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_it__1a798b96__s001__tb65536__v004\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3991.785129772483, \"nll_base\": 8.291993810775432, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s001__tb65536__v004\", \"oracle_delta\": 0.3...
{ "index": "lambada_openai_mt_it__1a798b96__s001__tb65536__v004", "uid": "lambada_openai_mt_it__1a798b96__s001__tb65536__v004", "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_it__1a798b96__s002__tb131072__v008\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1762.1613496790662, \"nll_base\": 7.474296374198718, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s002__tb131072__v008\", \"oracle_delta\":...
{ "index": "lambada_openai_mt_it__1a798b96__s002__tb131072__v008", "uid": "lambada_openai_mt_it__1a798b96__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_it__1a798b96__s002__tb262144__v006\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1762.1613496790662, \"nll_base\": 7.474296374198718, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s002__tb262144__v006\", \"oracle_delta\":...
{ "index": "lambada_openai_mt_it__1a798b96__s002__tb262144__v006", "uid": "lambada_openai_mt_it__1a798b96__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_it__1a798b96__s002__tb65536__v007\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1762.1613496790662, \"nll_base\": 7.474296374198718, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s002__tb65536__v007\", \"oracle_delta\": 0....
{ "index": "lambada_openai_mt_it__1a798b96__s002__tb65536__v007", "uid": "lambada_openai_mt_it__1a798b96__s002__tb65536__v007", "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_it__1a798b96__s003__tb131072__v011\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3255.947649057793, \"nll_base\": 8.08823864869397, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s003__tb131072__v011\", \"oracle_delta\": 0...
{ "index": "lambada_openai_mt_it__1a798b96__s003__tb131072__v011", "uid": "lambada_openai_mt_it__1a798b96__s003__tb131072__v011", "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_it__1a798b96__s003__tb262144__v009\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3255.947649057793, \"nll_base\": 8.08823864869397, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s003__tb262144__v009\", \"oracle_delta\": 1...
{ "index": "lambada_openai_mt_it__1a798b96__s003__tb262144__v009", "uid": "lambada_openai_mt_it__1a798b96__s003__tb262144__v009", "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_it__1a798b96__s003__tb65536__v010\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3255.947649057793, \"nll_base\": 8.08823864869397, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s003__tb65536__v010\", \"oracle_delta\": 0.38...
{ "index": "lambada_openai_mt_it__1a798b96__s003__tb65536__v010", "uid": "lambada_openai_mt_it__1a798b96__s003__tb65536__v010", "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_it__1a798b96__s004__tb131072__v014\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1939.863667688589, \"nll_base\": 7.570372975200688, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s004__tb131072__v014\", \"oracle_delta\": ...
{ "index": "lambada_openai_mt_it__1a798b96__s004__tb131072__v014", "uid": "lambada_openai_mt_it__1a798b96__s004__tb131072__v014", "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_it__1a798b96__s004__tb262144__v012\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1939.863667688589, \"nll_base\": 7.570372975200688, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s004__tb262144__v012\", \"oracle_delta\": ...
{ "index": "lambada_openai_mt_it__1a798b96__s004__tb262144__v012", "uid": "lambada_openai_mt_it__1a798b96__s004__tb262144__v012", "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_it__1a798b96__s004__tb65536__v013\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1939.863667688589, \"nll_base\": 7.570372975200688, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s004__tb65536__v013\", \"oracle_delta\": 0.2...
{ "index": "lambada_openai_mt_it__1a798b96__s004__tb65536__v013", "uid": "lambada_openai_mt_it__1a798b96__s004__tb65536__v013", "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_it__1a798b96__s005__tb131072__v017\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4205.8089049355, \"nll_base\": 8.344221921311211, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s005__tb131072__v017\", \"oracle_delta\": 0....
{ "index": "lambada_openai_mt_it__1a798b96__s005__tb131072__v017", "uid": "lambada_openai_mt_it__1a798b96__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_it__1a798b96__s005__tb262144__v015\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 4205.8089049355, \"nll_base\": 8.344221921311211, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s005__tb262144__v015\", \"oracle_delta\": 1....
{ "index": "lambada_openai_mt_it__1a798b96__s005__tb262144__v015", "uid": "lambada_openai_mt_it__1a798b96__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_it__1a798b96__s005__tb65536__v016\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4205.8089049355, \"nll_base\": 8.344221921311211, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s005__tb65536__v016\", \"oracle_delta\": 0.316...
{ "index": "lambada_openai_mt_it__1a798b96__s005__tb65536__v016", "uid": "lambada_openai_mt_it__1a798b96__s005__tb65536__v016", "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_it__1a798b96__s006__tb131072__v020\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2503.5158665890026, \"nll_base\": 7.825451369512649, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s006__tb131072__v020\", \"oracle_delta\":...
{ "index": "lambada_openai_mt_it__1a798b96__s006__tb131072__v020", "uid": "lambada_openai_mt_it__1a798b96__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_it__1a798b96__s006__tb262144__v018\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2503.5158665890026, \"nll_base\": 7.825451369512649, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s006__tb262144__v018\", \"oracle_delta\":...
{ "index": "lambada_openai_mt_it__1a798b96__s006__tb262144__v018", "uid": "lambada_openai_mt_it__1a798b96__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_it__1a798b96__s006__tb65536__v019\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2503.5158665890026, \"nll_base\": 7.825451369512649, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s006__tb65536__v019\", \"oracle_delta\": 0....
{ "index": "lambada_openai_mt_it__1a798b96__s006__tb65536__v019", "uid": "lambada_openai_mt_it__1a798b96__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_it__1a798b96__s007__tb131072__v023\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 5266.289407020804, \"nll_base\": 8.569081296202956, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s007__tb131072__v023\", \"oracle_delta\": ...
{ "index": "lambada_openai_mt_it__1a798b96__s007__tb131072__v023", "uid": "lambada_openai_mt_it__1a798b96__s007__tb131072__v023", "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_it__1a798b96__s007__tb262144__v021\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 5266.289407020804, \"nll_base\": 8.569081296202956, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s007__tb262144__v021\", \"oracle_delta\": ...
{ "index": "lambada_openai_mt_it__1a798b96__s007__tb262144__v021", "uid": "lambada_openai_mt_it__1a798b96__s007__tb262144__v021", "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_it__1a798b96__s007__tb65536__v022\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 5266.289407020804, \"nll_base\": 8.569081296202956, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s007__tb65536__v022\", \"oracle_delta\": 0.3...
{ "index": "lambada_openai_mt_it__1a798b96__s007__tb65536__v022", "uid": "lambada_openai_mt_it__1a798b96__s007__tb65536__v022", "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_it__1a798b96__s008__tb131072__v026\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3096.4950920485862, \"nll_base\": 8.03802613537721, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s008__tb131072__v026\", \"oracle_delta\": ...
{ "index": "lambada_openai_mt_it__1a798b96__s008__tb131072__v026", "uid": "lambada_openai_mt_it__1a798b96__s008__tb131072__v026", "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_it__1a798b96__s008__tb262144__v024\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3096.4950920485862, \"nll_base\": 8.03802613537721, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s008__tb262144__v024\", \"oracle_delta\": ...
{ "index": "lambada_openai_mt_it__1a798b96__s008__tb262144__v024", "uid": "lambada_openai_mt_it__1a798b96__s008__tb262144__v024", "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_it__1a798b96__s008__tb65536__v025\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3096.4950920485862, \"nll_base\": 8.03802613537721, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s008__tb65536__v025\", \"oracle_delta\": 0.3...
{ "index": "lambada_openai_mt_it__1a798b96__s008__tb65536__v025", "uid": "lambada_openai_mt_it__1a798b96__s008__tb65536__v025", "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_it__1a798b96__s009__tb131072__v029\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 6749.197868396281, \"nll_base\": 8.817178942493557, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s009__tb131072__v029\", \"oracle_delta\": ...
{ "index": "lambada_openai_mt_it__1a798b96__s009__tb131072__v029", "uid": "lambada_openai_mt_it__1a798b96__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_it__1a798b96__s009__tb262144__v027\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 6749.197868396281, \"nll_base\": 8.817178942493557, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s009__tb262144__v027\", \"oracle_delta\": ...
{ "index": "lambada_openai_mt_it__1a798b96__s009__tb262144__v027", "uid": "lambada_openai_mt_it__1a798b96__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_it__1a798b96__s009__tb65536__v028\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 6749.197868396281, \"nll_base\": 8.817178942493557, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s009__tb65536__v028\", \"oracle_delta\": 0.2...
{ "index": "lambada_openai_mt_it__1a798b96__s009__tb65536__v028", "uid": "lambada_openai_mt_it__1a798b96__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_it__1a798b96__s010__tb131072__v032\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 5163.9938896058475, \"nll_base\": 8.549465568698182, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s010__tb131072__v032\", \"oracle_delta\":...
{ "index": "lambada_openai_mt_it__1a798b96__s010__tb131072__v032", "uid": "lambada_openai_mt_it__1a798b96__s010__tb131072__v032", "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_it__1a798b96__s010__tb262144__v030\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 5163.9938896058475, \"nll_base\": 8.549465568698182, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s010__tb262144__v030\", \"oracle_delta\":...
{ "index": "lambada_openai_mt_it__1a798b96__s010__tb262144__v030", "uid": "lambada_openai_mt_it__1a798b96__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_it__1a798b96__s010__tb65536__v031\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 5163.9938896058475, \"nll_base\": 8.549465568698182, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s010__tb65536__v031\", \"oracle_delta\": 0....
{ "index": "lambada_openai_mt_it__1a798b96__s010__tb65536__v031", "uid": "lambada_openai_mt_it__1a798b96__s010__tb65536__v031", "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_it__1a798b96__s011__tb131072__v035\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2635.1672774321773, \"nll_base\": 7.876701941856971, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s011__tb131072__v035\", \"oracle_delta\":...
{ "index": "lambada_openai_mt_it__1a798b96__s011__tb131072__v035", "uid": "lambada_openai_mt_it__1a798b96__s011__tb131072__v035", "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_it__1a798b96__s011__tb262144__v033\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2635.1672774321773, \"nll_base\": 7.876701941856971, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s011__tb262144__v033\", \"oracle_delta\":...
{ "index": "lambada_openai_mt_it__1a798b96__s011__tb262144__v033", "uid": "lambada_openai_mt_it__1a798b96__s011__tb262144__v033", "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_it__1a798b96__s011__tb65536__v034\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2635.1672774321773, \"nll_base\": 7.876701941856971, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s011__tb65536__v034\", \"oracle_delta\": 0....
{ "index": "lambada_openai_mt_it__1a798b96__s011__tb65536__v034", "uid": "lambada_openai_mt_it__1a798b96__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_it__1a798b96__s012__tb131072__v038\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2902.099566520956, \"nll_base\": 7.973189742476852, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s012__tb131072__v038\", \"oracle_delta\": ...
{ "index": "lambada_openai_mt_it__1a798b96__s012__tb131072__v038", "uid": "lambada_openai_mt_it__1a798b96__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_it__1a798b96__s012__tb262144__v036\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2902.099566520956, \"nll_base\": 7.973189742476852, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s012__tb262144__v036\", \"oracle_delta\": ...
{ "index": "lambada_openai_mt_it__1a798b96__s012__tb262144__v036", "uid": "lambada_openai_mt_it__1a798b96__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_it__1a798b96__s012__tb65536__v037\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2902.099566520956, \"nll_base\": 7.973189742476852, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s012__tb65536__v037\", \"oracle_delta\": 0.2...
{ "index": "lambada_openai_mt_it__1a798b96__s012__tb65536__v037", "uid": "lambada_openai_mt_it__1a798b96__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_it__1a798b96__s013__tb131072__v041\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 6224.7869346600455, \"nll_base\": 8.73629449338329, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s013__tb131072__v041\", \"oracle_delta\": ...
{ "index": "lambada_openai_mt_it__1a798b96__s013__tb131072__v041", "uid": "lambada_openai_mt_it__1a798b96__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_it__1a798b96__s013__tb262144__v039\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 6224.7869346600455, \"nll_base\": 8.73629449338329, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s013__tb262144__v039\", \"oracle_delta\": ...
{ "index": "lambada_openai_mt_it__1a798b96__s013__tb262144__v039", "uid": "lambada_openai_mt_it__1a798b96__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_it__1a798b96__s013__tb65536__v040\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 6224.7869346600455, \"nll_base\": 8.73629449338329, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s013__tb65536__v040\", \"oracle_delta\": 0.3...
{ "index": "lambada_openai_mt_it__1a798b96__s013__tb65536__v040", "uid": "lambada_openai_mt_it__1a798b96__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_it__1a798b96__s014__tb131072__v044\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3974.1501076402033, \"nll_base\": 8.287566194829253, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s014__tb131072__v044\", \"oracle_delta\":...
{ "index": "lambada_openai_mt_it__1a798b96__s014__tb131072__v044", "uid": "lambada_openai_mt_it__1a798b96__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_it__1a798b96__s014__tb262144__v042\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3974.1501076402033, \"nll_base\": 8.287566194829253, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s014__tb262144__v042\", \"oracle_delta\":...
{ "index": "lambada_openai_mt_it__1a798b96__s014__tb262144__v042", "uid": "lambada_openai_mt_it__1a798b96__s014__tb262144__v042", "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_it__1a798b96__s014__tb65536__v043\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3974.1501076402033, \"nll_base\": 8.287566194829253, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s014__tb65536__v043\", \"oracle_delta\": 0....
{ "index": "lambada_openai_mt_it__1a798b96__s014__tb65536__v043", "uid": "lambada_openai_mt_it__1a798b96__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_it__1a798b96__s015__tb131072__v047\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2137.702321571075, \"nll_base\": 7.667486849698153, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s015__tb131072__v047\", \"oracle_delta\": ...
{ "index": "lambada_openai_mt_it__1a798b96__s015__tb131072__v047", "uid": "lambada_openai_mt_it__1a798b96__s015__tb131072__v047", "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_it__1a798b96__s015__tb262144__v045\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2137.702321571075, \"nll_base\": 7.667486849698153, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s015__tb262144__v045\", \"oracle_delta\": ...
{ "index": "lambada_openai_mt_it__1a798b96__s015__tb262144__v045", "uid": "lambada_openai_mt_it__1a798b96__s015__tb262144__v045", "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_it__1a798b96__s015__tb65536__v046\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2137.702321571075, \"nll_base\": 7.667486849698153, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s015__tb65536__v046\", \"oracle_delta\": 0.2...
{ "index": "lambada_openai_mt_it__1a798b96__s015__tb65536__v046", "uid": "lambada_openai_mt_it__1a798b96__s015__tb65536__v046", "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_it__1a798b96__s016__tb131072__v050\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3724.038910480974, \"nll_base\": 8.222564086914062, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s016__tb131072__v050\", \"oracle_delta\": ...
{ "index": "lambada_openai_mt_it__1a798b96__s016__tb131072__v050", "uid": "lambada_openai_mt_it__1a798b96__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_it__1a798b96__s016__tb262144__v048\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3724.038910480974, \"nll_base\": 8.222564086914062, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s016__tb262144__v048\", \"oracle_delta\": ...
{ "index": "lambada_openai_mt_it__1a798b96__s016__tb262144__v048", "uid": "lambada_openai_mt_it__1a798b96__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_it__1a798b96__s016__tb65536__v049\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3724.038910480974, \"nll_base\": 8.222564086914062, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s016__tb65536__v049\", \"oracle_delta\": 0.3...
{ "index": "lambada_openai_mt_it__1a798b96__s016__tb65536__v049", "uid": "lambada_openai_mt_it__1a798b96__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_it__1a798b96__s017__tb131072__v053\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2650.8474927035113, \"nll_base\": 7.882634676419771, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s017__tb131072__v053\", \"oracle_delta\":...
{ "index": "lambada_openai_mt_it__1a798b96__s017__tb131072__v053", "uid": "lambada_openai_mt_it__1a798b96__s017__tb131072__v053", "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_it__1a798b96__s017__tb262144__v051\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2650.8474927035113, \"nll_base\": 7.882634676419771, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s017__tb262144__v051\", \"oracle_delta\":...
{ "index": "lambada_openai_mt_it__1a798b96__s017__tb262144__v051", "uid": "lambada_openai_mt_it__1a798b96__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_it__1a798b96__s017__tb65536__v052\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2650.8474927035113, \"nll_base\": 7.882634676419771, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s017__tb65536__v052\", \"oracle_delta\": 0....
{ "index": "lambada_openai_mt_it__1a798b96__s017__tb65536__v052", "uid": "lambada_openai_mt_it__1a798b96__s017__tb65536__v052", "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_it__1a798b96__s018__tb131072__v056\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4521.029129788038, \"nll_base\": 8.416494930491728, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s018__tb131072__v056\", \"oracle_delta\": ...
{ "index": "lambada_openai_mt_it__1a798b96__s018__tb131072__v056", "uid": "lambada_openai_mt_it__1a798b96__s018__tb131072__v056", "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_it__1a798b96__s018__tb262144__v054\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 4521.029129788038, \"nll_base\": 8.416494930491728, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s018__tb262144__v054\", \"oracle_delta\": ...
{ "index": "lambada_openai_mt_it__1a798b96__s018__tb262144__v054", "uid": "lambada_openai_mt_it__1a798b96__s018__tb262144__v054", "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_it__1a798b96__s018__tb65536__v055\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4521.029129788038, \"nll_base\": 8.416494930491728, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s018__tb65536__v055\", \"oracle_delta\": 0.2...
{ "index": "lambada_openai_mt_it__1a798b96__s018__tb65536__v055", "uid": "lambada_openai_mt_it__1a798b96__s018__tb65536__v055", "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_it__1a798b96__s019__tb262144__v057\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3273.6787350047675, \"nll_base\": 8.093669626972464, \"eval_ref\": \"lambada_openai_mt_it__1a798b96__s019__tb262144__v057\", \"oracle_delta\":...
{ "index": "lambada_openai_mt_it__1a798b96__s019__tb262144__v057", "uid": "lambada_openai_mt_it__1a798b96__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_de__11084541__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2631.784859833331, \"nll_base\": 7.875417548919392, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s000__tb131072__v001\", ...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s000__tb131072__v001", "uid": "lambada_openai_mt_stablelm_de__11084541__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_de__11084541__s000__tb262144__v002\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2631.784859833331, \"nll_base\": 7.875417548919392, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s000__tb262144__v002\", ...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s000__tb262144__v002", "uid": "lambada_openai_mt_stablelm_de__11084541__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_de__11084541__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2631.784859833331, \"nll_base\": 7.875417548919392, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s000__tb65536__v000\", \"o...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s000__tb65536__v000", "uid": "lambada_openai_mt_stablelm_de__11084541__s000__tb65536__v000", "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__s001__tb131072__v004\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1668.0611854472234, \"nll_base\": 7.419417264168723, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s001__tb131072__v004\",...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s001__tb131072__v004", "uid": "lambada_openai_mt_stablelm_de__11084541__s001__tb131072__v004", "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__s001__tb262144__v005\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1668.0611854472234, \"nll_base\": 7.419417264168723, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s001__tb262144__v005\",...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s001__tb262144__v005", "uid": "lambada_openai_mt_stablelm_de__11084541__s001__tb262144__v005", "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__s001__tb65536__v003\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1668.0611854472234, \"nll_base\": 7.419417264168723, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s001__tb65536__v003\", \"...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s001__tb65536__v003", "uid": "lambada_openai_mt_stablelm_de__11084541__s001__tb65536__v003", "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__s002__tb131072__v007\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2771.14791757082, \"nll_base\": 7.927016924012382, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s002__tb131072__v007\", \...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s002__tb131072__v007", "uid": "lambada_openai_mt_stablelm_de__11084541__s002__tb131072__v007", "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__s002__tb262144__v008\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2771.14791757082, \"nll_base\": 7.927016924012382, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s002__tb262144__v008\", \...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s002__tb262144__v008", "uid": "lambada_openai_mt_stablelm_de__11084541__s002__tb262144__v008", "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__s002__tb65536__v006\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2771.14791757082, \"nll_base\": 7.927016924012382, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s002__tb65536__v006\", \"or...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s002__tb65536__v006", "uid": "lambada_openai_mt_stablelm_de__11084541__s002__tb65536__v006", "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__s003__tb131072__v010\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1913.8327012169113, \"nll_base\": 7.556863160259955, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s003__tb131072__v010\",...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s003__tb131072__v010", "uid": "lambada_openai_mt_stablelm_de__11084541__s003__tb131072__v010", "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__s003__tb262144__v011\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1913.8327012169113, \"nll_base\": 7.556863160259955, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s003__tb262144__v011\",...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s003__tb262144__v011", "uid": "lambada_openai_mt_stablelm_de__11084541__s003__tb262144__v011", "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__s003__tb65536__v009\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1913.8327012169113, \"nll_base\": 7.556863160259955, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s003__tb65536__v009\", \"...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s003__tb65536__v009", "uid": "lambada_openai_mt_stablelm_de__11084541__s003__tb65536__v009", "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__s004__tb131072__v013\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 6349.812919364363, \"nll_base\": 8.756180629935315, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s004__tb131072__v013\", ...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s004__tb131072__v013", "uid": "lambada_openai_mt_stablelm_de__11084541__s004__tb131072__v013", "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__s004__tb262144__v014\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 6349.812919364363, \"nll_base\": 8.756180629935315, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s004__tb262144__v014\", ...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s004__tb262144__v014", "uid": "lambada_openai_mt_stablelm_de__11084541__s004__tb262144__v014", "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__s004__tb65536__v012\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 6349.812919364363, \"nll_base\": 8.756180629935315, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s004__tb65536__v012\", \"o...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s004__tb65536__v012", "uid": "lambada_openai_mt_stablelm_de__11084541__s004__tb65536__v012", "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__s005__tb131072__v016\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2940.4373475187804, \"nll_base\": 7.986313606928853, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s005__tb131072__v016\",...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s005__tb131072__v016", "uid": "lambada_openai_mt_stablelm_de__11084541__s005__tb131072__v016", "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__s005__tb262144__v017\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2940.4373475187804, \"nll_base\": 7.986313606928853, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s005__tb262144__v017\",...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s005__tb262144__v017", "uid": "lambada_openai_mt_stablelm_de__11084541__s005__tb262144__v017", "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__s005__tb65536__v015\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2940.4373475187804, \"nll_base\": 7.986313606928853, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s005__tb65536__v015\", \"...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s005__tb65536__v015", "uid": "lambada_openai_mt_stablelm_de__11084541__s005__tb65536__v015", "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__s006__tb131072__v019\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1582.04963536741, \"nll_base\": 7.366476522909628, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s006__tb131072__v019\", \...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s006__tb131072__v019", "uid": "lambada_openai_mt_stablelm_de__11084541__s006__tb131072__v019", "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__s006__tb262144__v020\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1582.04963536741, \"nll_base\": 7.366476522909628, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s006__tb262144__v020\", \...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s006__tb262144__v020", "uid": "lambada_openai_mt_stablelm_de__11084541__s006__tb262144__v020", "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__s006__tb65536__v018\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1582.04963536741, \"nll_base\": 7.366476522909628, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s006__tb65536__v018\", \"or...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s006__tb65536__v018", "uid": "lambada_openai_mt_stablelm_de__11084541__s006__tb65536__v018", "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__s007__tb131072__v022\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1558.2416541693324, \"nll_base\": 7.351313319774943, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s007__tb131072__v022\",...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s007__tb131072__v022", "uid": "lambada_openai_mt_stablelm_de__11084541__s007__tb131072__v022", "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__s007__tb262144__v023\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1558.2416541693324, \"nll_base\": 7.351313319774943, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s007__tb262144__v023\",...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s007__tb262144__v023", "uid": "lambada_openai_mt_stablelm_de__11084541__s007__tb262144__v023", "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__s007__tb65536__v021\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1558.2416541693324, \"nll_base\": 7.351313319774943, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s007__tb65536__v021\", \"...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s007__tb65536__v021", "uid": "lambada_openai_mt_stablelm_de__11084541__s007__tb65536__v021", "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__s008__tb131072__v025\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1465.913396528774, \"nll_base\": 7.290233806028205, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s008__tb131072__v025\", ...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s008__tb131072__v025", "uid": "lambada_openai_mt_stablelm_de__11084541__s008__tb131072__v025", "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__s008__tb262144__v026\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1465.913396528774, \"nll_base\": 7.290233806028205, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s008__tb262144__v026\", ...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s008__tb262144__v026", "uid": "lambada_openai_mt_stablelm_de__11084541__s008__tb262144__v026", "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__s008__tb65536__v024\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1465.913396528774, \"nll_base\": 7.290233806028205, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s008__tb65536__v024\", \"o...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s008__tb65536__v024", "uid": "lambada_openai_mt_stablelm_de__11084541__s008__tb65536__v024", "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__s009__tb131072__v028\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3367.220112014925, \"nll_base\": 8.121842790358137, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s009__tb131072__v028\", ...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s009__tb131072__v028", "uid": "lambada_openai_mt_stablelm_de__11084541__s009__tb131072__v028", "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__s009__tb262144__v029\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3367.220112014925, \"nll_base\": 8.121842790358137, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s009__tb262144__v029\", ...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s009__tb262144__v029", "uid": "lambada_openai_mt_stablelm_de__11084541__s009__tb262144__v029", "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__s009__tb65536__v027\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3367.220112014925, \"nll_base\": 8.121842790358137, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s009__tb65536__v027\", \"o...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s009__tb65536__v027", "uid": "lambada_openai_mt_stablelm_de__11084541__s009__tb65536__v027", "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__s010__tb131072__v031\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4941.509280171577, \"nll_base\": 8.505426085821473, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s010__tb131072__v031\", ...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s010__tb131072__v031", "uid": "lambada_openai_mt_stablelm_de__11084541__s010__tb131072__v031", "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__s010__tb262144__v032\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 4941.509280171577, \"nll_base\": 8.505426085821473, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s010__tb262144__v032\", ...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s010__tb262144__v032", "uid": "lambada_openai_mt_stablelm_de__11084541__s010__tb262144__v032", "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__s010__tb65536__v030\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4941.509280171577, \"nll_base\": 8.505426085821473, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s010__tb65536__v030\", \"o...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s010__tb65536__v030", "uid": "lambada_openai_mt_stablelm_de__11084541__s010__tb65536__v030", "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__s011__tb131072__v034\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 7578.670598157369, \"nll_base\": 8.933093080426207, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s011__tb131072__v034\", ...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s011__tb131072__v034", "uid": "lambada_openai_mt_stablelm_de__11084541__s011__tb131072__v034", "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__s011__tb262144__v035\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 7578.670598157369, \"nll_base\": 8.933093080426207, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s011__tb262144__v035\", ...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s011__tb262144__v035", "uid": "lambada_openai_mt_stablelm_de__11084541__s011__tb262144__v035", "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__s011__tb65536__v033\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 7578.670598157369, \"nll_base\": 8.933093080426207, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s011__tb65536__v033\", \"o...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s011__tb65536__v033", "uid": "lambada_openai_mt_stablelm_de__11084541__s011__tb65536__v033", "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__s012__tb131072__v037\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2352.209570320574, \"nll_base\": 7.763110408076534, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s012__tb131072__v037\", ...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s012__tb131072__v037", "uid": "lambada_openai_mt_stablelm_de__11084541__s012__tb131072__v037", "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__s012__tb262144__v038\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2352.209570320574, \"nll_base\": 7.763110408076534, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s012__tb262144__v038\", ...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s012__tb262144__v038", "uid": "lambada_openai_mt_stablelm_de__11084541__s012__tb262144__v038", "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__s012__tb65536__v036\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2352.209570320574, \"nll_base\": 7.763110408076534, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s012__tb65536__v036\", \"o...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s012__tb65536__v036", "uid": "lambada_openai_mt_stablelm_de__11084541__s012__tb65536__v036", "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__s013__tb131072__v040\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1256.7797819619727, \"nll_base\": 7.136307999894425, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s013__tb131072__v040\",...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s013__tb131072__v040", "uid": "lambada_openai_mt_stablelm_de__11084541__s013__tb131072__v040", "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__s013__tb262144__v041\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1256.7797819619727, \"nll_base\": 7.136307999894425, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s013__tb262144__v041\",...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s013__tb262144__v041", "uid": "lambada_openai_mt_stablelm_de__11084541__s013__tb262144__v041", "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__s013__tb65536__v039\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1256.7797819619727, \"nll_base\": 7.136307999894425, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s013__tb65536__v039\", \"...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s013__tb65536__v039", "uid": "lambada_openai_mt_stablelm_de__11084541__s013__tb65536__v039", "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__s014__tb131072__v043\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3077.198378865719, \"nll_base\": 8.031774844763413, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s014__tb131072__v043\", ...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s014__tb131072__v043", "uid": "lambada_openai_mt_stablelm_de__11084541__s014__tb131072__v043", "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__s014__tb262144__v044\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3077.198378865719, \"nll_base\": 8.031774844763413, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s014__tb262144__v044\", ...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s014__tb262144__v044", "uid": "lambada_openai_mt_stablelm_de__11084541__s014__tb262144__v044", "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__s014__tb65536__v042\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3077.198378865719, \"nll_base\": 8.031774844763413, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s014__tb65536__v042\", \"o...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s014__tb65536__v042", "uid": "lambada_openai_mt_stablelm_de__11084541__s014__tb65536__v042", "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__s015__tb131072__v046\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2035.3998442646714, \"nll_base\": 7.618447562244451, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s015__tb131072__v046\",...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s015__tb131072__v046", "uid": "lambada_openai_mt_stablelm_de__11084541__s015__tb131072__v046", "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__s015__tb262144__v047\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2035.3998442646714, \"nll_base\": 7.618447562244451, \"eval_ref\": \"lambada_openai_mt_stablelm_de__11084541__s015__tb262144__v047\",...
{ "index": "lambada_openai_mt_stablelm_de__11084541__s015__tb262144__v047", "uid": "lambada_openai_mt_stablelm_de__11084541__s015__tb262144__v047", "axis": "multilingual_ppl", "budget": 2048, "token_budget": 262144, "variant": "filter" }