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_cot_translate_wol_prompt_5__df4ace54__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 221.6577181612342, \"nll_base\": 5.401134381975447, \"eval_ref\": \"afrimgsm_cot_translate_wol_prompt_5__df4ace54__s000__tb65536_... | {
"index": "afrimgsm_cot_translate_wol_prompt_5__df4ace54__s000__tb65536__v000",
"uid": "afrimgsm_cot_translate_wol_prompt_5__df4ace54__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_cot_translate_xho_prompt_1__4aea2c88__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 197.92761139662778, \"nll_base\": 5.287901364839994, \"eval_ref\": \"afrimgsm_cot_translate_xho_prompt_1__4aea2c88__s000__tb65536... | {
"index": "afrimgsm_cot_translate_xho_prompt_1__4aea2c88__s000__tb65536__v001",
"uid": "afrimgsm_cot_translate_xho_prompt_1__4aea2c88__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_cot_translate_xho_prompt_2__8a790135__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 223.41225060889656, \"nll_base\": 5.409018722234988, \"eval_ref\": \"afrimgsm_cot_translate_xho_prompt_2__8a790135__s000__tb131... | {
"index": "afrimgsm_cot_translate_xho_prompt_2__8a790135__s000__tb131072__v000",
"uid": "afrimgsm_cot_translate_xho_prompt_2__8a790135__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_cot_translate_xho_prompt_3__bf02f8d6__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 208.07894527312325, \"nll_base\": 5.337917552274816, \"eval_ref\": \"afrimgsm_cot_translate_xho_prompt_3__bf02f8d6__s000__tb65536... | {
"index": "afrimgsm_cot_translate_xho_prompt_3__bf02f8d6__s000__tb65536__v000",
"uid": "afrimgsm_cot_translate_xho_prompt_3__bf02f8d6__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_cot_translate_xho_prompt_4__26b3b18f__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 207.66941511251105, \"nll_base\": 5.3359474649234695, \"eval_ref\": \"afrimgsm_cot_translate_xho_prompt_4__26b3b18f__s000__tb26... | {
"index": "afrimgsm_cot_translate_xho_prompt_4__26b3b18f__s000__tb262144__v000",
"uid": "afrimgsm_cot_translate_xho_prompt_4__26b3b18f__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_cot_translate_xho_prompt_4__26b3b18f__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 207.66941511251105, \"nll_base\": 5.3359474649234695, \"eval_ref\": \"afrimgsm_cot_translate_xho_prompt_4__26b3b18f__s000__tb6553... | {
"index": "afrimgsm_cot_translate_xho_prompt_4__26b3b18f__s000__tb65536__v001",
"uid": "afrimgsm_cot_translate_xho_prompt_4__26b3b18f__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_cot_translate_xho_prompt_5__d3cd6965__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 221.7742262726738, \"nll_base\": 5.4016598656063985, \"eval_ref\": \"afrimgsm_cot_translate_xho_prompt_5__d3cd6965__s000__tb131... | {
"index": "afrimgsm_cot_translate_xho_prompt_5__d3cd6965__s000__tb131072__v000",
"uid": "afrimgsm_cot_translate_xho_prompt_5__d3cd6965__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_cot_translate_xho_prompt_5__d3cd6965__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 221.7742262726738, \"nll_base\": 5.4016598656063985, \"eval_ref\": \"afrimgsm_cot_translate_xho_prompt_5__d3cd6965__s000__tb262... | {
"index": "afrimgsm_cot_translate_xho_prompt_5__d3cd6965__s000__tb262144__v001",
"uid": "afrimgsm_cot_translate_xho_prompt_5__d3cd6965__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_cot_translate_yor_prompt_1__ce2e8a04__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 210.39099511402992, \"nll_base\": 5.348967681059966, \"eval_ref\": \"afrimgsm_cot_translate_yor_prompt_1__ce2e8a04__s000__tb131... | {
"index": "afrimgsm_cot_translate_yor_prompt_1__ce2e8a04__s000__tb131072__v001",
"uid": "afrimgsm_cot_translate_yor_prompt_1__ce2e8a04__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_cot_translate_yor_prompt_2__d8802988__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 187.52849359337853, \"nll_base\": 5.23393079969618, \"eval_ref\": \"afrimgsm_cot_translate_yor_prompt_2__d8802988__s000__tb1310... | {
"index": "afrimgsm_cot_translate_yor_prompt_2__d8802988__s000__tb131072__v001",
"uid": "afrimgsm_cot_translate_yor_prompt_2__d8802988__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_cot_translate_yor_prompt_2__d8802988__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 187.52849359337853, \"nll_base\": 5.23393079969618, \"eval_ref\": \"afrimgsm_cot_translate_yor_prompt_2__d8802988__s000__tb65536_... | {
"index": "afrimgsm_cot_translate_yor_prompt_2__d8802988__s000__tb65536__v000",
"uid": "afrimgsm_cot_translate_yor_prompt_2__d8802988__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_cot_translate_yor_prompt_3__8037e67b__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 183.8381715723156, \"nll_base\": 5.214055868295523, \"eval_ref\": \"afrimgsm_cot_translate_yor_prompt_3__8037e67b__s000__tb65536_... | {
"index": "afrimgsm_cot_translate_yor_prompt_3__8037e67b__s000__tb65536__v001",
"uid": "afrimgsm_cot_translate_yor_prompt_3__8037e67b__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_cot_translate_yor_prompt_4__f9c75370__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 220.74530757495168, \"nll_base\": 5.397009582519531, \"eval_ref\": \"afrimgsm_cot_translate_yor_prompt_4__f9c75370__s000__tb131... | {
"index": "afrimgsm_cot_translate_yor_prompt_4__f9c75370__s000__tb131072__v000",
"uid": "afrimgsm_cot_translate_yor_prompt_4__f9c75370__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_cot_translate_yor_prompt_4__f9c75370__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 220.74530757495168, \"nll_base\": 5.397009582519531, \"eval_ref\": \"afrimgsm_cot_translate_yor_prompt_4__f9c75370__s000__tb262... | {
"index": "afrimgsm_cot_translate_yor_prompt_4__f9c75370__s000__tb262144__v001",
"uid": "afrimgsm_cot_translate_yor_prompt_4__f9c75370__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_cot_translate_yor_prompt_5__cdfe639b__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 219.2132243779525, \"nll_base\": 5.390044883445457, \"eval_ref\": \"afrimgsm_cot_translate_yor_prompt_5__cdfe639b__s000__tb2621... | {
"index": "afrimgsm_cot_translate_yor_prompt_5__cdfe639b__s000__tb262144__v000",
"uid": "afrimgsm_cot_translate_yor_prompt_5__cdfe639b__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_cot_translate_yor_prompt_5__cdfe639b__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 219.2132243779525, \"nll_base\": 5.390044883445457, \"eval_ref\": \"afrimgsm_cot_translate_yor_prompt_5__cdfe639b__s000__tb65536_... | {
"index": "afrimgsm_cot_translate_yor_prompt_5__cdfe639b__s000__tb65536__v001",
"uid": "afrimgsm_cot_translate_yor_prompt_5__cdfe639b__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_cot_translate_zul_prompt_1__8513cb05__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 207.48659823899263, \"nll_base\": 5.3350667507849, \"eval_ref\": \"afrimgsm_cot_translate_zul_prompt_1__8513cb05__s000__tb65536__... | {
"index": "afrimgsm_cot_translate_zul_prompt_1__8513cb05__s000__tb65536__v001",
"uid": "afrimgsm_cot_translate_zul_prompt_1__8513cb05__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_cot_translate_zul_prompt_2__451b48e3__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 150.21342918624288, \"nll_base\": 5.012057144030005, \"eval_ref\": \"afrimgsm_cot_translate_zul_prompt_2__451b48e3__s000__tb65536... | {
"index": "afrimgsm_cot_translate_zul_prompt_2__451b48e3__s000__tb65536__v001",
"uid": "afrimgsm_cot_translate_zul_prompt_2__451b48e3__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_cot_translate_zul_prompt_3__964a6193__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 167.01423855971393, \"nll_base\": 5.118079069618867, \"eval_ref\": \"afrimgsm_cot_translate_zul_prompt_3__964a6193__s000__tb131... | {
"index": "afrimgsm_cot_translate_zul_prompt_3__964a6193__s000__tb131072__v000",
"uid": "afrimgsm_cot_translate_zul_prompt_3__964a6193__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_cot_translate_zul_prompt_4__62350e7b__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 250.86087156478777, \"nll_base\": 5.524898488898026, \"eval_ref\": \"afrimgsm_cot_translate_zul_prompt_4__62350e7b__s000__tb262... | {
"index": "afrimgsm_cot_translate_zul_prompt_4__62350e7b__s000__tb262144__v000",
"uid": "afrimgsm_cot_translate_zul_prompt_4__62350e7b__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_cot_translate_zul_prompt_4__62350e7b__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 250.86087156478777, \"nll_base\": 5.524898488898026, \"eval_ref\": \"afrimgsm_cot_translate_zul_prompt_4__62350e7b__s000__tb65536... | {
"index": "afrimgsm_cot_translate_zul_prompt_4__62350e7b__s000__tb65536__v001",
"uid": "afrimgsm_cot_translate_zul_prompt_4__62350e7b__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_cot_translate_zul_prompt_5__ce0caf6e__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 241.23646220812208, \"nll_base\": 5.485777623445085, \"eval_ref\": \"afrimgsm_cot_translate_zul_prompt_5__ce0caf6e__s000__tb131... | {
"index": "afrimgsm_cot_translate_zul_prompt_5__ce0caf6e__s000__tb131072__v000",
"uid": "afrimgsm_cot_translate_zul_prompt_5__ce0caf6e__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_cot_twi_prompt_1__28cb33cd__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 384.23450553292616, \"nll_base\": 5.951253057683556, \"eval_ref\": \"afrimgsm_cot_twi_prompt_1__28cb33cd__s000__tb131072__v000\", \"oracl... | {
"index": "afrimgsm_cot_twi_prompt_1__28cb33cd__s000__tb131072__v000",
"uid": "afrimgsm_cot_twi_prompt_1__28cb33cd__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_cot_twi_prompt_1__28cb33cd__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 384.23450553292616, \"nll_base\": 5.951253057683556, \"eval_ref\": \"afrimgsm_cot_twi_prompt_1__28cb33cd__s000__tb262144__v001\", \"oracl... | {
"index": "afrimgsm_cot_twi_prompt_1__28cb33cd__s000__tb262144__v001",
"uid": "afrimgsm_cot_twi_prompt_1__28cb33cd__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_cot_twi_prompt_4__d3e08cb4__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 148.33294312919296, \"nll_base\": 4.99945936290496, \"eval_ref\": \"afrimgsm_cot_twi_prompt_4__d3e08cb4__s000__tb131072__v000\", \"oracle... | {
"index": "afrimgsm_cot_twi_prompt_4__d3e08cb4__s000__tb131072__v000",
"uid": "afrimgsm_cot_twi_prompt_4__d3e08cb4__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_cot_twi_prompt_4__d3e08cb4__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 148.33294312919296, \"nll_base\": 4.99945936290496, \"eval_ref\": \"afrimgsm_cot_twi_prompt_4__d3e08cb4__s000__tb262144__v001\", \"oracle... | {
"index": "afrimgsm_cot_twi_prompt_4__d3e08cb4__s000__tb262144__v001",
"uid": "afrimgsm_cot_twi_prompt_4__d3e08cb4__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_cot_wol_prompt_1__829cafcb__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 305.7749389919895, \"nll_base\": 5.722849337854118, \"eval_ref\": \"afrimgsm_cot_wol_prompt_1__829cafcb__s000__tb131072__v000\", \"oracle... | {
"index": "afrimgsm_cot_wol_prompt_1__829cafcb__s000__tb131072__v000",
"uid": "afrimgsm_cot_wol_prompt_1__829cafcb__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_cot_wol_prompt_1__829cafcb__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 305.7749389919895, \"nll_base\": 5.722849337854118, \"eval_ref\": \"afrimgsm_cot_wol_prompt_1__829cafcb__s000__tb262144__v001\", \"oracle... | {
"index": "afrimgsm_cot_wol_prompt_1__829cafcb__s000__tb262144__v001",
"uid": "afrimgsm_cot_wol_prompt_1__829cafcb__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_cot_wol_prompt_2__cb379ab9__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 198.76811340952412, \"nll_base\": 5.292138886033443, \"eval_ref\": \"afrimgsm_cot_wol_prompt_2__cb379ab9__s000__tb65536__v000\", \"oracle_d... | {
"index": "afrimgsm_cot_wol_prompt_2__cb379ab9__s000__tb65536__v000",
"uid": "afrimgsm_cot_wol_prompt_2__cb379ab9__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_cot_wol_prompt_3__4600ddf2__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 178.62459799102888, \"nll_base\": 5.185286385672433, \"eval_ref\": \"afrimgsm_cot_wol_prompt_3__4600ddf2__s000__tb262144__v000\", \"oracl... | {
"index": "afrimgsm_cot_wol_prompt_3__4600ddf2__s000__tb262144__v000",
"uid": "afrimgsm_cot_wol_prompt_3__4600ddf2__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_cot_wol_prompt_3__4600ddf2__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 178.62459799102888, \"nll_base\": 5.185286385672433, \"eval_ref\": \"afrimgsm_cot_wol_prompt_3__4600ddf2__s000__tb65536__v001\", \"oracle_d... | {
"index": "afrimgsm_cot_wol_prompt_3__4600ddf2__s000__tb65536__v001",
"uid": "afrimgsm_cot_wol_prompt_3__4600ddf2__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_cot_wol_prompt_4__cc456f0b__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 144.01555014335742, \"nll_base\": 4.9699212808524615, \"eval_ref\": \"afrimgsm_cot_wol_prompt_4__cc456f0b__s000__tb262144__v001\", \"orac... | {
"index": "afrimgsm_cot_wol_prompt_4__cc456f0b__s000__tb262144__v001",
"uid": "afrimgsm_cot_wol_prompt_4__cc456f0b__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_cot_wol_prompt_5__17259cfa__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 259.62952317851176, \"nll_base\": 5.559255704012784, \"eval_ref\": \"afrimgsm_cot_wol_prompt_5__17259cfa__s000__tb131072__v001\", \"oracl... | {
"index": "afrimgsm_cot_wol_prompt_5__17259cfa__s000__tb131072__v001",
"uid": "afrimgsm_cot_wol_prompt_5__17259cfa__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_cot_wol_prompt_5__17259cfa__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 259.62952317851176, \"nll_base\": 5.559255704012784, \"eval_ref\": \"afrimgsm_cot_wol_prompt_5__17259cfa__s000__tb65536__v000\", \"oracle_d... | {
"index": "afrimgsm_cot_wol_prompt_5__17259cfa__s000__tb65536__v000",
"uid": "afrimgsm_cot_wol_prompt_5__17259cfa__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_cot_yor_prompt_1__c3805c30__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 332.57421756411156, \"nll_base\": 5.806863045907235, \"eval_ref\": \"afrimgsm_cot_yor_prompt_1__c3805c30__s000__tb262144__v000\", \"oracl... | {
"index": "afrimgsm_cot_yor_prompt_1__c3805c30__s000__tb262144__v000",
"uid": "afrimgsm_cot_yor_prompt_1__c3805c30__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_cot_yor_prompt_1__c3805c30__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 332.57421756411156, \"nll_base\": 5.806863045907235, \"eval_ref\": \"afrimgsm_cot_yor_prompt_1__c3805c30__s000__tb65536__v001\", \"oracle_d... | {
"index": "afrimgsm_cot_yor_prompt_1__c3805c30__s000__tb65536__v001",
"uid": "afrimgsm_cot_yor_prompt_1__c3805c30__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_cot_yor_prompt_2__b289abb5__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 297.37169438383444, \"nll_base\": 5.694982852575914, \"eval_ref\": \"afrimgsm_cot_yor_prompt_2__b289abb5__s000__tb262144__v000\", \"oracl... | {
"index": "afrimgsm_cot_yor_prompt_2__b289abb5__s000__tb262144__v000",
"uid": "afrimgsm_cot_yor_prompt_2__b289abb5__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_cot_yor_prompt_2__b289abb5__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 297.37169438383444, \"nll_base\": 5.694982852575914, \"eval_ref\": \"afrimgsm_cot_yor_prompt_2__b289abb5__s000__tb65536__v001\", \"oracle_d... | {
"index": "afrimgsm_cot_yor_prompt_2__b289abb5__s000__tb65536__v001",
"uid": "afrimgsm_cot_yor_prompt_2__b289abb5__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_cot_yor_prompt_3__a7c33120__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 185.58656856902488, \"nll_base\": 5.223521450109649, \"eval_ref\": \"afrimgsm_cot_yor_prompt_3__a7c33120__s000__tb65536__v000\", \"oracle_d... | {
"index": "afrimgsm_cot_yor_prompt_3__a7c33120__s000__tb65536__v000",
"uid": "afrimgsm_cot_yor_prompt_3__a7c33120__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_cot_yor_prompt_4__251b06bf__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 147.06339217747092, \"nll_base\": 4.990863733124315, \"eval_ref\": \"afrimgsm_cot_yor_prompt_4__251b06bf__s000__tb262144__v000\", \"oracl... | {
"index": "afrimgsm_cot_yor_prompt_4__251b06bf__s000__tb262144__v000",
"uid": "afrimgsm_cot_yor_prompt_4__251b06bf__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_cot_yor_prompt_4__251b06bf__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 147.06339217747092, \"nll_base\": 4.990863733124315, \"eval_ref\": \"afrimgsm_cot_yor_prompt_4__251b06bf__s000__tb65536__v001\", \"oracle_d... | {
"index": "afrimgsm_cot_yor_prompt_4__251b06bf__s000__tb65536__v001",
"uid": "afrimgsm_cot_yor_prompt_4__251b06bf__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_cot_yor_prompt_5__f9cdd33d__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 463.81872431797643, \"nll_base\": 6.1394937955416164, \"eval_ref\": \"afrimgsm_cot_yor_prompt_5__f9cdd33d__s000__tb131072__v001\", \"orac... | {
"index": "afrimgsm_cot_yor_prompt_5__f9cdd33d__s000__tb131072__v001",
"uid": "afrimgsm_cot_yor_prompt_5__f9cdd33d__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_cot_yor_prompt_5__f9cdd33d__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 463.81872431797643, \"nll_base\": 6.1394937955416164, \"eval_ref\": \"afrimgsm_cot_yor_prompt_5__f9cdd33d__s000__tb65536__v000\", \"oracle_... | {
"index": "afrimgsm_cot_yor_prompt_5__f9cdd33d__s000__tb65536__v000",
"uid": "afrimgsm_cot_yor_prompt_5__f9cdd33d__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_eng_prompt_1__5ef1188e__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 202.07551585266214, \"nll_base\": 5.308641468394886, \"eval_ref\": \"afrimgsm_eng_prompt_1__5ef1188e__s000__tb65536__v000\", \"oracle_delta\": ... | {
"index": "afrimgsm_eng_prompt_1__5ef1188e__s000__tb65536__v000",
"uid": "afrimgsm_eng_prompt_1__5ef1188e__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_eng_prompt_3__7dbc7a6f__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 186.60799409046018, \"nll_base\": 5.229010128280492, \"eval_ref\": \"afrimgsm_eng_prompt_3__7dbc7a6f__s000__tb131072__v000\", \"oracle_delta\... | {
"index": "afrimgsm_eng_prompt_3__7dbc7a6f__s000__tb131072__v000",
"uid": "afrimgsm_eng_prompt_3__7dbc7a6f__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_eng_prompt_3__7dbc7a6f__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 186.60799409046018, \"nll_base\": 5.229010128280492, \"eval_ref\": \"afrimgsm_eng_prompt_3__7dbc7a6f__s000__tb262144__v001\", \"oracle_delta\... | {
"index": "afrimgsm_eng_prompt_3__7dbc7a6f__s000__tb262144__v001",
"uid": "afrimgsm_eng_prompt_3__7dbc7a6f__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_eng_prompt_4__1dd8cb4b__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 115.6788462496173, \"nll_base\": 4.750817784723246, \"eval_ref\": \"afrimgsm_eng_prompt_4__1dd8cb4b__s000__tb131072__v000\", \"oracle_delta\"... | {
"index": "afrimgsm_eng_prompt_4__1dd8cb4b__s000__tb131072__v000",
"uid": "afrimgsm_eng_prompt_4__1dd8cb4b__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_eng_prompt_4__1dd8cb4b__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 115.6788462496173, \"nll_base\": 4.750817784723246, \"eval_ref\": \"afrimgsm_eng_prompt_4__1dd8cb4b__s000__tb262144__v001\", \"oracle_delta\"... | {
"index": "afrimgsm_eng_prompt_4__1dd8cb4b__s000__tb262144__v001",
"uid": "afrimgsm_eng_prompt_4__1dd8cb4b__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_eng_prompt_5__acbcc7ca__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 377.1378723225865, \"nll_base\": 5.932610829671224, \"eval_ref\": \"afrimgsm_eng_prompt_5__acbcc7ca__s000__tb262144__v000\", \"oracle_delta\"... | {
"index": "afrimgsm_eng_prompt_5__acbcc7ca__s000__tb262144__v000",
"uid": "afrimgsm_eng_prompt_5__acbcc7ca__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_eng_prompt_5__acbcc7ca__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 377.1378723225865, \"nll_base\": 5.932610829671224, \"eval_ref\": \"afrimgsm_eng_prompt_5__acbcc7ca__s000__tb65536__v001\", \"oracle_delta\": 0... | {
"index": "afrimgsm_eng_prompt_5__acbcc7ca__s000__tb65536__v001",
"uid": "afrimgsm_eng_prompt_5__acbcc7ca__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_ewe_prompt_1__89615904__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 647.4279461628407, \"nll_base\": 6.473007507324219, \"eval_ref\": \"afrimgsm_ewe_prompt_1__89615904__s000__tb131072__v000\", \"oracle_delta\"... | {
"index": "afrimgsm_ewe_prompt_1__89615904__s000__tb131072__v000",
"uid": "afrimgsm_ewe_prompt_1__89615904__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_ewe_prompt_1__89615904__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 647.4279461628407, \"nll_base\": 6.473007507324219, \"eval_ref\": \"afrimgsm_ewe_prompt_1__89615904__s000__tb262144__v001\", \"oracle_delta\"... | {
"index": "afrimgsm_ewe_prompt_1__89615904__s000__tb262144__v001",
"uid": "afrimgsm_ewe_prompt_1__89615904__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_ewe_prompt_2__d9be9c6b__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 317.2192408022492, \"nll_base\": 5.759593146187918, \"eval_ref\": \"afrimgsm_ewe_prompt_2__d9be9c6b__s000__tb65536__v001\", \"oracle_delta\": 0... | {
"index": "afrimgsm_ewe_prompt_2__d9be9c6b__s000__tb65536__v001",
"uid": "afrimgsm_ewe_prompt_2__d9be9c6b__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_ewe_prompt_3__898513d5__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 233.1358105293637, \"nll_base\": 5.45162116156684, \"eval_ref\": \"afrimgsm_ewe_prompt_3__898513d5__s000__tb262144__v000\", \"oracle_delta\":... | {
"index": "afrimgsm_ewe_prompt_3__898513d5__s000__tb262144__v000",
"uid": "afrimgsm_ewe_prompt_3__898513d5__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_ewe_prompt_3__898513d5__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 233.1358105293637, \"nll_base\": 5.45162116156684, \"eval_ref\": \"afrimgsm_ewe_prompt_3__898513d5__s000__tb65536__v001\", \"oracle_delta\": 0.... | {
"index": "afrimgsm_ewe_prompt_3__898513d5__s000__tb65536__v001",
"uid": "afrimgsm_ewe_prompt_3__898513d5__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_ewe_prompt_4__520f7819__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 189.92767538776332, \"nll_base\": 5.2466433438387785, \"eval_ref\": \"afrimgsm_ewe_prompt_4__520f7819__s000__tb65536__v001\", \"oracle_delta\":... | {
"index": "afrimgsm_ewe_prompt_4__520f7819__s000__tb65536__v001",
"uid": "afrimgsm_ewe_prompt_4__520f7819__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_ewe_prompt_5__722b1f3a__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 338.8598744362893, \"nll_base\": 5.825586672182436, \"eval_ref\": \"afrimgsm_ewe_prompt_5__722b1f3a__s000__tb131072__v000\", \"oracle_delta\"... | {
"index": "afrimgsm_ewe_prompt_5__722b1f3a__s000__tb131072__v000",
"uid": "afrimgsm_ewe_prompt_5__722b1f3a__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_ewe_prompt_5__722b1f3a__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 338.8598744362893, \"nll_base\": 5.825586672182436, \"eval_ref\": \"afrimgsm_ewe_prompt_5__722b1f3a__s000__tb262144__v001\", \"oracle_delta\"... | {
"index": "afrimgsm_ewe_prompt_5__722b1f3a__s000__tb262144__v001",
"uid": "afrimgsm_ewe_prompt_5__722b1f3a__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_fra_prompt_1__0eef4007__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 299.65781242390904, \"nll_base\": 5.702641198394495, \"eval_ref\": \"afrimgsm_fra_prompt_1__0eef4007__s000__tb131072__v000\", \"oracle_delta\... | {
"index": "afrimgsm_fra_prompt_1__0eef4007__s000__tb131072__v000",
"uid": "afrimgsm_fra_prompt_1__0eef4007__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_fra_prompt_1__0eef4007__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 299.65781242390904, \"nll_base\": 5.702641198394495, \"eval_ref\": \"afrimgsm_fra_prompt_1__0eef4007__s000__tb262144__v001\", \"oracle_delta\... | {
"index": "afrimgsm_fra_prompt_1__0eef4007__s000__tb262144__v001",
"uid": "afrimgsm_fra_prompt_1__0eef4007__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_fra_prompt_2__6147806b__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 290.34979257940927, \"nll_base\": 5.671086377436572, \"eval_ref\": \"afrimgsm_fra_prompt_2__6147806b__s000__tb131072__v000\", \"oracle_delta\... | {
"index": "afrimgsm_fra_prompt_2__6147806b__s000__tb131072__v000",
"uid": "afrimgsm_fra_prompt_2__6147806b__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_fra_prompt_3__6c0ba55a__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 225.47105353624428, \"nll_base\": 5.418191785010222, \"eval_ref\": \"afrimgsm_fra_prompt_3__6c0ba55a__s000__tb65536__v001\", \"oracle_delta\": ... | {
"index": "afrimgsm_fra_prompt_3__6c0ba55a__s000__tb65536__v001",
"uid": "afrimgsm_fra_prompt_3__6c0ba55a__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_fra_prompt_4__ec7c8f46__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 164.63052471930024, \"nll_base\": 5.103703718919021, \"eval_ref\": \"afrimgsm_fra_prompt_4__ec7c8f46__s000__tb131072__v000\", \"oracle_delta\... | {
"index": "afrimgsm_fra_prompt_4__ec7c8f46__s000__tb131072__v000",
"uid": "afrimgsm_fra_prompt_4__ec7c8f46__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_fra_prompt_5__18340401__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 233.38353715090352, \"nll_base\": 5.452683182450028, \"eval_ref\": \"afrimgsm_fra_prompt_5__18340401__s000__tb65536__v001\", \"oracle_delta\": ... | {
"index": "afrimgsm_fra_prompt_5__18340401__s000__tb65536__v001",
"uid": "afrimgsm_fra_prompt_5__18340401__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_hau_prompt_1__d8975f36__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 277.12023675214533, \"nll_base\": 5.624451479704484, \"eval_ref\": \"afrimgsm_hau_prompt_1__d8975f36__s000__tb262144__v000\", \"oracle_delta\... | {
"index": "afrimgsm_hau_prompt_1__d8975f36__s000__tb262144__v000",
"uid": "afrimgsm_hau_prompt_1__d8975f36__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_hau_prompt_1__d8975f36__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 277.12023675214533, \"nll_base\": 5.624451479704484, \"eval_ref\": \"afrimgsm_hau_prompt_1__d8975f36__s000__tb65536__v001\", \"oracle_delta\": ... | {
"index": "afrimgsm_hau_prompt_1__d8975f36__s000__tb65536__v001",
"uid": "afrimgsm_hau_prompt_1__d8975f36__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_hau_prompt_2__44ea0d3b__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 241.97273189114873, \"nll_base\": 5.488825041672279, \"eval_ref\": \"afrimgsm_hau_prompt_2__44ea0d3b__s000__tb131072__v000\", \"oracle_delta\... | {
"index": "afrimgsm_hau_prompt_2__44ea0d3b__s000__tb131072__v000",
"uid": "afrimgsm_hau_prompt_2__44ea0d3b__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_hau_prompt_2__44ea0d3b__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 241.97273189114873, \"nll_base\": 5.488825041672279, \"eval_ref\": \"afrimgsm_hau_prompt_2__44ea0d3b__s000__tb262144__v001\", \"oracle_delta\... | {
"index": "afrimgsm_hau_prompt_2__44ea0d3b__s000__tb262144__v001",
"uid": "afrimgsm_hau_prompt_2__44ea0d3b__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_hau_prompt_3__80af7bd6__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 257.21200950865847, \"nll_base\": 5.549900684541869, \"eval_ref\": \"afrimgsm_hau_prompt_3__80af7bd6__s000__tb131072__v000\", \"oracle_delta\... | {
"index": "afrimgsm_hau_prompt_3__80af7bd6__s000__tb131072__v000",
"uid": "afrimgsm_hau_prompt_3__80af7bd6__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_hau_prompt_3__80af7bd6__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 257.21200950865847, \"nll_base\": 5.549900684541869, \"eval_ref\": \"afrimgsm_hau_prompt_3__80af7bd6__s000__tb262144__v001\", \"oracle_delta\... | {
"index": "afrimgsm_hau_prompt_3__80af7bd6__s000__tb262144__v001",
"uid": "afrimgsm_hau_prompt_3__80af7bd6__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_hau_prompt_4__57246647__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 227.03752269966168, \"nll_base\": 5.4251153020575495, \"eval_ref\": \"afrimgsm_hau_prompt_4__57246647__s000__tb131072__v000\", \"oracle_delta... | {
"index": "afrimgsm_hau_prompt_4__57246647__s000__tb131072__v000",
"uid": "afrimgsm_hau_prompt_4__57246647__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_hau_prompt_4__57246647__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 227.03752269966168, \"nll_base\": 5.4251153020575495, \"eval_ref\": \"afrimgsm_hau_prompt_4__57246647__s000__tb262144__v001\", \"oracle_delta... | {
"index": "afrimgsm_hau_prompt_4__57246647__s000__tb262144__v001",
"uid": "afrimgsm_hau_prompt_4__57246647__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_hau_prompt_5__b827efed__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 337.0954511146612, \"nll_base\": 5.820366127977094, \"eval_ref\": \"afrimgsm_hau_prompt_5__b827efed__s000__tb131072__v000\", \"oracle_delta\"... | {
"index": "afrimgsm_hau_prompt_5__b827efed__s000__tb131072__v000",
"uid": "afrimgsm_hau_prompt_5__b827efed__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_hau_prompt_5__b827efed__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 337.0954511146612, \"nll_base\": 5.820366127977094, \"eval_ref\": \"afrimgsm_hau_prompt_5__b827efed__s000__tb262144__v001\", \"oracle_delta\"... | {
"index": "afrimgsm_hau_prompt_5__b827efed__s000__tb262144__v001",
"uid": "afrimgsm_hau_prompt_5__b827efed__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_ibo_prompt_1__2fb78415__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 521.850731389368, \"nll_base\": 6.257381591796875, \"eval_ref\": \"afrimgsm_ibo_prompt_1__2fb78415__s000__tb131072__v000\", \"oracle_delta\":... | {
"index": "afrimgsm_ibo_prompt_1__2fb78415__s000__tb131072__v000",
"uid": "afrimgsm_ibo_prompt_1__2fb78415__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_ibo_prompt_1__2fb78415__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 521.850731389368, \"nll_base\": 6.257381591796875, \"eval_ref\": \"afrimgsm_ibo_prompt_1__2fb78415__s000__tb262144__v001\", \"oracle_delta\":... | {
"index": "afrimgsm_ibo_prompt_1__2fb78415__s000__tb262144__v001",
"uid": "afrimgsm_ibo_prompt_1__2fb78415__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_ibo_prompt_2__20bf0a27__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 380.39344500579557, \"nll_base\": 5.941206098676802, \"eval_ref\": \"afrimgsm_ibo_prompt_2__20bf0a27__s000__tb131072__v000\", \"oracle_delta\... | {
"index": "afrimgsm_ibo_prompt_2__20bf0a27__s000__tb131072__v000",
"uid": "afrimgsm_ibo_prompt_2__20bf0a27__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_ibo_prompt_2__20bf0a27__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 380.39344500579557, \"nll_base\": 5.941206098676802, \"eval_ref\": \"afrimgsm_ibo_prompt_2__20bf0a27__s000__tb262144__v001\", \"oracle_delta\... | {
"index": "afrimgsm_ibo_prompt_2__20bf0a27__s000__tb262144__v001",
"uid": "afrimgsm_ibo_prompt_2__20bf0a27__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_ibo_prompt_3__69c8695e__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 208.76014857570033, \"nll_base\": 5.341185978480747, \"eval_ref\": \"afrimgsm_ibo_prompt_3__69c8695e__s000__tb131072__v000\", \"oracle_delta\... | {
"index": "afrimgsm_ibo_prompt_3__69c8695e__s000__tb131072__v000",
"uid": "afrimgsm_ibo_prompt_3__69c8695e__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_ibo_prompt_3__69c8695e__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 208.76014857570033, \"nll_base\": 5.341185978480747, \"eval_ref\": \"afrimgsm_ibo_prompt_3__69c8695e__s000__tb262144__v001\", \"oracle_delta\... | {
"index": "afrimgsm_ibo_prompt_3__69c8695e__s000__tb262144__v001",
"uid": "afrimgsm_ibo_prompt_3__69c8695e__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_ibo_prompt_4__daac28c2__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 215.62251582179857, \"nll_base\": 5.373529266892818, \"eval_ref\": \"afrimgsm_ibo_prompt_4__daac28c2__s000__tb262144__v000\", \"oracle_delta\... | {
"index": "afrimgsm_ibo_prompt_4__daac28c2__s000__tb262144__v000",
"uid": "afrimgsm_ibo_prompt_4__daac28c2__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_ibo_prompt_4__daac28c2__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 215.62251582179857, \"nll_base\": 5.373529266892818, \"eval_ref\": \"afrimgsm_ibo_prompt_4__daac28c2__s000__tb65536__v001\", \"oracle_delta\": ... | {
"index": "afrimgsm_ibo_prompt_4__daac28c2__s000__tb65536__v001",
"uid": "afrimgsm_ibo_prompt_4__daac28c2__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_ibo_prompt_5__82113355__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 505.4769224594105, \"nll_base\": 6.225502384519114, \"eval_ref\": \"afrimgsm_ibo_prompt_5__82113355__s000__tb131072__v000\", \"oracle_delta\"... | {
"index": "afrimgsm_ibo_prompt_5__82113355__s000__tb131072__v000",
"uid": "afrimgsm_ibo_prompt_5__82113355__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_ibo_prompt_5__82113355__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 505.4769224594105, \"nll_base\": 6.225502384519114, \"eval_ref\": \"afrimgsm_ibo_prompt_5__82113355__s000__tb262144__v001\", \"oracle_delta\"... | {
"index": "afrimgsm_ibo_prompt_5__82113355__s000__tb262144__v001",
"uid": "afrimgsm_ibo_prompt_5__82113355__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_kin_prompt_1__3d940d77__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 316.49065059758266, \"nll_base\": 5.757293701171875, \"eval_ref\": \"afrimgsm_kin_prompt_1__3d940d77__s000__tb131072__v000\", \"oracle_delta\... | {
"index": "afrimgsm_kin_prompt_1__3d940d77__s000__tb131072__v000",
"uid": "afrimgsm_kin_prompt_1__3d940d77__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_kin_prompt_1__3d940d77__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 316.49065059758266, \"nll_base\": 5.757293701171875, \"eval_ref\": \"afrimgsm_kin_prompt_1__3d940d77__s000__tb262144__v001\", \"oracle_delta\... | {
"index": "afrimgsm_kin_prompt_1__3d940d77__s000__tb262144__v001",
"uid": "afrimgsm_kin_prompt_1__3d940d77__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_kin_prompt_2__c6746194__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 385.8785157474673, \"nll_base\": 5.955522593885365, \"eval_ref\": \"afrimgsm_kin_prompt_2__c6746194__s000__tb131072__v000\", \"oracle_delta\"... | {
"index": "afrimgsm_kin_prompt_2__c6746194__s000__tb131072__v000",
"uid": "afrimgsm_kin_prompt_2__c6746194__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_kin_prompt_2__c6746194__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 385.8785157474673, \"nll_base\": 5.955522593885365, \"eval_ref\": \"afrimgsm_kin_prompt_2__c6746194__s000__tb262144__v001\", \"oracle_delta\"... | {
"index": "afrimgsm_kin_prompt_2__c6746194__s000__tb262144__v001",
"uid": "afrimgsm_kin_prompt_2__c6746194__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_kin_prompt_3__3e3359ec__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 294.56453399463334, \"nll_base\": 5.685498109797842, \"eval_ref\": \"afrimgsm_kin_prompt_3__3e3359ec__s000__tb131072__v000\", \"oracle_delta\... | {
"index": "afrimgsm_kin_prompt_3__3e3359ec__s000__tb131072__v000",
"uid": "afrimgsm_kin_prompt_3__3e3359ec__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_kin_prompt_3__3e3359ec__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 294.56453399463334, \"nll_base\": 5.685498109797842, \"eval_ref\": \"afrimgsm_kin_prompt_3__3e3359ec__s000__tb262144__v001\", \"oracle_delta\... | {
"index": "afrimgsm_kin_prompt_3__3e3359ec__s000__tb262144__v001",
"uid": "afrimgsm_kin_prompt_3__3e3359ec__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_kin_prompt_4__1c571793__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 166.15067608563885, \"nll_base\": 5.112895063920455, \"eval_ref\": \"afrimgsm_kin_prompt_4__1c571793__s000__tb131072__v001\", \"oracle_delta\... | {
"index": "afrimgsm_kin_prompt_4__1c571793__s000__tb131072__v001",
"uid": "afrimgsm_kin_prompt_4__1c571793__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_kin_prompt_4__1c571793__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 166.15067608563885, \"nll_base\": 5.112895063920455, \"eval_ref\": \"afrimgsm_kin_prompt_4__1c571793__s000__tb65536__v000\", \"oracle_delta\": ... | {
"index": "afrimgsm_kin_prompt_4__1c571793__s000__tb65536__v000",
"uid": "afrimgsm_kin_prompt_4__1c571793__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_kin_prompt_5__fc19f824__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 291.8919035230138, \"nll_base\": 5.676383540315448, \"eval_ref\": \"afrimgsm_kin_prompt_5__fc19f824__s000__tb131072__v001\", \"oracle_delta\"... | {
"index": "afrimgsm_kin_prompt_5__fc19f824__s000__tb131072__v001",
"uid": "afrimgsm_kin_prompt_5__fc19f824__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_kin_prompt_5__fc19f824__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 291.8919035230138, \"nll_base\": 5.676383540315448, \"eval_ref\": \"afrimgsm_kin_prompt_5__fc19f824__s000__tb65536__v000\", \"oracle_delta\": 0... | {
"index": "afrimgsm_kin_prompt_5__fc19f824__s000__tb65536__v000",
"uid": "afrimgsm_kin_prompt_5__fc19f824__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_lin_prompt_1__bc33a5c1__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 308.910802943953, \"nll_base\": 5.733052571614583, \"eval_ref\": \"afrimgsm_lin_prompt_1__bc33a5c1__s000__tb131072__v001\", \"oracle_delta\":... | {
"index": "afrimgsm_lin_prompt_1__bc33a5c1__s000__tb131072__v001",
"uid": "afrimgsm_lin_prompt_1__bc33a5c1__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_lin_prompt_1__bc33a5c1__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 308.910802943953, \"nll_base\": 5.733052571614583, \"eval_ref\": \"afrimgsm_lin_prompt_1__bc33a5c1__s000__tb65536__v000\", \"oracle_delta\": 0.... | {
"index": "afrimgsm_lin_prompt_1__bc33a5c1__s000__tb65536__v000",
"uid": "afrimgsm_lin_prompt_1__bc33a5c1__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_lin_prompt_2__ce68a6c4__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 177.7872085011589, \"nll_base\": 5.180587377303686, \"eval_ref\": \"afrimgsm_lin_prompt_2__ce68a6c4__s000__tb65536__v001\", \"oracle_delta\": 0... | {
"index": "afrimgsm_lin_prompt_2__ce68a6c4__s000__tb65536__v001",
"uid": "afrimgsm_lin_prompt_2__ce68a6c4__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_lin_prompt_3__a219f6a0__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 208.77490987327016, \"nll_base\": 5.341256685346086, \"eval_ref\": \"afrimgsm_lin_prompt_3__a219f6a0__s000__tb262144__v000\", \"oracle_delta\... | {
"index": "afrimgsm_lin_prompt_3__a219f6a0__s000__tb262144__v000",
"uid": "afrimgsm_lin_prompt_3__a219f6a0__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_lin_prompt_3__a219f6a0__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 208.77490987327016, \"nll_base\": 5.341256685346086, \"eval_ref\": \"afrimgsm_lin_prompt_3__a219f6a0__s000__tb65536__v001\", \"oracle_delta\": ... | {
"index": "afrimgsm_lin_prompt_3__a219f6a0__s000__tb65536__v001",
"uid": "afrimgsm_lin_prompt_3__a219f6a0__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_lin_prompt_4__5b669bea__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 119.36625822490349, \"nll_base\": 4.782196566589877, \"eval_ref\": \"afrimgsm_lin_prompt_4__5b669bea__s000__tb262144__v000\", \"oracle_delta\... | {
"index": "afrimgsm_lin_prompt_4__5b669bea__s000__tb262144__v000",
"uid": "afrimgsm_lin_prompt_4__5b669bea__s000__tb262144__v000",
"axis": "math",
"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.