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\": \"AraDiCE_ArabicMMLU_high_language_arabic-language_egy__49a26627__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 487.99032057790066, \"nll_base\": 6.190295570775082, \"eval_ref\": \"AraDiCE_ArabicMMLU_high_language_arabic-lan... | {
"index": "AraDiCE_ArabicMMLU_high_language_arabic-language_egy__49a26627__s000__tb65536__v000",
"uid": "AraDiCE_ArabicMMLU_high_language_arabic-language_egy__49a26627__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\": \"AraDiCE_ArabicMMLU_middle_language_arabic-language_lev__059c37a6__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 425.194343068896, \"nll_base\": 6.052546342213948, \"eval_ref\": \"AraDiCE_ArabicMMLU_middle_language_arabic... | {
"index": "AraDiCE_ArabicMMLU_middle_language_arabic-language_lev__059c37a6__s000__tb262144__v000",
"uid": "AraDiCE_ArabicMMLU_middle_language_arabic-language_lev__059c37a6__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\": \"AraDiCE_ArabicMMLU_middle_other_general-knowledge_egy__ea669b06__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 353.0687574456074, \"nll_base\": 5.866662818265248, \"eval_ref\": \"AraDiCE_ArabicMMLU_middle_other_general-kno... | {
"index": "AraDiCE_ArabicMMLU_middle_other_general-knowledge_egy__ea669b06__s000__tb65536__v000",
"uid": "AraDiCE_ArabicMMLU_middle_other_general-knowledge_egy__ea669b06__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\": \"AraDiCE_ArabicMMLU_middle_social-science_geography_egy__026d4b85__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 319.0885627576991, \"nll_base\": 5.765468690453506, \"eval_ref\": \"AraDiCE_ArabicMMLU_middle_social-science... | {
"index": "AraDiCE_ArabicMMLU_middle_social-science_geography_egy__026d4b85__s000__tb262144__v000",
"uid": "AraDiCE_ArabicMMLU_middle_social-science_geography_egy__026d4b85__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\": \"AraDiCE_ArabicMMLU_primary_stem_math_egy__12a26356__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 224.15840863843314, \"nll_base\": 5.412352983341661, \"eval_ref\": \"AraDiCE_ArabicMMLU_primary_stem_math_egy__12a26356__s00... | {
"index": "AraDiCE_ArabicMMLU_primary_stem_math_egy__12a26356__s000__tb65536__v000",
"uid": "AraDiCE_ArabicMMLU_primary_stem_math_egy__12a26356__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\": \"AraDiCE_ArabicMMLU_univ_social-science_accounting_egy__b7d1d838__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 416.52540738226054, \"nll_base\": 6.0319474618646165, \"eval_ref\": \"AraDiCE_ArabicMMLU_univ_social-science_... | {
"index": "AraDiCE_ArabicMMLU_univ_social-science_accounting_egy__b7d1d838__s000__tb131072__v000",
"uid": "AraDiCE_ArabicMMLU_univ_social-science_accounting_egy__b7d1d838__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\": \"AraDiCE_ArabicMMLU_univ_social-science_economics_egy__6b896521__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 497.75937653091887, \"nll_base\": 6.210116780598958, \"eval_ref\": \"AraDiCE_ArabicMMLU_univ_social-science_econ... | {
"index": "AraDiCE_ArabicMMLU_univ_social-science_economics_egy__6b896521__s000__tb65536__v000",
"uid": "AraDiCE_ArabicMMLU_univ_social-science_economics_egy__6b896521__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\": \"aclue_polysemy_resolution__89542a15__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 793.811689143417, \"nll_base\": 6.676846265792847, \"eval_ref\": \"aclue_polysemy_resolution__89542a15__s000__tb65536__v000\", \"oracle_del... | {
"index": "aclue_polysemy_resolution__89542a15__s000__tb65536__v000",
"uid": "aclue_polysemy_resolution__89542a15__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\": \"acp_just_bool__fd790e79__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 12357.324249038631, \"nll_base\": 9.422004222869873, \"eval_ref\": \"acp_just_bool__fd790e79__s000__tb65536__v000\", \"oracle_delta\": -0.0604796409606... | {
"index": "acp_just_bool__fd790e79__s000__tb65536__v000",
"uid": "acp_just_bool__fd790e79__s000__tb65536__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"acp_reach_bool__2ecaff9f__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 16608.002277456657, \"nll_base\": 9.717639923095703, \"eval_ref\": \"acp_reach_bool__2ecaff9f__s000__tb131072__v000\", \"oracle_delta\": 0.407837867... | {
"index": "acp_reach_bool__2ecaff9f__s000__tb131072__v000",
"uid": "acp_reach_bool__2ecaff9f__s000__tb131072__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"acp_val_bool__1ce850de__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 15652.564450030064, \"nll_base\": 9.658390045166016, \"eval_ref\": \"acp_val_bool__1ce850de__s000__tb131072__v000\", \"oracle_delta\": 0.0775561332702... | {
"index": "acp_val_bool__1ce850de__s000__tb131072__v000",
"uid": "acp_val_bool__1ce850de__s000__tb131072__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"advanced_ai_risk_fewshot-coordinate-itself__fd688499__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 239.57569906686936, \"nll_base\": 5.478869438171387, \"eval_ref\": \"advanced_ai_risk_fewshot-coordinate-itself__fd688499_... | {
"index": "advanced_ai_risk_fewshot-coordinate-itself__fd688499__s000__tb65536__v000",
"uid": "advanced_ai_risk_fewshot-coordinate-itself__fd688499__s000__tb65536__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"advanced_ai_risk_fewshot-power-seeking-inclination__4bf345db__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 194.76642008847645, \"nll_base\": 5.271800994873047, \"eval_ref\": \"advanced_ai_risk_fewshot-power-seeking-incl... | {
"index": "advanced_ai_risk_fewshot-power-seeking-inclination__4bf345db__s000__tb131072__v000",
"uid": "advanced_ai_risk_fewshot-power-seeking-inclination__4bf345db__s000__tb131072__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"advanced_ai_risk_fewshot-self-awareness-good-text-model__da894797__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 190.3644642954135, \"nll_base\": 5.248940467834473, \"eval_ref\": \"advanced_ai_risk_fewshot-self-awareness-g... | {
"index": "advanced_ai_risk_fewshot-self-awareness-good-text-model__da894797__s000__tb65536__v000",
"uid": "advanced_ai_risk_fewshot-self-awareness-good-text-model__da894797__s000__tb65536__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"advanced_ai_risk_fewshot-self-awareness-training-web-gpt__eb7dbaf3__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 214.70553048544875, \"nll_base\": 5.369267463684082, \"eval_ref\": \"advanced_ai_risk_fewshot-self-awarene... | {
"index": "advanced_ai_risk_fewshot-self-awareness-training-web-gpt__eb7dbaf3__s000__tb131072__v000",
"uid": "advanced_ai_risk_fewshot-self-awareness-training-web-gpt__eb7dbaf3__s000__tb131072__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"advanced_ai_risk_human-self-awareness-good-text-model__3b02f583__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 261.1155067239219, \"nll_base\": 5.564962863922119, \"eval_ref\": \"advanced_ai_risk_human-self-awareness-goo... | {
"index": "advanced_ai_risk_human-self-awareness-good-text-model__3b02f583__s000__tb131072__v000",
"uid": "advanced_ai_risk_human-self-awareness-good-text-model__3b02f583__s000__tb131072__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"advanced_ai_risk_human-self-awareness-web-gpt__69b176b2__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 260.4095596214393, \"nll_base\": 5.562255620956421, \"eval_ref\": \"advanced_ai_risk_human-self-awareness-web-gpt__69... | {
"index": "advanced_ai_risk_human-self-awareness-web-gpt__69b176b2__s000__tb131072__v000",
"uid": "advanced_ai_risk_human-self-awareness-web-gpt__69b176b2__s000__tb131072__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"advanced_ai_risk_lm-myopic-reward__804ac07f__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 238.16627285220918, \"nll_base\": 5.472969055175781, \"eval_ref\": \"advanced_ai_risk_lm-myopic-reward__804ac07f__s000__tb131072_... | {
"index": "advanced_ai_risk_lm-myopic-reward__804ac07f__s000__tb131072__v000",
"uid": "advanced_ai_risk_lm-myopic-reward__804ac07f__s000__tb131072__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"afrimgsm_cot_eng_prompt_1__6f2504f2__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 199.5513240017189, \"nll_base\": 5.296071466409935, \"eval_ref\": \"afrimgsm_cot_eng_prompt_1__6f2504f2__s000__tb131072__v000\", \"oracle... | {
"index": "afrimgsm_cot_eng_prompt_1__6f2504f2__s000__tb131072__v000",
"uid": "afrimgsm_cot_eng_prompt_1__6f2504f2__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_eng_prompt_1__6f2504f2__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 199.5513240017189, \"nll_base\": 5.296071466409935, \"eval_ref\": \"afrimgsm_cot_eng_prompt_1__6f2504f2__s000__tb262144__v001\", \"oracle... | {
"index": "afrimgsm_cot_eng_prompt_1__6f2504f2__s000__tb262144__v001",
"uid": "afrimgsm_cot_eng_prompt_1__6f2504f2__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_eng_prompt_2__0bd42e3c__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 181.91822732404924, \"nll_base\": 5.203557285693807, \"eval_ref\": \"afrimgsm_cot_eng_prompt_2__0bd42e3c__s000__tb65536__v001\", \"oracle_d... | {
"index": "afrimgsm_cot_eng_prompt_2__0bd42e3c__s000__tb65536__v001",
"uid": "afrimgsm_cot_eng_prompt_2__0bd42e3c__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_eng_prompt_3__2d6d2958__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 153.01587225096256, \"nll_base\": 5.0305416562141625, \"eval_ref\": \"afrimgsm_cot_eng_prompt_3__2d6d2958__s000__tb65536__v001\", \"oracle_... | {
"index": "afrimgsm_cot_eng_prompt_3__2d6d2958__s000__tb65536__v001",
"uid": "afrimgsm_cot_eng_prompt_3__2d6d2958__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_eng_prompt_4__c53149e6__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 146.80952888839067, \"nll_base\": 4.989136024757668, \"eval_ref\": \"afrimgsm_cot_eng_prompt_4__c53149e6__s000__tb131072__v000\", \"oracl... | {
"index": "afrimgsm_cot_eng_prompt_4__c53149e6__s000__tb131072__v000",
"uid": "afrimgsm_cot_eng_prompt_4__c53149e6__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_eng_prompt_4__c53149e6__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 146.80952888839067, \"nll_base\": 4.989136024757668, \"eval_ref\": \"afrimgsm_cot_eng_prompt_4__c53149e6__s000__tb262144__v001\", \"oracl... | {
"index": "afrimgsm_cot_eng_prompt_4__c53149e6__s000__tb262144__v001",
"uid": "afrimgsm_cot_eng_prompt_4__c53149e6__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_eng_prompt_5__232a1ffa__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 172.1262529985372, \"nll_base\": 5.148228236607143, \"eval_ref\": \"afrimgsm_cot_eng_prompt_5__232a1ffa__s000__tb131072__v000\", \"oracle... | {
"index": "afrimgsm_cot_eng_prompt_5__232a1ffa__s000__tb131072__v000",
"uid": "afrimgsm_cot_eng_prompt_5__232a1ffa__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_eng_prompt_5__232a1ffa__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 172.1262529985372, \"nll_base\": 5.148228236607143, \"eval_ref\": \"afrimgsm_cot_eng_prompt_5__232a1ffa__s000__tb262144__v001\", \"oracle... | {
"index": "afrimgsm_cot_eng_prompt_5__232a1ffa__s000__tb262144__v001",
"uid": "afrimgsm_cot_eng_prompt_5__232a1ffa__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_ewe_prompt_1__47b331b6__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 470.0663050681154, \"nll_base\": 6.152873759367028, \"eval_ref\": \"afrimgsm_cot_ewe_prompt_1__47b331b6__s000__tb131072__v001\", \"oracle... | {
"index": "afrimgsm_cot_ewe_prompt_1__47b331b6__s000__tb131072__v001",
"uid": "afrimgsm_cot_ewe_prompt_1__47b331b6__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_ewe_prompt_1__47b331b6__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 470.0663050681154, \"nll_base\": 6.152873759367028, \"eval_ref\": \"afrimgsm_cot_ewe_prompt_1__47b331b6__s000__tb65536__v000\", \"oracle_de... | {
"index": "afrimgsm_cot_ewe_prompt_1__47b331b6__s000__tb65536__v000",
"uid": "afrimgsm_cot_ewe_prompt_1__47b331b6__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_ewe_prompt_3__165e8625__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 164.91932008250302, \"nll_base\": 5.105456385119208, \"eval_ref\": \"afrimgsm_cot_ewe_prompt_3__165e8625__s000__tb262144__v000\", \"oracl... | {
"index": "afrimgsm_cot_ewe_prompt_3__165e8625__s000__tb262144__v000",
"uid": "afrimgsm_cot_ewe_prompt_3__165e8625__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_ewe_prompt_3__165e8625__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 164.91932008250302, \"nll_base\": 5.105456385119208, \"eval_ref\": \"afrimgsm_cot_ewe_prompt_3__165e8625__s000__tb65536__v001\", \"oracle_d... | {
"index": "afrimgsm_cot_ewe_prompt_3__165e8625__s000__tb65536__v001",
"uid": "afrimgsm_cot_ewe_prompt_3__165e8625__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_ewe_prompt_4__3981749c__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 115.91189493056049, \"nll_base\": 4.752830376059322, \"eval_ref\": \"afrimgsm_cot_ewe_prompt_4__3981749c__s000__tb131072__v001\", \"oracl... | {
"index": "afrimgsm_cot_ewe_prompt_4__3981749c__s000__tb131072__v001",
"uid": "afrimgsm_cot_ewe_prompt_4__3981749c__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_ewe_prompt_4__3981749c__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 115.91189493056049, \"nll_base\": 4.752830376059322, \"eval_ref\": \"afrimgsm_cot_ewe_prompt_4__3981749c__s000__tb65536__v000\", \"oracle_d... | {
"index": "afrimgsm_cot_ewe_prompt_4__3981749c__s000__tb65536__v000",
"uid": "afrimgsm_cot_ewe_prompt_4__3981749c__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_ewe_prompt_5__2430bcac__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 349.85841494740123, \"nll_base\": 5.857528543917932, \"eval_ref\": \"afrimgsm_cot_ewe_prompt_5__2430bcac__s000__tb131072__v000\", \"oracl... | {
"index": "afrimgsm_cot_ewe_prompt_5__2430bcac__s000__tb131072__v000",
"uid": "afrimgsm_cot_ewe_prompt_5__2430bcac__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_ewe_prompt_5__2430bcac__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 349.85841494740123, \"nll_base\": 5.857528543917932, \"eval_ref\": \"afrimgsm_cot_ewe_prompt_5__2430bcac__s000__tb262144__v001\", \"oracl... | {
"index": "afrimgsm_cot_ewe_prompt_5__2430bcac__s000__tb262144__v001",
"uid": "afrimgsm_cot_ewe_prompt_5__2430bcac__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_fra_prompt_1__b6555682__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 246.61010823779566, \"nll_base\": 5.507808580311067, \"eval_ref\": \"afrimgsm_cot_fra_prompt_1__b6555682__s000__tb262144__v000\", \"oracl... | {
"index": "afrimgsm_cot_fra_prompt_1__b6555682__s000__tb262144__v000",
"uid": "afrimgsm_cot_fra_prompt_1__b6555682__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_fra_prompt_1__b6555682__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 246.61010823779566, \"nll_base\": 5.507808580311067, \"eval_ref\": \"afrimgsm_cot_fra_prompt_1__b6555682__s000__tb65536__v001\", \"oracle_d... | {
"index": "afrimgsm_cot_fra_prompt_1__b6555682__s000__tb65536__v001",
"uid": "afrimgsm_cot_fra_prompt_1__b6555682__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_fra_prompt_2__a38c2fbc__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 163.32869284775077, \"nll_base\": 5.095764690896739, \"eval_ref\": \"afrimgsm_cot_fra_prompt_2__a38c2fbc__s000__tb65536__v000\", \"oracle_d... | {
"index": "afrimgsm_cot_fra_prompt_2__a38c2fbc__s000__tb65536__v000",
"uid": "afrimgsm_cot_fra_prompt_2__a38c2fbc__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_fra_prompt_3__ad57ed1f__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 239.6885842468695, \"nll_base\": 5.479340515136719, \"eval_ref\": \"afrimgsm_cot_fra_prompt_3__ad57ed1f__s000__tb131072__v001\", \"oracle... | {
"index": "afrimgsm_cot_fra_prompt_3__ad57ed1f__s000__tb131072__v001",
"uid": "afrimgsm_cot_fra_prompt_3__ad57ed1f__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_fra_prompt_3__ad57ed1f__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 239.6885842468695, \"nll_base\": 5.479340515136719, \"eval_ref\": \"afrimgsm_cot_fra_prompt_3__ad57ed1f__s000__tb65536__v000\", \"oracle_de... | {
"index": "afrimgsm_cot_fra_prompt_3__ad57ed1f__s000__tb65536__v000",
"uid": "afrimgsm_cot_fra_prompt_3__ad57ed1f__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_fra_prompt_4__b381b79b__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 145.29145300003006, \"nll_base\": 4.9787417457217265, \"eval_ref\": \"afrimgsm_cot_fra_prompt_4__b381b79b__s000__tb262144__v000\", \"orac... | {
"index": "afrimgsm_cot_fra_prompt_4__b381b79b__s000__tb262144__v000",
"uid": "afrimgsm_cot_fra_prompt_4__b381b79b__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_fra_prompt_5__81494d90__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 203.1631208324441, \"nll_base\": 5.3140092072663485, \"eval_ref\": \"afrimgsm_cot_fra_prompt_5__81494d90__s000__tb262144__v000\", \"oracl... | {
"index": "afrimgsm_cot_fra_prompt_5__81494d90__s000__tb262144__v000",
"uid": "afrimgsm_cot_fra_prompt_5__81494d90__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_fra_prompt_5__81494d90__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 203.1631208324441, \"nll_base\": 5.3140092072663485, \"eval_ref\": \"afrimgsm_cot_fra_prompt_5__81494d90__s000__tb65536__v001\", \"oracle_d... | {
"index": "afrimgsm_cot_fra_prompt_5__81494d90__s000__tb65536__v001",
"uid": "afrimgsm_cot_fra_prompt_5__81494d90__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_hau_prompt_1__ae36bc1e__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 336.05270516686375, \"nll_base\": 5.81726800827753, \"eval_ref\": \"afrimgsm_cot_hau_prompt_1__ae36bc1e__s000__tb262144__v000\", \"oracle... | {
"index": "afrimgsm_cot_hau_prompt_1__ae36bc1e__s000__tb262144__v000",
"uid": "afrimgsm_cot_hau_prompt_1__ae36bc1e__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_hau_prompt_1__ae36bc1e__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 336.05270516686375, \"nll_base\": 5.81726800827753, \"eval_ref\": \"afrimgsm_cot_hau_prompt_1__ae36bc1e__s000__tb65536__v001\", \"oracle_de... | {
"index": "afrimgsm_cot_hau_prompt_1__ae36bc1e__s000__tb65536__v001",
"uid": "afrimgsm_cot_hau_prompt_1__ae36bc1e__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_hau_prompt_2__cb660f10__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 297.20816100985286, \"nll_base\": 5.694432772122896, \"eval_ref\": \"afrimgsm_cot_hau_prompt_2__cb660f10__s000__tb262144__v000\", \"oracl... | {
"index": "afrimgsm_cot_hau_prompt_2__cb660f10__s000__tb262144__v000",
"uid": "afrimgsm_cot_hau_prompt_2__cb660f10__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_hau_prompt_2__cb660f10__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 297.20816100985286, \"nll_base\": 5.694432772122896, \"eval_ref\": \"afrimgsm_cot_hau_prompt_2__cb660f10__s000__tb65536__v001\", \"oracle_d... | {
"index": "afrimgsm_cot_hau_prompt_2__cb660f10__s000__tb65536__v001",
"uid": "afrimgsm_cot_hau_prompt_2__cb660f10__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_hau_prompt_3__f9e1b333__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 313.61197996266714, \"nll_base\": 5.748156489158164, \"eval_ref\": \"afrimgsm_cot_hau_prompt_3__f9e1b333__s000__tb131072__v000\", \"oracl... | {
"index": "afrimgsm_cot_hau_prompt_3__f9e1b333__s000__tb131072__v000",
"uid": "afrimgsm_cot_hau_prompt_3__f9e1b333__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_hau_prompt_3__f9e1b333__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 313.61197996266714, \"nll_base\": 5.748156489158164, \"eval_ref\": \"afrimgsm_cot_hau_prompt_3__f9e1b333__s000__tb262144__v001\", \"oracl... | {
"index": "afrimgsm_cot_hau_prompt_3__f9e1b333__s000__tb262144__v001",
"uid": "afrimgsm_cot_hau_prompt_3__f9e1b333__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_hau_prompt_4__7617ded7__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 159.38358067535404, \"nll_base\": 5.071313753995028, \"eval_ref\": \"afrimgsm_cot_hau_prompt_4__7617ded7__s000__tb131072__v001\", \"oracl... | {
"index": "afrimgsm_cot_hau_prompt_4__7617ded7__s000__tb131072__v001",
"uid": "afrimgsm_cot_hau_prompt_4__7617ded7__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_hau_prompt_4__7617ded7__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 159.38358067535404, \"nll_base\": 5.071313753995028, \"eval_ref\": \"afrimgsm_cot_hau_prompt_4__7617ded7__s000__tb65536__v000\", \"oracle_d... | {
"index": "afrimgsm_cot_hau_prompt_4__7617ded7__s000__tb65536__v000",
"uid": "afrimgsm_cot_hau_prompt_4__7617ded7__s000__tb65536__v000",
"axis": "math",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
End of preview. Expand in Data Studio
README.md exists but content is empty.
- Downloads last month
- 59