__8042a13c__s000__tb262144__v001",
"axis": "math",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_yor_prompt_4__a93c623e__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 153.47321205425368, \"nll_base\": 5.033526037493322, \"eval_ref\": \"afrimgsm_yor_prompt_4__a93c623e__s000__tb131072__v000\", \"oracle_delta\... | {
"index": "afrimgsm_yor_prompt_4__a93c623e__s000__tb131072__v000",
"uid": "afrimgsm_yor_prompt_4__a93c623e__s000__tb131072__v000",
"axis": "math",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_yor_prompt_5__54e0749e__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 544.0738998856418, \"nll_base\": 6.299085083007813, \"eval_ref\": \"afrimgsm_yor_prompt_5__54e0749e__s000__tb131072__v000\", \"oracle_delta\"... | {
"index": "afrimgsm_yor_prompt_5__54e0749e__s000__tb131072__v000",
"uid": "afrimgsm_yor_prompt_5__54e0749e__s000__tb131072__v000",
"axis": "math",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_yor_prompt_5__54e0749e__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 544.0738998856418, \"nll_base\": 6.299085083007813, \"eval_ref\": \"afrimgsm_yor_prompt_5__54e0749e__s000__tb262144__v001\", \"oracle_delta\"... | {
"index": "afrimgsm_yor_prompt_5__54e0749e__s000__tb262144__v001",
"uid": "afrimgsm_yor_prompt_5__54e0749e__s000__tb262144__v001",
"axis": "math",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afriqa_bem_prompt_4__11b0fb87__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 13.181836847714754, \"nll_base\": 2.5788398856547343, \"eval_ref\": \"afriqa_bem_prompt_4__11b0fb87__s000__tb65536__v000\", \"oracle_delta\": 0.1... | {
"index": "afriqa_bem_prompt_4__11b0fb87__s000__tb65536__v000",
"uid": "afriqa_bem_prompt_4__11b0fb87__s000__tb65536__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afriqa_fon_prompt_3__9e5f6aac__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 10.96011552458987, \"nll_base\": 2.3942628220290874, \"eval_ref\": \"afriqa_fon_prompt_3__9e5f6aac__s000__tb262144__v000\", \"oracle_delta\": 0... | {
"index": "afriqa_fon_prompt_3__9e5f6aac__s000__tb262144__v000",
"uid": "afriqa_fon_prompt_3__9e5f6aac__s000__tb262144__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afriqa_swa_prompt_4__e7d155d3__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 28.214685579707528, \"nll_base\": 3.3398426076307652, \"eval_ref\": \"afriqa_swa_prompt_4__e7d155d3__s000__tb65536__v000\", \"oracle_delta\": 0.2... | {
"index": "afriqa_swa_prompt_4__e7d155d3__s000__tb65536__v000",
"uid": "afriqa_swa_prompt_4__e7d155d3__s000__tb65536__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrisenti_pcm_prompt_1__5ffb8ab6__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3480.124344201818, \"nll_base\": 8.154823303222656, \"eval_ref\": \"afrisenti_pcm_prompt_1__5ffb8ab6__s000__tb262144__v000\", \"oracle_delta... | {
"index": "afrisenti_pcm_prompt_1__5ffb8ab6__s000__tb262144__v000",
"uid": "afrisenti_pcm_prompt_1__5ffb8ab6__s000__tb262144__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrisenti_pcm_prompt_4__007c181f__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 12345.250666736312, \"nll_base\": 9.421026706695557, \"eval_ref\": \"afrisenti_pcm_prompt_4__007c181f__s000__tb131072__v000\", \"oracle_delt... | {
"index": "afrisenti_pcm_prompt_4__007c181f__s000__tb131072__v000",
"uid": "afrisenti_pcm_prompt_4__007c181f__s000__tb131072__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrisenti_pcm_prompt_5__32877717__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4364.5106256018735, \"nll_base\": 8.381261348724365, \"eval_ref\": \"afrisenti_pcm_prompt_5__32877717__s000__tb131072__v000\", \"oracle_delt... | {
"index": "afrisenti_pcm_prompt_5__32877717__s000__tb131072__v000",
"uid": "afrisenti_pcm_prompt_5__32877717__s000__tb131072__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrisenti_swa_prompt_2__c3322db6__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3296.961762047336, \"nll_base\": 8.100756645202637, \"eval_ref\": \"afrisenti_swa_prompt_2__c3322db6__s000__tb65536__v000\", \"oracle_delta\":... | {
"index": "afrisenti_swa_prompt_2__c3322db6__s000__tb65536__v000",
"uid": "afrisenti_swa_prompt_2__c3322db6__s000__tb65536__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrisenti_tir_prompt_3__e7a3b7eb__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 5556.6980903549465, \"nll_base\": 8.622759342193604, \"eval_ref\": \"afrisenti_tir_prompt_3__e7a3b7eb__s000__tb65536__v000\", \"oracle_delta\"... | {
"index": "afrisenti_tir_prompt_3__e7a3b7eb__s000__tb65536__v000",
"uid": "afrisenti_tir_prompt_3__e7a3b7eb__s000__tb65536__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrisenti_tir_prompt_4__ba472423__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 16084.99682737101, \"nll_base\": 9.68564224243164, \"eval_ref\": \"afrisenti_tir_prompt_4__ba472423__s000__tb65536__v000\", \"oracle_delta\": ... | {
"index": "afrisenti_tir_prompt_4__ba472423__s000__tb65536__v000",
"uid": "afrisenti_tir_prompt_4__ba472423__s000__tb65536__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrixnli_en_direct_ibo__2727c73c__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3275.6455047664203, \"nll_base\": 8.0942702293396, \"eval_ref\": \"afrixnli_en_direct_ibo__2727c73c__s000__tb65536__v000\", \"oracle_delta\": ... | {
"index": "afrixnli_en_direct_ibo__2727c73c__s000__tb65536__v000",
"uid": "afrixnli_en_direct_ibo__2727c73c__s000__tb65536__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrixnli_eng_prompt_2__6d1ba70f__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 942.2736097658632, \"nll_base\": 6.84829568862915, \"eval_ref\": \"afrixnli_eng_prompt_2__6d1ba70f__s000__tb65536__v000\", \"oracle_delta\": 0.... | {
"index": "afrixnli_eng_prompt_2__6d1ba70f__s000__tb65536__v000",
"uid": "afrixnli_eng_prompt_2__6d1ba70f__s000__tb65536__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrixnli_ewe_prompt_1__aa7f011e__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 332.3761138747743, \"nll_base\": 5.8062672008167615, \"eval_ref\": \"afrixnli_ewe_prompt_1__aa7f011e__s000__tb65536__v000\", \"oracle_delta\": ... | {
"index": "afrixnli_ewe_prompt_1__aa7f011e__s000__tb65536__v000",
"uid": "afrixnli_ewe_prompt_1__aa7f011e__s000__tb65536__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrixnli_lug_prompt_3__cff8630b__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 573.7203291184062, \"nll_base\": 6.352142046082695, \"eval_ref\": \"afrixnli_lug_prompt_3__cff8630b__s000__tb65536__v000\", \"oracle_delta\": 0... | {
"index": "afrixnli_lug_prompt_3__cff8630b__s000__tb65536__v000",
"uid": "afrixnli_lug_prompt_3__cff8630b__s000__tb65536__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrixnli_manual_translate_swa__dcbfe31e__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 413.1201177202598, \"nll_base\": 6.02373839261239, \"eval_ref\": \"afrixnli_manual_translate_swa__dcbfe31e__s000__tb262144__v000\", \... | {
"index": "afrixnli_manual_translate_swa__dcbfe31e__s000__tb262144__v000",
"uid": "afrixnli_manual_translate_swa__dcbfe31e__s000__tb262144__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrixnli_manual_translate_yor__5ef7c4da__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 575.0705636391249, \"nll_base\": 6.354492752640335, \"eval_ref\": \"afrixnli_manual_translate_yor__5ef7c4da__s000__tb131072__v000\", ... | {
"index": "afrixnli_manual_translate_yor__5ef7c4da__s000__tb131072__v000",
"uid": "afrixnli_manual_translate_yor__5ef7c4da__s000__tb131072__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrixnli_swa_prompt_2__9915ef89__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3255.870241690236, \"nll_base\": 8.088214874267578, \"eval_ref\": \"afrixnli_swa_prompt_2__9915ef89__s000__tb131072__v000\", \"oracle_delta\"... | {
"index": "afrixnli_swa_prompt_2__9915ef89__s000__tb131072__v000",
"uid": "afrixnli_swa_prompt_2__9915ef89__s000__tb131072__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrixnli_swa_prompt_4__edfd184f__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 244.3789445727566, \"nll_base\": 5.498720072083554, \"eval_ref\": \"afrixnli_swa_prompt_4__edfd184f__s000__tb262144__v000\", \"oracle_delta\"... | {
"index": "afrixnli_swa_prompt_4__edfd184f__s000__tb262144__v000",
"uid": "afrixnli_swa_prompt_4__edfd184f__s000__tb262144__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrixnli_translate_fra_prompt_4__afc4a269__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 641.7488399372336, \"nll_base\": 6.464197012094351, \"eval_ref\": \"afrixnli_translate_fra_prompt_4__afc4a269__s000__tb65536__v000\",... | {
"index": "afrixnli_translate_fra_prompt_4__afc4a269__s000__tb65536__v000",
"uid": "afrixnli_translate_fra_prompt_4__afc4a269__s000__tb65536__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrixnli_translate_ibo__c159403f__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 838.5385913527005, \"nll_base\": 6.731660604476929, \"eval_ref\": \"afrixnli_translate_ibo__c159403f__s000__tb65536__v000\", \"oracle_delta\":... | {
"index": "afrixnli_translate_ibo__c159403f__s000__tb65536__v000",
"uid": "afrixnli_translate_ibo__c159403f__s000__tb65536__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrixnli_translate_lin_prompt_3__67edd97a__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1303.767582740594, \"nll_base\": 7.173013492506378, \"eval_ref\": \"afrixnli_translate_lin_prompt_3__67edd97a__s000__tb65536__v000\",... | {
"index": "afrixnli_translate_lin_prompt_3__67edd97a__s000__tb65536__v000",
"uid": "afrixnli_translate_lin_prompt_3__67edd97a__s000__tb65536__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrixnli_translate_lin_prompt_5__c4477b9f__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 917.9676615835592, \"nll_base\": 6.822162162966844, \"eval_ref\": \"afrixnli_translate_lin_prompt_5__c4477b9f__s000__tb65536__v000\",... | {
"index": "afrixnli_translate_lin_prompt_5__c4477b9f__s000__tb65536__v000",
"uid": "afrixnli_translate_lin_prompt_5__c4477b9f__s000__tb65536__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrixnli_translate_wol__a6c2f1ab__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 706.5377577889392, \"nll_base\": 6.5603766441345215, \"eval_ref\": \"afrixnli_translate_wol__a6c2f1ab__s000__tb131072__v000\", \"oracle_delt... | {
"index": "afrixnli_translate_wol__a6c2f1ab__s000__tb131072__v000",
"uid": "afrixnli_translate_wol__a6c2f1ab__s000__tb131072__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrixnli_translate_yor_prompt_1__dba58b13__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 508.1721664684617, \"nll_base\": 6.230820300532322, \"eval_ref\": \"afrixnli_translate_yor_prompt_1__dba58b13__s000__tb65536__v000\",... | {
"index": "afrixnli_translate_yor_prompt_1__dba58b13__s000__tb65536__v000",
"uid": "afrixnli_translate_yor_prompt_1__dba58b13__s000__tb65536__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrixnli_translate_zul__57704379__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 605.3941514762113, \"nll_base\": 6.405879735946655, \"eval_ref\": \"afrixnli_translate_zul__57704379__s000__tb65536__v000\", \"oracle_delta\":... | {
"index": "afrixnli_translate_zul__57704379__s000__tb65536__v000",
"uid": "afrixnli_translate_zul__57704379__s000__tb65536__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrixnli_translate_zul_prompt_5__5b12e827__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 838.0630160329724, \"nll_base\": 6.731093295784884, \"eval_ref\": \"afrixnli_translate_zul_prompt_5__5b12e827__s000__tb262144__v000... | {
"index": "afrixnli_translate_zul_prompt_5__5b12e827__s000__tb262144__v000",
"uid": "afrixnli_translate_zul_prompt_5__5b12e827__s000__tb262144__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrixnli_yor_prompt_3__5bb235dd__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 706.917689946593, \"nll_base\": 6.560914237544222, \"eval_ref\": \"afrixnli_yor_prompt_3__5bb235dd__s000__tb262144__v000\", \"oracle_delta\":... | {
"index": "afrixnli_yor_prompt_3__5bb235dd__s000__tb262144__v000",
"uid": "afrixnli_yor_prompt_3__5bb235dd__s000__tb262144__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrixnli_zul_prompt_3__807acaa2__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 466.6590008704022, \"nll_base\": 6.14559880009404, \"eval_ref\": \"afrixnli_zul_prompt_3__807acaa2__s000__tb262144__v000\", \"oracle_delta\":... | {
"index": "afrixnli_zul_prompt_3__807acaa2__s000__tb262144__v000",
"uid": "afrixnli_zul_prompt_3__807acaa2__s000__tb262144__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"agieval_aqua_rat__58dbb277__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 27.126856324985162, \"nll_base\": 3.3005242454631447, \"eval_ref\": \"agieval_aqua_rat__58dbb277__s000__tb65536__v000\", \"oracle_delta\": 0.3571510... | {
"index": "agieval_aqua_rat__58dbb277__s000__tb65536__v000",
"uid": "agieval_aqua_rat__58dbb277__s000__tb65536__v000",
"axis": "knowledge",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"agieval_gaokao_mathcloze__ee254fe6__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 19.055645316867874, \"nll_base\": 2.9473633997648667, \"eval_ref\": \"agieval_gaokao_mathcloze__ee254fe6__s000__tb131072__v000\", \"oracle... | {
"index": "agieval_gaokao_mathcloze__ee254fe6__s000__tb131072__v000",
"uid": "agieval_gaokao_mathcloze__ee254fe6__s000__tb131072__v000",
"axis": "knowledge",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"agieval_logiqa_en__08e0b500__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2.4472800907326535, \"nll_base\": 0.894977240816839, \"eval_ref\": \"agieval_logiqa_en__08e0b500__s000__tb131072__v000\", \"oracle_delta\": 0.088... | {
"index": "agieval_logiqa_en__08e0b500__s000__tb131072__v000",
"uid": "agieval_logiqa_en__08e0b500__s000__tb131072__v000",
"axis": "knowledge",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"agieval_logiqa_en__08e0b500__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2.4472800907326535, \"nll_base\": 0.894977240816839, \"eval_ref\": \"agieval_logiqa_en__08e0b500__s000__tb262144__v001\", \"oracle_delta\": 0.174... | {
"index": "agieval_logiqa_en__08e0b500__s000__tb262144__v001",
"uid": "agieval_logiqa_en__08e0b500__s000__tb262144__v001",
"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.