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\": \"lambada_openai_mt_stablelm_de__11084541__s015__tb65536__v045\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2035.3998442646714, \"nll_base\": 7.618447562244451, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s015__tb65536__v045\", \"...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s015__tb65536__v045", "uid": "lambada_openai_mt_stablelm_de__11084541__s015__tb65536__v045", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_de__11084541__s016__tb131072__v049\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2401.400464551492, \"nll_base\": 7.783807373046875, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s016__tb131072__v049\", ...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s016__tb131072__v049", "uid": "lambada_openai_mt_stablelm_de__11084541__s016__tb131072__v049", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_de__11084541__s016__tb262144__v050\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2401.400464551492, \"nll_base\": 7.783807373046875, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s016__tb262144__v050\", ...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s016__tb262144__v050", "uid": "lambada_openai_mt_stablelm_de__11084541__s016__tb262144__v050", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_de__11084541__s016__tb65536__v048\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2401.400464551492, \"nll_base\": 7.783807373046875, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s016__tb65536__v048\", \"o...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s016__tb65536__v048", "uid": "lambada_openai_mt_stablelm_de__11084541__s016__tb65536__v048", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_de__11084541__s017__tb131072__v052\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1920.8260057316013, \"nll_base\": 7.560510583826013, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s017__tb131072__v052\",...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s017__tb131072__v052", "uid": "lambada_openai_mt_stablelm_de__11084541__s017__tb131072__v052", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_de__11084541__s017__tb262144__v053\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1920.8260057316013, \"nll_base\": 7.560510583826013, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s017__tb262144__v053\",...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s017__tb262144__v053", "uid": "lambada_openai_mt_stablelm_de__11084541__s017__tb262144__v053", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_de__11084541__s017__tb65536__v051\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1920.8260057316013, \"nll_base\": 7.560510583826013, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s017__tb65536__v051\", \"...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s017__tb65536__v051", "uid": "lambada_openai_mt_stablelm_de__11084541__s017__tb65536__v051", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_de__11084541__s018__tb131072__v055\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1869.8516017846011, \"nll_base\": 7.5336143493652346, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s018__tb131072__v055\"...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s018__tb131072__v055", "uid": "lambada_openai_mt_stablelm_de__11084541__s018__tb131072__v055", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_de__11084541__s018__tb262144__v056\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1869.8516017846011, \"nll_base\": 7.5336143493652346, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s018__tb262144__v056\"...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s018__tb262144__v056", "uid": "lambada_openai_mt_stablelm_de__11084541__s018__tb262144__v056", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_de__11084541__s018__tb65536__v054\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1869.8516017846011, \"nll_base\": 7.5336143493652346, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s018__tb65536__v054\", \...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s018__tb65536__v054", "uid": "lambada_openai_mt_stablelm_de__11084541__s018__tb65536__v054", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_de__11084541__s019__tb131072__v058\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2091.330051641516, \"nll_base\": 7.645555530894886, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s019__tb131072__v058\", ...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s019__tb131072__v058", "uid": "lambada_openai_mt_stablelm_de__11084541__s019__tb131072__v058", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_de__11084541__s019__tb65536__v057\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2091.330051641516, \"nll_base\": 7.645555530894886, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s019__tb65536__v057\", \"o...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s019__tb65536__v057", "uid": "lambada_openai_mt_stablelm_de__11084541__s019__tb65536__v057", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_en__1cca186a__s000__tb131072__v002\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 33884.58031861122, \"nll_base\": 10.43071533203125, \"eval_ref\": \"lambada_openai_mt_stablelm_en__1cca186a__s000__tb131072__v002\", ...
{ "index": "lambada_openai_mt_stablelm_en__1cca186a__s000__tb131072__v002", "uid": "lambada_openai_mt_stablelm_en__1cca186a__s000__tb131072__v002", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_en__1cca186a__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 33884.58031861122, \"nll_base\": 10.43071533203125, \"eval_ref\": \"lambada_openai_mt_stablelm_en__1cca186a__s000__tb262144__v000\", ...
{ "index": "lambada_openai_mt_stablelm_en__1cca186a__s000__tb262144__v000", "uid": "lambada_openai_mt_stablelm_en__1cca186a__s000__tb262144__v000", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_en__1cca186a__s002__tb131072__v008\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 15381.252552421805, \"nll_base\": 9.64090468008307, \"eval_ref\": \"lambada_openai_mt_stablelm_en__1cca186a__s002__tb131072__v008\", ...
{ "index": "lambada_openai_mt_stablelm_en__1cca186a__s002__tb131072__v008", "uid": "lambada_openai_mt_stablelm_en__1cca186a__s002__tb131072__v008", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_en__1cca186a__s002__tb262144__v006\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 15381.252552421805, \"nll_base\": 9.64090468008307, \"eval_ref\": \"lambada_openai_mt_stablelm_en__1cca186a__s002__tb262144__v006\", ...
{ "index": "lambada_openai_mt_stablelm_en__1cca186a__s002__tb262144__v006", "uid": "lambada_openai_mt_stablelm_en__1cca186a__s002__tb262144__v006", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_en__1cca186a__s005__tb131072__v017\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 25885.59593824333, \"nll_base\": 10.161441951603084, \"eval_ref\": \"lambada_openai_mt_stablelm_en__1cca186a__s005__tb131072__v017\",...
{ "index": "lambada_openai_mt_stablelm_en__1cca186a__s005__tb131072__v017", "uid": "lambada_openai_mt_stablelm_en__1cca186a__s005__tb131072__v017", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_en__1cca186a__s005__tb262144__v015\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 25885.59593824333, \"nll_base\": 10.161441951603084, \"eval_ref\": \"lambada_openai_mt_stablelm_en__1cca186a__s005__tb262144__v015\",...
{ "index": "lambada_openai_mt_stablelm_en__1cca186a__s005__tb262144__v015", "uid": "lambada_openai_mt_stablelm_en__1cca186a__s005__tb262144__v015", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_en__1cca186a__s006__tb131072__v020\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 19844.359588809737, \"nll_base\": 9.895675094039351, \"eval_ref\": \"lambada_openai_mt_stablelm_en__1cca186a__s006__tb131072__v020\",...
{ "index": "lambada_openai_mt_stablelm_en__1cca186a__s006__tb131072__v020", "uid": "lambada_openai_mt_stablelm_en__1cca186a__s006__tb131072__v020", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_en__1cca186a__s006__tb262144__v018\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 19844.359588809737, \"nll_base\": 9.895675094039351, \"eval_ref\": \"lambada_openai_mt_stablelm_en__1cca186a__s006__tb262144__v018\",...
{ "index": "lambada_openai_mt_stablelm_en__1cca186a__s006__tb262144__v018", "uid": "lambada_openai_mt_stablelm_en__1cca186a__s006__tb262144__v018", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_en__1cca186a__s006__tb65536__v019\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 19844.359588809737, \"nll_base\": 9.895675094039351, \"eval_ref\": \"lambada_openai_mt_stablelm_en__1cca186a__s006__tb65536__v019\", \"...
{ "index": "lambada_openai_mt_stablelm_en__1cca186a__s006__tb65536__v019", "uid": "lambada_openai_mt_stablelm_en__1cca186a__s006__tb65536__v019", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_en__1cca186a__s009__tb131072__v029\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 79901.21036375256, \"nll_base\": 11.288546280122139, \"eval_ref\": \"lambada_openai_mt_stablelm_en__1cca186a__s009__tb131072__v029\",...
{ "index": "lambada_openai_mt_stablelm_en__1cca186a__s009__tb131072__v029", "uid": "lambada_openai_mt_stablelm_en__1cca186a__s009__tb131072__v029", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_en__1cca186a__s009__tb262144__v027\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 79901.21036375256, \"nll_base\": 11.288546280122139, \"eval_ref\": \"lambada_openai_mt_stablelm_en__1cca186a__s009__tb262144__v027\",...
{ "index": "lambada_openai_mt_stablelm_en__1cca186a__s009__tb262144__v027", "uid": "lambada_openai_mt_stablelm_en__1cca186a__s009__tb262144__v027", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_en__1cca186a__s009__tb65536__v028\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 79901.21036375256, \"nll_base\": 11.288546280122139, \"eval_ref\": \"lambada_openai_mt_stablelm_en__1cca186a__s009__tb65536__v028\", \"...
{ "index": "lambada_openai_mt_stablelm_en__1cca186a__s009__tb65536__v028", "uid": "lambada_openai_mt_stablelm_en__1cca186a__s009__tb65536__v028", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_en__1cca186a__s010__tb262144__v030\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 23615.46325559739, \"nll_base\": 10.069656999143836, \"eval_ref\": \"lambada_openai_mt_stablelm_en__1cca186a__s010__tb262144__v030\",...
{ "index": "lambada_openai_mt_stablelm_en__1cca186a__s010__tb262144__v030", "uid": "lambada_openai_mt_stablelm_en__1cca186a__s010__tb262144__v030", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_en__1cca186a__s011__tb65536__v034\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 20702.282446234465, \"nll_base\": 9.937999236278046, \"eval_ref\": \"lambada_openai_mt_stablelm_en__1cca186a__s011__tb65536__v034\", \"...
{ "index": "lambada_openai_mt_stablelm_en__1cca186a__s011__tb65536__v034", "uid": "lambada_openai_mt_stablelm_en__1cca186a__s011__tb65536__v034", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_en__1cca186a__s012__tb131072__v038\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 13185.057773223441, \"nll_base\": 9.486839480516387, \"eval_ref\": \"lambada_openai_mt_stablelm_en__1cca186a__s012__tb131072__v038\",...
{ "index": "lambada_openai_mt_stablelm_en__1cca186a__s012__tb131072__v038", "uid": "lambada_openai_mt_stablelm_en__1cca186a__s012__tb131072__v038", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_en__1cca186a__s012__tb262144__v036\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 13185.057773223441, \"nll_base\": 9.486839480516387, \"eval_ref\": \"lambada_openai_mt_stablelm_en__1cca186a__s012__tb262144__v036\",...
{ "index": "lambada_openai_mt_stablelm_en__1cca186a__s012__tb262144__v036", "uid": "lambada_openai_mt_stablelm_en__1cca186a__s012__tb262144__v036", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_en__1cca186a__s012__tb65536__v037\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 13185.057773223441, \"nll_base\": 9.486839480516387, \"eval_ref\": \"lambada_openai_mt_stablelm_en__1cca186a__s012__tb65536__v037\", \"...
{ "index": "lambada_openai_mt_stablelm_en__1cca186a__s012__tb65536__v037", "uid": "lambada_openai_mt_stablelm_en__1cca186a__s012__tb65536__v037", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_en__1cca186a__s013__tb131072__v041\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 26280.685698156205, \"nll_base\": 10.176589564273232, \"eval_ref\": \"lambada_openai_mt_stablelm_en__1cca186a__s013__tb131072__v041\"...
{ "index": "lambada_openai_mt_stablelm_en__1cca186a__s013__tb131072__v041", "uid": "lambada_openai_mt_stablelm_en__1cca186a__s013__tb131072__v041", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_en__1cca186a__s013__tb262144__v039\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 26280.685698156205, \"nll_base\": 10.176589564273232, \"eval_ref\": \"lambada_openai_mt_stablelm_en__1cca186a__s013__tb262144__v039\"...
{ "index": "lambada_openai_mt_stablelm_en__1cca186a__s013__tb262144__v039", "uid": "lambada_openai_mt_stablelm_en__1cca186a__s013__tb262144__v039", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_en__1cca186a__s013__tb65536__v040\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 26280.685698156205, \"nll_base\": 10.176589564273232, \"eval_ref\": \"lambada_openai_mt_stablelm_en__1cca186a__s013__tb65536__v040\", \...
{ "index": "lambada_openai_mt_stablelm_en__1cca186a__s013__tb65536__v040", "uid": "lambada_openai_mt_stablelm_en__1cca186a__s013__tb65536__v040", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_en__1cca186a__s014__tb131072__v044\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 26891.50823048215, \"nll_base\": 10.199565836588542, \"eval_ref\": \"lambada_openai_mt_stablelm_en__1cca186a__s014__tb131072__v044\",...
{ "index": "lambada_openai_mt_stablelm_en__1cca186a__s014__tb131072__v044", "uid": "lambada_openai_mt_stablelm_en__1cca186a__s014__tb131072__v044", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_en__1cca186a__s014__tb65536__v043\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 26891.50823048215, \"nll_base\": 10.199565836588542, \"eval_ref\": \"lambada_openai_mt_stablelm_en__1cca186a__s014__tb65536__v043\", \"...
{ "index": "lambada_openai_mt_stablelm_en__1cca186a__s014__tb65536__v043", "uid": "lambada_openai_mt_stablelm_en__1cca186a__s014__tb65536__v043", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_en__1cca186a__s016__tb131072__v050\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 18155.72127474596, \"nll_base\": 9.806741011770148, \"eval_ref\": \"lambada_openai_mt_stablelm_en__1cca186a__s016__tb131072__v050\", ...
{ "index": "lambada_openai_mt_stablelm_en__1cca186a__s016__tb131072__v050", "uid": "lambada_openai_mt_stablelm_en__1cca186a__s016__tb131072__v050", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_en__1cca186a__s016__tb262144__v048\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 18155.72127474596, \"nll_base\": 9.806741011770148, \"eval_ref\": \"lambada_openai_mt_stablelm_en__1cca186a__s016__tb262144__v048\", ...
{ "index": "lambada_openai_mt_stablelm_en__1cca186a__s016__tb262144__v048", "uid": "lambada_openai_mt_stablelm_en__1cca186a__s016__tb262144__v048", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_en__1cca186a__s016__tb65536__v049\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 18155.72127474596, \"nll_base\": 9.806741011770148, \"eval_ref\": \"lambada_openai_mt_stablelm_en__1cca186a__s016__tb65536__v049\", \"o...
{ "index": "lambada_openai_mt_stablelm_en__1cca186a__s016__tb65536__v049", "uid": "lambada_openai_mt_stablelm_en__1cca186a__s016__tb65536__v049", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_en__1cca186a__s017__tb262144__v051\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 114905.58654197231, \"nll_base\": 11.651866083559783, \"eval_ref\": \"lambada_openai_mt_stablelm_en__1cca186a__s017__tb262144__v051\"...
{ "index": "lambada_openai_mt_stablelm_en__1cca186a__s017__tb262144__v051", "uid": "lambada_openai_mt_stablelm_en__1cca186a__s017__tb262144__v051", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_en__1cca186a__s019__tb262144__v057\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 56298.833283769745, \"nll_base\": 10.938429090711805, \"eval_ref\": \"lambada_openai_mt_stablelm_en__1cca186a__s019__tb262144__v057\"...
{ "index": "lambada_openai_mt_stablelm_en__1cca186a__s019__tb262144__v057", "uid": "lambada_openai_mt_stablelm_en__1cca186a__s019__tb262144__v057", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3269.298270600144, \"nll_base\": 8.092330644715506, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s000__tb131072__v000\", ...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s000__tb131072__v000", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s000__tb131072__v000", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3269.298270600144, \"nll_base\": 8.092330644715506, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s000__tb262144__v001\", ...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s000__tb262144__v001", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s000__tb262144__v001", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s000__tb65536__v002\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3269.298270600144, \"nll_base\": 8.092330644715506, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s000__tb65536__v002\", \"o...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s000__tb65536__v002", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s000__tb65536__v002", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s001__tb131072__v003\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2273.3074051461404, \"nll_base\": 7.728991056743421, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s001__tb131072__v003\",...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s001__tb131072__v003", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s001__tb131072__v003", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s001__tb262144__v004\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2273.3074051461404, \"nll_base\": 7.728991056743421, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s001__tb262144__v004\",...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s001__tb262144__v004", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s001__tb262144__v004", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s001__tb65536__v005\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2273.3074051461404, \"nll_base\": 7.728991056743421, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s001__tb65536__v005\", \"...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s001__tb65536__v005", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s001__tb65536__v005", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s002__tb131072__v006\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3455.770425358257, \"nll_base\": 8.147800699869792, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s002__tb131072__v006\", ...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s002__tb131072__v006", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s002__tb131072__v006", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s002__tb262144__v007\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3455.770425358257, \"nll_base\": 8.147800699869792, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s002__tb262144__v007\", ...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s002__tb262144__v007", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s002__tb262144__v007", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s002__tb65536__v008\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3455.770425358257, \"nll_base\": 8.147800699869792, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s002__tb65536__v008\", \"o...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s002__tb65536__v008", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s002__tb65536__v008", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s003__tb131072__v009\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3246.2579154695522, \"nll_base\": 8.08525820131655, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s003__tb131072__v009\", ...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s003__tb131072__v009", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s003__tb131072__v009", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s003__tb262144__v010\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3246.2579154695522, \"nll_base\": 8.08525820131655, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s003__tb262144__v010\", ...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s003__tb262144__v010", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s003__tb262144__v010", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s003__tb65536__v011\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3246.2579154695522, \"nll_base\": 8.08525820131655, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s003__tb65536__v011\", \"o...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s003__tb65536__v011", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s003__tb65536__v011", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s004__tb131072__v012\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2200.4135732277364, \"nll_base\": 7.696400609509698, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s004__tb131072__v012\",...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s004__tb131072__v012", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s004__tb131072__v012", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s004__tb262144__v013\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2200.4135732277364, \"nll_base\": 7.696400609509698, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s004__tb262144__v013\",...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s004__tb262144__v013", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s004__tb262144__v013", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s004__tb65536__v014\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2200.4135732277364, \"nll_base\": 7.696400609509698, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s004__tb65536__v014\", \"...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s004__tb65536__v014", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s004__tb65536__v014", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s005__tb131072__v015\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1969.7186156199293, \"nll_base\": 7.58564597681949, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s005__tb131072__v015\", ...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s005__tb131072__v015", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s005__tb131072__v015", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s005__tb262144__v016\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1969.7186156199293, \"nll_base\": 7.58564597681949, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s005__tb262144__v016\", ...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s005__tb262144__v016", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s005__tb262144__v016", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s005__tb65536__v017\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1969.7186156199293, \"nll_base\": 7.58564597681949, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s005__tb65536__v017\", \"o...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s005__tb65536__v017", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s005__tb65536__v017", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s006__tb131072__v018\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3728.3687200238332, \"nll_base\": 8.223726076500439, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s006__tb131072__v018\",...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s006__tb131072__v018", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s006__tb131072__v018", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s006__tb262144__v019\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3728.3687200238332, \"nll_base\": 8.223726076500439, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s006__tb262144__v019\",...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s006__tb262144__v019", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s006__tb262144__v019", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s006__tb65536__v020\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3728.3687200238332, \"nll_base\": 8.223726076500439, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s006__tb65536__v020\", \"...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s006__tb65536__v020", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s006__tb65536__v020", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s007__tb131072__v021\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3411.461228856044, \"nll_base\": 8.134895991353156, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s007__tb131072__v021\", ...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s007__tb131072__v021", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s007__tb131072__v021", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s007__tb262144__v022\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3411.461228856044, \"nll_base\": 8.134895991353156, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s007__tb262144__v022\", ...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s007__tb262144__v022", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s007__tb262144__v022", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s007__tb65536__v023\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3411.461228856044, \"nll_base\": 8.134895991353156, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s007__tb65536__v023\", \"o...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s007__tb65536__v023", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s007__tb65536__v023", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s008__tb131072__v024\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2799.8028851684567, \"nll_base\": 7.937304295531106, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s008__tb131072__v024\",...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s008__tb131072__v024", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s008__tb131072__v024", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s008__tb262144__v025\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2799.8028851684567, \"nll_base\": 7.937304295531106, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s008__tb262144__v025\",...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s008__tb262144__v025", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s008__tb262144__v025", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s008__tb65536__v026\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2799.8028851684567, \"nll_base\": 7.937304295531106, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s008__tb65536__v026\", \"...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s008__tb65536__v026", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s008__tb65536__v026", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s009__tb131072__v027\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1672.6824243965852, \"nll_base\": 7.42218385893723, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s009__tb131072__v027\", ...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s009__tb131072__v027", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s009__tb131072__v027", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s009__tb262144__v028\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1672.6824243965852, \"nll_base\": 7.42218385893723, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s009__tb262144__v028\", ...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s009__tb262144__v028", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s009__tb262144__v028", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s009__tb65536__v029\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1672.6824243965852, \"nll_base\": 7.42218385893723, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s009__tb65536__v029\", \"o...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s009__tb65536__v029", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s009__tb65536__v029", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s010__tb131072__v030\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3299.369838518349, \"nll_base\": 8.101486771195024, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s010__tb131072__v030\", ...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s010__tb131072__v030", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s010__tb131072__v030", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s010__tb262144__v031\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3299.369838518349, \"nll_base\": 8.101486771195024, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s010__tb262144__v031\", ...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s010__tb262144__v031", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s010__tb262144__v031", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s010__tb65536__v032\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3299.369838518349, \"nll_base\": 8.101486771195024, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s010__tb65536__v032\", \"o...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s010__tb65536__v032", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s010__tb65536__v032", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s011__tb131072__v033\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4477.4067904858175, \"nll_base\": 8.40679931640625, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s011__tb131072__v033\", ...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s011__tb131072__v033", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s011__tb131072__v033", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s011__tb262144__v034\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 4477.4067904858175, \"nll_base\": 8.40679931640625, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s011__tb262144__v034\", ...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s011__tb262144__v034", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s011__tb262144__v034", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s011__tb65536__v035\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4477.4067904858175, \"nll_base\": 8.40679931640625, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s011__tb65536__v035\", \"o...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s011__tb65536__v035", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s011__tb65536__v035", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s012__tb131072__v036\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4692.216158601288, \"nll_base\": 8.453660278320312, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s012__tb131072__v036\", ...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s012__tb131072__v036", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s012__tb131072__v036", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s012__tb262144__v037\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 4692.216158601288, \"nll_base\": 8.453660278320312, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s012__tb262144__v037\", ...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s012__tb262144__v037", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s012__tb262144__v037", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s012__tb65536__v038\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4692.216158601288, \"nll_base\": 8.453660278320312, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s012__tb65536__v038\", \"o...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s012__tb65536__v038", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s012__tb65536__v038", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s013__tb131072__v039\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2379.7001723307394, \"nll_base\": 7.77472978071733, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s013__tb131072__v039\", ...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s013__tb131072__v039", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s013__tb131072__v039", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s013__tb262144__v040\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2379.7001723307394, \"nll_base\": 7.77472978071733, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s013__tb262144__v040\", ...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s013__tb262144__v040", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s013__tb262144__v040", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s013__tb65536__v041\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2379.7001723307394, \"nll_base\": 7.77472978071733, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s013__tb65536__v041\", \"o...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s013__tb65536__v041", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s013__tb65536__v041", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s014__tb131072__v042\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1868.570061288748, \"nll_base\": 7.5329287442294035, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s014__tb131072__v042\",...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s014__tb131072__v042", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s014__tb131072__v042", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s014__tb262144__v043\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1868.570061288748, \"nll_base\": 7.5329287442294035, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s014__tb262144__v043\",...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s014__tb262144__v043", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s014__tb262144__v043", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s014__tb65536__v044\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1868.570061288748, \"nll_base\": 7.5329287442294035, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s014__tb65536__v044\", \"...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s014__tb65536__v044", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s014__tb65536__v044", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s015__tb131072__v045\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2052.0758583848624, \"nll_base\": 7.626607173631172, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s015__tb131072__v045\",...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s015__tb131072__v045", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s015__tb131072__v045", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s015__tb262144__v046\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2052.0758583848624, \"nll_base\": 7.626607173631172, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s015__tb262144__v046\",...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s015__tb262144__v046", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s015__tb262144__v046", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s015__tb65536__v047\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2052.0758583848624, \"nll_base\": 7.626607173631172, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s015__tb65536__v047\", \"...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s015__tb65536__v047", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s015__tb65536__v047", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s016__tb131072__v048\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3313.239922503803, \"nll_base\": 8.105681818181818, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s016__tb131072__v048\", ...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s016__tb131072__v048", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s016__tb131072__v048", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s016__tb262144__v049\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3313.239922503803, \"nll_base\": 8.105681818181818, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s016__tb262144__v049\", ...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s016__tb262144__v049", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s016__tb262144__v049", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s016__tb65536__v050\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3313.239922503803, \"nll_base\": 8.105681818181818, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s016__tb65536__v050\", \"o...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s016__tb65536__v050", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s016__tb65536__v050", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s017__tb131072__v051\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4590.609734941979, \"nll_base\": 8.431768134088799, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s017__tb131072__v051\", ...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s017__tb131072__v051", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s017__tb131072__v051", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s017__tb262144__v052\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 4590.609734941979, \"nll_base\": 8.431768134088799, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s017__tb262144__v052\", ...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s017__tb262144__v052", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s017__tb262144__v052", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s017__tb65536__v053\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4590.609734941979, \"nll_base\": 8.431768134088799, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s017__tb65536__v053\", \"o...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s017__tb65536__v053", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s017__tb65536__v053", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s018__tb131072__v054\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3463.7866532006988, \"nll_base\": 8.150117678062939, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s018__tb131072__v054\",...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s018__tb131072__v054", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s018__tb131072__v054", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s018__tb262144__v055\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3463.7866532006988, \"nll_base\": 8.150117678062939, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s018__tb262144__v055\",...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s018__tb262144__v055", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s018__tb262144__v055", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s018__tb65536__v056\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3463.7866532006988, \"nll_base\": 8.150117678062939, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s018__tb65536__v056\", \"...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s018__tb65536__v056", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s018__tb65536__v056", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_es__aecb9487__s019__tb131072__v057\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3330.7554642461128, \"nll_base\": 8.110954423384232, \"eval_ref\": \"lambada_openai_mt_stablelm_es__aecb9487__s019__tb131072__v057\",...
{ "index": "lambada_openai_mt_stablelm_es__aecb9487__s019__tb131072__v057", "uid": "lambada_openai_mt_stablelm_es__aecb9487__s019__tb131072__v057", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_fr__ddbab25c__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 5941.273421611535, \"nll_base\": 8.689678770123106, \"eval_ref\": \"lambada_openai_mt_stablelm_fr__ddbab25c__s000__tb131072__v001\", ...
{ "index": "lambada_openai_mt_stablelm_fr__ddbab25c__s000__tb131072__v001", "uid": "lambada_openai_mt_stablelm_fr__ddbab25c__s000__tb131072__v001", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_fr__ddbab25c__s000__tb262144__v002\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 5941.273421611535, \"nll_base\": 8.689678770123106, \"eval_ref\": \"lambada_openai_mt_stablelm_fr__ddbab25c__s000__tb262144__v002\", ...
{ "index": "lambada_openai_mt_stablelm_fr__ddbab25c__s000__tb262144__v002", "uid": "lambada_openai_mt_stablelm_fr__ddbab25c__s000__tb262144__v002", "axis": "multilingual_ppl", "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\": \"lambada_openai_mt_stablelm_fr__ddbab25c__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 5941.273421611535, \"nll_base\": 8.689678770123106, \"eval_ref\": \"lambada_openai_mt_stablelm_fr__ddbab25c__s000__tb65536__v000\", \"o...
{ "index": "lambada_openai_mt_stablelm_fr__ddbab25c__s000__tb65536__v000", "uid": "lambada_openai_mt_stablelm_fr__ddbab25c__s000__tb65536__v000", "axis": "multilingual_ppl", "budget": 2048, "token_budget": 65536, "variant": "filter" }