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\": \"bigbench_odd_one_out_multiple_choice__b4135a10__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 12967.5265860091, \"nll_base\": 9.470203556426585, \"eval_ref\": \"bigbench_odd_one_out_multiple_choice__b4135a10__s000__tb65536...
{ "index": "bigbench_odd_one_out_multiple_choice__b4135a10__s000__tb65536__v000", "uid": "bigbench_odd_one_out_multiple_choice__b4135a10__s000__tb65536__v000", "axis": "reasoning", "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\": \"bigbench_operators_generate_until__c4b5a2af__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 134.96357656436382, \"nll_base\": 4.905004938808056, \"eval_ref\": \"bigbench_operators_generate_until__c4b5a2af__s000__tb131072_...
{ "index": "bigbench_operators_generate_until__c4b5a2af__s000__tb131072__v000", "uid": "bigbench_operators_generate_until__c4b5a2af__s000__tb131072__v000", "axis": "reasoning", "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\": \"bigbench_parsinlu_reading_comprehension_generate_until__10af368e__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2.7284202809600755, \"nll_base\": 1.003722790050406, \"eval_ref\": \"bigbench_parsinlu_reading_comprehension...
{ "index": "bigbench_parsinlu_reading_comprehension_generate_until__10af368e__s000__tb131072__v001", "uid": "bigbench_parsinlu_reading_comprehension_generate_until__10af368e__s000__tb131072__v001", "axis": "reasoning", "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\": \"bigbench_penguins_in_a_table_generate_until__72eeac5c__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 282.32055256561966, \"nll_base\": 5.643043136596679, \"eval_ref\": \"bigbench_penguins_in_a_table_generate_until__72eea...
{ "index": "bigbench_penguins_in_a_table_generate_until__72eeac5c__s000__tb131072__v001", "uid": "bigbench_penguins_in_a_table_generate_until__72eeac5c__s000__tb131072__v001", "axis": "reasoning", "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\": \"bigbench_penguins_in_a_table_generate_until__72eeac5c__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 282.32055256561966, \"nll_base\": 5.643043136596679, \"eval_ref\": \"bigbench_penguins_in_a_table_generate_until__72eeac5...
{ "index": "bigbench_penguins_in_a_table_generate_until__72eeac5c__s000__tb65536__v000", "uid": "bigbench_penguins_in_a_table_generate_until__72eeac5c__s000__tb65536__v000", "axis": "reasoning", "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\": \"bigbench_penguins_in_a_table_multiple_choice__ce3dac51__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 217.61921111295283, \"nll_base\": 5.382746797449449, \"eval_ref\": \"bigbench_penguins_in_a_table_multiple_choice__ce3...
{ "index": "bigbench_penguins_in_a_table_multiple_choice__ce3dac51__s000__tb131072__v001", "uid": "bigbench_penguins_in_a_table_multiple_choice__ce3dac51__s000__tb131072__v001", "axis": "reasoning", "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\": \"bigbench_penguins_in_a_table_multiple_choice__ce3dac51__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 217.61921111295283, \"nll_base\": 5.382746797449449, \"eval_ref\": \"bigbench_penguins_in_a_table_multiple_choice__ce3da...
{ "index": "bigbench_penguins_in_a_table_multiple_choice__ce3dac51__s000__tb65536__v000", "uid": "bigbench_penguins_in_a_table_multiple_choice__ce3dac51__s000__tb65536__v000", "axis": "reasoning", "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\": \"bigbench_periodic_elements_generate_until__07ee2486__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2525.0344570362886, \"nll_base\": 7.834009987967355, \"eval_ref\": \"bigbench_periodic_elements_generate_until__07ee2486_...
{ "index": "bigbench_periodic_elements_generate_until__07ee2486__s000__tb131072__v001", "uid": "bigbench_periodic_elements_generate_until__07ee2486__s000__tb131072__v001", "axis": "reasoning", "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\": \"bigbench_periodic_elements_generate_until__07ee2486__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2525.0344570362886, \"nll_base\": 7.834009987967355, \"eval_ref\": \"bigbench_periodic_elements_generate_until__07ee2486__s...
{ "index": "bigbench_periodic_elements_generate_until__07ee2486__s000__tb65536__v000", "uid": "bigbench_periodic_elements_generate_until__07ee2486__s000__tb65536__v000", "axis": "reasoning", "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\": \"bigbench_phrase_relatedness_generate_until__21c74fc2__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 160.55525438804773, \"nll_base\": 5.078638147424768, \"eval_ref\": \"bigbench_phrase_relatedness_generate_until__21c74fc...
{ "index": "bigbench_phrase_relatedness_generate_until__21c74fc2__s000__tb131072__v001", "uid": "bigbench_phrase_relatedness_generate_until__21c74fc2__s000__tb131072__v001", "axis": "reasoning", "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\": \"bigbench_phrase_relatedness_generate_until__21c74fc2__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 160.55525438804773, \"nll_base\": 5.078638147424768, \"eval_ref\": \"bigbench_phrase_relatedness_generate_until__21c74fc2_...
{ "index": "bigbench_phrase_relatedness_generate_until__21c74fc2__s000__tb65536__v000", "uid": "bigbench_phrase_relatedness_generate_until__21c74fc2__s000__tb65536__v000", "axis": "reasoning", "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\": \"bigbench_phrase_relatedness_multiple_choice__75fa4c7d__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 83.25568208185886, \"nll_base\": 4.421916379766949, \"eval_ref\": \"bigbench_phrase_relatedness_multiple_choice__75fa4c...
{ "index": "bigbench_phrase_relatedness_multiple_choice__75fa4c7d__s000__tb131072__v001", "uid": "bigbench_phrase_relatedness_multiple_choice__75fa4c7d__s000__tb131072__v001", "axis": "reasoning", "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\": \"bigbench_phrase_relatedness_multiple_choice__75fa4c7d__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 83.25568208185886, \"nll_base\": 4.421916379766949, \"eval_ref\": \"bigbench_phrase_relatedness_multiple_choice__75fa4c7d...
{ "index": "bigbench_phrase_relatedness_multiple_choice__75fa4c7d__s000__tb65536__v000", "uid": "bigbench_phrase_relatedness_multiple_choice__75fa4c7d__s000__tb65536__v000", "axis": "reasoning", "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\": \"bigbench_physical_intuition_generate_until__8c2a63a0__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 114.64630802410002, \"nll_base\": 4.741851806640625, \"eval_ref\": \"bigbench_physical_intuition_generate_until__8c2a63a...
{ "index": "bigbench_physical_intuition_generate_until__8c2a63a0__s000__tb131072__v001", "uid": "bigbench_physical_intuition_generate_until__8c2a63a0__s000__tb131072__v001", "axis": "reasoning", "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\": \"bigbench_physical_intuition_generate_until__8c2a63a0__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 114.64630802410002, \"nll_base\": 4.741851806640625, \"eval_ref\": \"bigbench_physical_intuition_generate_until__8c2a63a0_...
{ "index": "bigbench_physical_intuition_generate_until__8c2a63a0__s000__tb65536__v000", "uid": "bigbench_physical_intuition_generate_until__8c2a63a0__s000__tb65536__v000", "axis": "reasoning", "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\": \"bigbench_physical_intuition_multiple_choice__104ff474__s000__tb131072__v002\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 111.07587531051072, \"nll_base\": 4.7102135291466345, \"eval_ref\": \"bigbench_physical_intuition_multiple_choice__104f...
{ "index": "bigbench_physical_intuition_multiple_choice__104ff474__s000__tb131072__v002", "uid": "bigbench_physical_intuition_multiple_choice__104ff474__s000__tb131072__v002", "axis": "reasoning", "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\": \"bigbench_physical_intuition_multiple_choice__104ff474__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 111.07587531051072, \"nll_base\": 4.7102135291466345, \"eval_ref\": \"bigbench_physical_intuition_multiple_choice__104ff4...
{ "index": "bigbench_physical_intuition_multiple_choice__104ff474__s000__tb65536__v001", "uid": "bigbench_physical_intuition_multiple_choice__104ff474__s000__tb65536__v001", "axis": "reasoning", "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\": \"bigbench_physics_generate_until__de6b0dd4__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 5.324444195219297, \"nll_base\": 1.6723083296632455, \"eval_ref\": \"bigbench_physics_generate_until__de6b0dd4__s000__tb131072__v00...
{ "index": "bigbench_physics_generate_until__de6b0dd4__s000__tb131072__v000", "uid": "bigbench_physics_generate_until__de6b0dd4__s000__tb131072__v000", "axis": "reasoning", "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\": \"bigbench_physics_generate_until__de6b0dd4__s000__tb65536__v002\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 5.324444195219297, \"nll_base\": 1.6723083296632455, \"eval_ref\": \"bigbench_physics_generate_until__de6b0dd4__s000__tb65536__v002\"...
{ "index": "bigbench_physics_generate_until__de6b0dd4__s000__tb65536__v002", "uid": "bigbench_physics_generate_until__de6b0dd4__s000__tb65536__v002", "axis": "reasoning", "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\": \"bigbench_physics_questions_generate_until__462db48a__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 34.884883514505304, \"nll_base\": 3.5520535982572117, \"eval_ref\": \"bigbench_physics_questions_generate_until__462db48a...
{ "index": "bigbench_physics_questions_generate_until__462db48a__s000__tb131072__v000", "uid": "bigbench_physics_questions_generate_until__462db48a__s000__tb131072__v000", "axis": "reasoning", "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\": \"bigbench_play_dialog_same_or_different_multiple_choice__8de73f57__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 66701.65336542574, \"nll_base\": 11.107985019683838, \"eval_ref\": \"bigbench_play_dialog_same_or_different_...
{ "index": "bigbench_play_dialog_same_or_different_multiple_choice__8de73f57__s000__tb131072__v000", "uid": "bigbench_play_dialog_same_or_different_multiple_choice__8de73f57__s000__tb131072__v000", "axis": "reasoning", "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\": \"bigbench_polish_sequence_labeling_generate_until__d95dab22__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 26.706215792408745, \"nll_base\": 3.284896339526017, \"eval_ref\": \"bigbench_polish_sequence_labeling_generate_un...
{ "index": "bigbench_polish_sequence_labeling_generate_until__d95dab22__s000__tb131072__v000", "uid": "bigbench_polish_sequence_labeling_generate_until__d95dab22__s000__tb131072__v000", "axis": "reasoning", "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\": \"bigbench_polish_sequence_labeling_generate_until__d95dab22__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 26.706215792408745, \"nll_base\": 3.284896339526017, \"eval_ref\": \"bigbench_polish_sequence_labeling_generate_un...
{ "index": "bigbench_polish_sequence_labeling_generate_until__d95dab22__s000__tb262144__v001", "uid": "bigbench_polish_sequence_labeling_generate_until__d95dab22__s000__tb262144__v001", "axis": "reasoning", "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\": \"bigbench_polish_sequence_labeling_generate_until__d95dab22__s000__tb65536__v002\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 26.706215792408745, \"nll_base\": 3.284896339526017, \"eval_ref\": \"bigbench_polish_sequence_labeling_generate_unti...
{ "index": "bigbench_polish_sequence_labeling_generate_until__d95dab22__s000__tb65536__v002", "uid": "bigbench_polish_sequence_labeling_generate_until__d95dab22__s000__tb65536__v002", "axis": "reasoning", "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\": \"bigbench_presuppositions_as_nli_generate_until__f3e7e8bf__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 560.24080443322, \"nll_base\": 6.32836669921875, \"eval_ref\": \"bigbench_presuppositions_as_nli_generate_until__f3e...
{ "index": "bigbench_presuppositions_as_nli_generate_until__f3e7e8bf__s000__tb131072__v001", "uid": "bigbench_presuppositions_as_nli_generate_until__f3e7e8bf__s000__tb131072__v001", "axis": "reasoning", "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\": \"bigbench_presuppositions_as_nli_generate_until__f3e7e8bf__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 560.24080443322, \"nll_base\": 6.32836669921875, \"eval_ref\": \"bigbench_presuppositions_as_nli_generate_until__f3e7e...
{ "index": "bigbench_presuppositions_as_nli_generate_until__f3e7e8bf__s000__tb65536__v000", "uid": "bigbench_presuppositions_as_nli_generate_until__f3e7e8bf__s000__tb65536__v000", "axis": "reasoning", "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\": \"bigbench_presuppositions_as_nli_multiple_choice__76027c99__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 494.1817175160119, \"nll_base\": 6.20290329877068, \"eval_ref\": \"bigbench_presuppositions_as_nli_multiple_choice_...
{ "index": "bigbench_presuppositions_as_nli_multiple_choice__76027c99__s000__tb131072__v001", "uid": "bigbench_presuppositions_as_nli_multiple_choice__76027c99__s000__tb131072__v001", "axis": "reasoning", "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\": \"bigbench_presuppositions_as_nli_multiple_choice__76027c99__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 494.1817175160119, \"nll_base\": 6.20290329877068, \"eval_ref\": \"bigbench_presuppositions_as_nli_multiple_choice__7...
{ "index": "bigbench_presuppositions_as_nli_multiple_choice__76027c99__s000__tb65536__v000", "uid": "bigbench_presuppositions_as_nli_multiple_choice__76027c99__s000__tb65536__v000", "axis": "reasoning", "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\": \"bigbench_qa_wikidata_generate_until__c1e6dea2__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 463.2553017998411, \"nll_base\": 6.138278309891864, \"eval_ref\": \"bigbench_qa_wikidata_generate_until__c1e6dea2__s000__tb2621...
{ "index": "bigbench_qa_wikidata_generate_until__c1e6dea2__s000__tb262144__v000", "uid": "bigbench_qa_wikidata_generate_until__c1e6dea2__s000__tb262144__v000", "axis": "reasoning", "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\": \"bigbench_qa_wikidata_generate_until__c1e6dea2__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 463.2553017998411, \"nll_base\": 6.138278309891864, \"eval_ref\": \"bigbench_qa_wikidata_generate_until__c1e6dea2__s000__tb65536_...
{ "index": "bigbench_qa_wikidata_generate_until__c1e6dea2__s000__tb65536__v001", "uid": "bigbench_qa_wikidata_generate_until__c1e6dea2__s000__tb65536__v001", "axis": "reasoning", "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\": \"bigbench_question_selection_generate_until__d1253ae1__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2.7042968760859494, \"nll_base\": 0.9948419436059817, \"eval_ref\": \"bigbench_question_selection_generate_until__d1253a...
{ "index": "bigbench_question_selection_generate_until__d1253ae1__s000__tb131072__v001", "uid": "bigbench_question_selection_generate_until__d1253ae1__s000__tb131072__v001", "axis": "reasoning", "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\": \"bigbench_question_selection_generate_until__d1253ae1__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2.7042968760859494, \"nll_base\": 0.9948419436059817, \"eval_ref\": \"bigbench_question_selection_generate_until__d1253ae1...
{ "index": "bigbench_question_selection_generate_until__d1253ae1__s000__tb65536__v000", "uid": "bigbench_question_selection_generate_until__d1253ae1__s000__tb65536__v000", "axis": "reasoning", "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\": \"bigbench_question_selection_multiple_choice__3e02c704__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2.706944436565861, \"nll_base\": 0.9958204845869517, \"eval_ref\": \"bigbench_question_selection_multiple_choice__3e02c...
{ "index": "bigbench_question_selection_multiple_choice__3e02c704__s000__tb131072__v001", "uid": "bigbench_question_selection_multiple_choice__3e02c704__s000__tb131072__v001", "axis": "reasoning", "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\": \"bigbench_question_selection_multiple_choice__3e02c704__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2.706944436565861, \"nll_base\": 0.9958204845869517, \"eval_ref\": \"bigbench_question_selection_multiple_choice__3e02c70...
{ "index": "bigbench_question_selection_multiple_choice__3e02c704__s000__tb65536__v000", "uid": "bigbench_question_selection_multiple_choice__3e02c704__s000__tb65536__v000", "axis": "reasoning", "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\": \"bigbench_real_or_fake_text_generate_until__a20dcf7c__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2.330414217630285, \"nll_base\": 0.8460460275843523, \"eval_ref\": \"bigbench_real_or_fake_text_generate_until__a20dcf7c_...
{ "index": "bigbench_real_or_fake_text_generate_until__a20dcf7c__s000__tb131072__v000", "uid": "bigbench_real_or_fake_text_generate_until__a20dcf7c__s000__tb131072__v000", "axis": "reasoning", "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\": \"bigbench_real_or_fake_text_multiple_choice__f53b32a6__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2.2259966738474004, \"nll_base\": 0.8002047586237553, \"eval_ref\": \"bigbench_real_or_fake_text_multiple_choice__f53b32a6...
{ "index": "bigbench_real_or_fake_text_multiple_choice__f53b32a6__s000__tb65536__v001", "uid": "bigbench_real_or_fake_text_multiple_choice__f53b32a6__s000__tb65536__v001", "axis": "reasoning", "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\": \"bigbench_reasoning_about_colored_objects_generate_until__b3c18e1b__s000__tb131072__v002\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3464.1499727114265, \"nll_base\": 8.150222563407791, \"eval_ref\": \"bigbench_reasoning_about_colored_objec...
{ "index": "bigbench_reasoning_about_colored_objects_generate_until__b3c18e1b__s000__tb131072__v002", "uid": "bigbench_reasoning_about_colored_objects_generate_until__b3c18e1b__s000__tb131072__v002", "axis": "reasoning", "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\": \"bigbench_reasoning_about_colored_objects_multiple_choice__86323738__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1895.4571507818955, \"nll_base\": 7.5472153289408626, \"eval_ref\": \"bigbench_reasoning_about_colored_obj...
{ "index": "bigbench_reasoning_about_colored_objects_multiple_choice__86323738__s000__tb131072__v000", "uid": "bigbench_reasoning_about_colored_objects_multiple_choice__86323738__s000__tb131072__v000", "axis": "reasoning", "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\": \"bigbench_rephrase_generate_until__e9bebb33__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 5.223264811359831, \"nll_base\": 1.6531226491759607, \"eval_ref\": \"bigbench_rephrase_generate_until__e9bebb33__s000__tb131072__v...
{ "index": "bigbench_rephrase_generate_until__e9bebb33__s000__tb131072__v000", "uid": "bigbench_rephrase_generate_until__e9bebb33__s000__tb131072__v000", "axis": "reasoning", "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\": \"bigbench_rephrase_generate_until__e9bebb33__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 5.223264811359831, \"nll_base\": 1.6531226491759607, \"eval_ref\": \"bigbench_rephrase_generate_until__e9bebb33__s000__tb262144__v...
{ "index": "bigbench_rephrase_generate_until__e9bebb33__s000__tb262144__v001", "uid": "bigbench_rephrase_generate_until__e9bebb33__s000__tb262144__v001", "axis": "reasoning", "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\": \"bigbench_riddle_sense_multiple_choice__6ce0ecff__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 5777.350801194293, \"nll_base\": 8.661700517703325, \"eval_ref\": \"bigbench_riddle_sense_multiple_choice__6ce0ecff__s000__tb...
{ "index": "bigbench_riddle_sense_multiple_choice__6ce0ecff__s000__tb262144__v000", "uid": "bigbench_riddle_sense_multiple_choice__6ce0ecff__s000__tb262144__v000", "axis": "reasoning", "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\": \"bigbench_ruin_names_generate_until__f583d146__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 108.54431902155112, \"nll_base\": 4.687158559823965, \"eval_ref\": \"bigbench_ruin_names_generate_until__f583d146__s000__tb13107...
{ "index": "bigbench_ruin_names_generate_until__f583d146__s000__tb131072__v000", "uid": "bigbench_ruin_names_generate_until__f583d146__s000__tb131072__v000", "axis": "reasoning", "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\": \"bigbench_ruin_names_generate_until__f583d146__s000__tb65536__v002\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 108.54431902155112, \"nll_base\": 4.687158559823965, \"eval_ref\": \"bigbench_ruin_names_generate_until__f583d146__s000__tb65536__...
{ "index": "bigbench_ruin_names_generate_until__f583d146__s000__tb65536__v002", "uid": "bigbench_ruin_names_generate_until__f583d146__s000__tb65536__v002", "axis": "reasoning", "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\": \"bigbench_salient_translation_error_detection_generate_until__d2b80073__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 50.59135760087737, \"nll_base\": 3.923780763304079, \"eval_ref\": \"bigbench_salient_translation_error_...
{ "index": "bigbench_salient_translation_error_detection_generate_until__d2b80073__s000__tb131072__v001", "uid": "bigbench_salient_translation_error_detection_generate_until__d2b80073__s000__tb131072__v001", "axis": "reasoning", "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\": \"bigbench_salient_translation_error_detection_generate_until__d2b80073__s000__tb262144__v002\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 50.59135760087737, \"nll_base\": 3.923780763304079, \"eval_ref\": \"bigbench_salient_translation_error_...
{ "index": "bigbench_salient_translation_error_detection_generate_until__d2b80073__s000__tb262144__v002", "uid": "bigbench_salient_translation_error_detection_generate_until__d2b80073__s000__tb262144__v002", "axis": "reasoning", "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\": \"bigbench_salient_translation_error_detection_generate_until__d2b80073__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 50.59135760087737, \"nll_base\": 3.923780763304079, \"eval_ref\": \"bigbench_salient_translation_error_de...
{ "index": "bigbench_salient_translation_error_detection_generate_until__d2b80073__s000__tb65536__v000", "uid": "bigbench_salient_translation_error_detection_generate_until__d2b80073__s000__tb65536__v000", "axis": "reasoning", "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\": \"bigbench_salient_translation_error_detection_multiple_choice__995289cf__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 43.49513224855878, \"nll_base\": 3.7726490295011232, \"eval_ref\": \"bigbench_salient_translation_erro...
{ "index": "bigbench_salient_translation_error_detection_multiple_choice__995289cf__s000__tb131072__v000", "uid": "bigbench_salient_translation_error_detection_multiple_choice__995289cf__s000__tb131072__v000", "axis": "reasoning", "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\": \"bigbench_salient_translation_error_detection_multiple_choice__995289cf__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 43.49513224855878, \"nll_base\": 3.7726490295011232, \"eval_ref\": \"bigbench_salient_translation_erro...
{ "index": "bigbench_salient_translation_error_detection_multiple_choice__995289cf__s000__tb262144__v001", "uid": "bigbench_salient_translation_error_detection_multiple_choice__995289cf__s000__tb262144__v001", "axis": "reasoning", "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\": \"bigbench_salient_translation_error_detection_multiple_choice__995289cf__s000__tb65536__v002\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 43.49513224855878, \"nll_base\": 3.7726490295011232, \"eval_ref\": \"bigbench_salient_translation_error_...
{ "index": "bigbench_salient_translation_error_detection_multiple_choice__995289cf__s000__tb65536__v002", "uid": "bigbench_salient_translation_error_detection_multiple_choice__995289cf__s000__tb65536__v002", "axis": "reasoning", "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\": \"bigbench_semantic_parsing_spider_generate_until__60c23093__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 18.941705284614212, \"nll_base\": 2.9413661198513954, \"eval_ref\": \"bigbench_semantic_parsing_spider_generate_unt...
{ "index": "bigbench_semantic_parsing_spider_generate_until__60c23093__s000__tb131072__v001", "uid": "bigbench_semantic_parsing_spider_generate_until__60c23093__s000__tb131072__v001", "axis": "reasoning", "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\": \"bigbench_sentence_ambiguity_multiple_choice__82923ecc__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 554.055859207658, \"nll_base\": 6.317265510559082, \"eval_ref\": \"bigbench_sentence_ambiguity_multiple_choice__82923ec...
{ "index": "bigbench_sentence_ambiguity_multiple_choice__82923ecc__s000__tb131072__v000", "uid": "bigbench_sentence_ambiguity_multiple_choice__82923ecc__s000__tb131072__v000", "axis": "reasoning", "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\": \"bigbench_sentence_ambiguity_multiple_choice__82923ecc__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 554.055859207658, \"nll_base\": 6.317265510559082, \"eval_ref\": \"bigbench_sentence_ambiguity_multiple_choice__82923ec...
{ "index": "bigbench_sentence_ambiguity_multiple_choice__82923ecc__s000__tb262144__v001", "uid": "bigbench_sentence_ambiguity_multiple_choice__82923ecc__s000__tb262144__v001", "axis": "reasoning", "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\": \"bigbench_similarities_abstraction_generate_until__cd44b38c__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 63.30055521413653, \"nll_base\": 4.1478941002670595, \"eval_ref\": \"bigbench_similarities_abstraction_generate_un...
{ "index": "bigbench_similarities_abstraction_generate_until__cd44b38c__s000__tb131072__v001", "uid": "bigbench_similarities_abstraction_generate_until__cd44b38c__s000__tb131072__v001", "axis": "reasoning", "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\": \"bigbench_similarities_abstraction_generate_until__cd44b38c__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 63.30055521413653, \"nll_base\": 4.1478941002670595, \"eval_ref\": \"bigbench_similarities_abstraction_generate_unti...
{ "index": "bigbench_similarities_abstraction_generate_until__cd44b38c__s000__tb65536__v000", "uid": "bigbench_similarities_abstraction_generate_until__cd44b38c__s000__tb65536__v000", "axis": "reasoning", "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\": \"bigbench_simple_arithmetic_json_generate_until__c8f65033__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 144.84933129679322, \"nll_base\": 4.975694107682737, \"eval_ref\": \"bigbench_simple_arithmetic_json_generate_until_...
{ "index": "bigbench_simple_arithmetic_json_generate_until__c8f65033__s000__tb262144__v000", "uid": "bigbench_simple_arithmetic_json_generate_until__c8f65033__s000__tb262144__v000", "axis": "reasoning", "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\": \"bigbench_simple_arithmetic_json_generate_until__c8f65033__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 144.84933129679322, \"nll_base\": 4.975694107682737, \"eval_ref\": \"bigbench_simple_arithmetic_json_generate_until__c...
{ "index": "bigbench_simple_arithmetic_json_generate_until__c8f65033__s000__tb65536__v001", "uid": "bigbench_simple_arithmetic_json_generate_until__c8f65033__s000__tb65536__v001", "axis": "reasoning", "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\": \"bigbench_simple_ethical_questions_generate_until__e3dba10d__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 10.52129323386175, \"nll_base\": 2.353401130729622, \"eval_ref\": \"bigbench_simple_ethical_questions_generate_unt...
{ "index": "bigbench_simple_ethical_questions_generate_until__e3dba10d__s000__tb131072__v001", "uid": "bigbench_simple_ethical_questions_generate_until__e3dba10d__s000__tb131072__v001", "axis": "reasoning", "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\": \"bigbench_simple_ethical_questions_generate_until__e3dba10d__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 10.52129323386175, \"nll_base\": 2.353401130729622, \"eval_ref\": \"bigbench_simple_ethical_questions_generate_until...
{ "index": "bigbench_simple_ethical_questions_generate_until__e3dba10d__s000__tb65536__v000", "uid": "bigbench_simple_ethical_questions_generate_until__e3dba10d__s000__tb65536__v000", "axis": "reasoning", "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\": \"bigbench_simple_text_editing_generate_until__4f7fdd5d__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 6.053222923158847, \"nll_base\": 1.8005908447724248, \"eval_ref\": \"bigbench_simple_text_editing_generate_until__4f7fd...
{ "index": "bigbench_simple_text_editing_generate_until__4f7fdd5d__s000__tb262144__v000", "uid": "bigbench_simple_text_editing_generate_until__4f7fdd5d__s000__tb262144__v000", "axis": "reasoning", "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\": \"bigbench_snarks_generate_until__e5a89331__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 81.71233077104378, \"nll_base\": 4.403204917907715, \"eval_ref\": \"bigbench_snarks_generate_until__e5a89331__s000__tb131072__v001\"...
{ "index": "bigbench_snarks_generate_until__e5a89331__s000__tb131072__v001", "uid": "bigbench_snarks_generate_until__e5a89331__s000__tb131072__v001", "axis": "reasoning", "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\": \"bigbench_social_iqa_generate_until__7e291293__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 18.015971245362866, \"nll_base\": 2.8912586558948865, \"eval_ref\": \"bigbench_social_iqa_generate_until__7e291293__s000__tb1310...
{ "index": "bigbench_social_iqa_generate_until__7e291293__s000__tb131072__v000", "uid": "bigbench_social_iqa_generate_until__7e291293__s000__tb131072__v000", "axis": "reasoning", "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\": \"bigbench_social_iqa_generate_until__7e291293__s000__tb65536__v002\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 18.015971245362866, \"nll_base\": 2.8912586558948865, \"eval_ref\": \"bigbench_social_iqa_generate_until__7e291293__s000__tb65536_...
{ "index": "bigbench_social_iqa_generate_until__7e291293__s000__tb65536__v002", "uid": "bigbench_social_iqa_generate_until__7e291293__s000__tb65536__v002", "axis": "reasoning", "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\": \"bigbench_social_iqa_multiple_choice__d507804f__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 18.840242449152974, \"nll_base\": 2.935995137916421, \"eval_ref\": \"bigbench_social_iqa_multiple_choice__d507804f__s000__tb131...
{ "index": "bigbench_social_iqa_multiple_choice__d507804f__s000__tb131072__v000", "uid": "bigbench_social_iqa_multiple_choice__d507804f__s000__tb131072__v000", "axis": "reasoning", "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\": \"bigbench_social_support_generate_until__e0bca058__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3167.9448901473756, \"nll_base\": 8.060838356996195, \"eval_ref\": \"bigbench_social_support_generate_until__e0bca058__s000_...
{ "index": "bigbench_social_support_generate_until__e0bca058__s000__tb131072__v001", "uid": "bigbench_social_support_generate_until__e0bca058__s000__tb131072__v001", "axis": "reasoning", "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\": \"bigbench_social_support_generate_until__e0bca058__s000__tb262144__v002\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3167.9448901473756, \"nll_base\": 8.060838356996195, \"eval_ref\": \"bigbench_social_support_generate_until__e0bca058__s000_...
{ "index": "bigbench_social_support_generate_until__e0bca058__s000__tb262144__v002", "uid": "bigbench_social_support_generate_until__e0bca058__s000__tb262144__v002", "axis": "reasoning", "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\": \"bigbench_social_support_generate_until__e0bca058__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3167.9448901473756, \"nll_base\": 8.060838356996195, \"eval_ref\": \"bigbench_social_support_generate_until__e0bca058__s000__t...
{ "index": "bigbench_social_support_generate_until__e0bca058__s000__tb65536__v000", "uid": "bigbench_social_support_generate_until__e0bca058__s000__tb65536__v000", "axis": "reasoning", "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\": \"bigbench_social_support_multiple_choice__3444a94e__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 5607.36357738437, \"nll_base\": 8.6318359375, \"eval_ref\": \"bigbench_social_support_multiple_choice__3444a94e__s000__tb13...
{ "index": "bigbench_social_support_multiple_choice__3444a94e__s000__tb131072__v000", "uid": "bigbench_social_support_multiple_choice__3444a94e__s000__tb131072__v000", "axis": "reasoning", "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\": \"bigbench_social_support_multiple_choice__3444a94e__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 5607.36357738437, \"nll_base\": 8.6318359375, \"eval_ref\": \"bigbench_social_support_multiple_choice__3444a94e__s000__tb26...
{ "index": "bigbench_social_support_multiple_choice__3444a94e__s000__tb262144__v001", "uid": "bigbench_social_support_multiple_choice__3444a94e__s000__tb262144__v001", "axis": "reasoning", "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\": \"bigbench_sports_understanding_multiple_choice__b5bfa4a4__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 89.84577225140788, \"nll_base\": 4.49809455871582, \"eval_ref\": \"bigbench_sports_understanding_multiple_choice__b5b...
{ "index": "bigbench_sports_understanding_multiple_choice__b5bfa4a4__s000__tb131072__v000", "uid": "bigbench_sports_understanding_multiple_choice__b5bfa4a4__s000__tb131072__v000", "axis": "reasoning", "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\": \"bigbench_strange_stories_generate_until__189bd7f3__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 102.17397017775916, \"nll_base\": 4.626676950395478, \"eval_ref\": \"bigbench_strange_stories_generate_until__189bd7f3__s00...
{ "index": "bigbench_strange_stories_generate_until__189bd7f3__s000__tb131072__v000", "uid": "bigbench_strange_stories_generate_until__189bd7f3__s000__tb131072__v000", "axis": "reasoning", "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\": \"bigbench_strange_stories_generate_until__189bd7f3__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 102.17397017775916, \"nll_base\": 4.626676950395478, \"eval_ref\": \"bigbench_strange_stories_generate_until__189bd7f3__s00...
{ "index": "bigbench_strange_stories_generate_until__189bd7f3__s000__tb262144__v001", "uid": "bigbench_strange_stories_generate_until__189bd7f3__s000__tb262144__v001", "axis": "reasoning", "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\": \"bigbench_strange_stories_multiple_choice__daa4f6e0__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 41.860532602416455, \"nll_base\": 3.7343434403582316, \"eval_ref\": \"bigbench_strange_stories_multiple_choice__daa4f6e0__...
{ "index": "bigbench_strange_stories_multiple_choice__daa4f6e0__s000__tb131072__v001", "uid": "bigbench_strange_stories_multiple_choice__daa4f6e0__s000__tb131072__v001", "axis": "reasoning", "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\": \"bigbench_strange_stories_multiple_choice__daa4f6e0__s000__tb262144__v002\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 41.860532602416455, \"nll_base\": 3.7343434403582316, \"eval_ref\": \"bigbench_strange_stories_multiple_choice__daa4f6e0__...
{ "index": "bigbench_strange_stories_multiple_choice__daa4f6e0__s000__tb262144__v002", "uid": "bigbench_strange_stories_multiple_choice__daa4f6e0__s000__tb262144__v002", "axis": "reasoning", "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\": \"bigbench_strange_stories_multiple_choice__daa4f6e0__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 41.860532602416455, \"nll_base\": 3.7343434403582316, \"eval_ref\": \"bigbench_strange_stories_multiple_choice__daa4f6e0__s0...
{ "index": "bigbench_strange_stories_multiple_choice__daa4f6e0__s000__tb65536__v000", "uid": "bigbench_strange_stories_multiple_choice__daa4f6e0__s000__tb65536__v000", "axis": "reasoning", "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\": \"bigbench_strategyqa_multiple_choice__abff5d97__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1448.2479419193396, \"nll_base\": 7.278109788894653, \"eval_ref\": \"bigbench_strategyqa_multiple_choice__abff5d97__s000__tb131...
{ "index": "bigbench_strategyqa_multiple_choice__abff5d97__s000__tb131072__v000", "uid": "bigbench_strategyqa_multiple_choice__abff5d97__s000__tb131072__v000", "axis": "reasoning", "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\": \"bigbench_sufficient_information_generate_until__e038b032__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 75.16666668235959, \"nll_base\": 4.319707870483398, \"eval_ref\": \"bigbench_sufficient_information_generate_until__e0...
{ "index": "bigbench_sufficient_information_generate_until__e038b032__s000__tb65536__v000", "uid": "bigbench_sufficient_information_generate_until__e038b032__s000__tb65536__v000", "axis": "reasoning", "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\": \"bigbench_suicide_risk_generate_until__ac1fa7ea__s000__tb65536__v002\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 425.80180129111267, \"nll_base\": 6.0539739828194135, \"eval_ref\": \"bigbench_suicide_risk_generate_until__ac1fa7ea__s000__tb65...
{ "index": "bigbench_suicide_risk_generate_until__ac1fa7ea__s000__tb65536__v002", "uid": "bigbench_suicide_risk_generate_until__ac1fa7ea__s000__tb65536__v002", "axis": "reasoning", "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\": \"bigbench_suicide_risk_multiple_choice__cf49518d__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 491.85267363157897, \"nll_base\": 6.198179227811796, \"eval_ref\": \"bigbench_suicide_risk_multiple_choice__cf49518d__s000__t...
{ "index": "bigbench_suicide_risk_multiple_choice__cf49518d__s000__tb131072__v001", "uid": "bigbench_suicide_risk_multiple_choice__cf49518d__s000__tb131072__v001", "axis": "reasoning", "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\": \"bigbench_suicide_risk_multiple_choice__cf49518d__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 491.85267363157897, \"nll_base\": 6.198179227811796, \"eval_ref\": \"bigbench_suicide_risk_multiple_choice__cf49518d__s000__tb6...
{ "index": "bigbench_suicide_risk_multiple_choice__cf49518d__s000__tb65536__v000", "uid": "bigbench_suicide_risk_multiple_choice__cf49518d__s000__tb65536__v000", "axis": "reasoning", "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\": \"bigbench_swahili_english_proverbs_generate_until__264114cd__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 9.17927960844265, \"nll_base\": 2.2169487275272965, \"eval_ref\": \"bigbench_swahili_english_proverbs_generate_unt...
{ "index": "bigbench_swahili_english_proverbs_generate_until__264114cd__s000__tb131072__v001", "uid": "bigbench_swahili_english_proverbs_generate_until__264114cd__s000__tb131072__v001", "axis": "reasoning", "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\": \"bigbench_swahili_english_proverbs_generate_until__264114cd__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 9.17927960844265, \"nll_base\": 2.2169487275272965, \"eval_ref\": \"bigbench_swahili_english_proverbs_generate_until...
{ "index": "bigbench_swahili_english_proverbs_generate_until__264114cd__s000__tb65536__v000", "uid": "bigbench_swahili_english_proverbs_generate_until__264114cd__s000__tb65536__v000", "axis": "reasoning", "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\": \"bigbench_swahili_english_proverbs_multiple_choice__29123686__s000__tb131072__v002\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 8.41483340772708, \"nll_base\": 2.1299960303852576, \"eval_ref\": \"bigbench_swahili_english_proverbs_multiple_ch...
{ "index": "bigbench_swahili_english_proverbs_multiple_choice__29123686__s000__tb131072__v002", "uid": "bigbench_swahili_english_proverbs_multiple_choice__29123686__s000__tb131072__v002", "axis": "reasoning", "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\": \"bigbench_swahili_english_proverbs_multiple_choice__29123686__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 8.41483340772708, \"nll_base\": 2.1299960303852576, \"eval_ref\": \"bigbench_swahili_english_proverbs_multiple_ch...
{ "index": "bigbench_swahili_english_proverbs_multiple_choice__29123686__s000__tb262144__v000", "uid": "bigbench_swahili_english_proverbs_multiple_choice__29123686__s000__tb262144__v000", "axis": "reasoning", "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\": \"bigbench_swahili_english_proverbs_multiple_choice__29123686__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 8.41483340772708, \"nll_base\": 2.1299960303852576, \"eval_ref\": \"bigbench_swahili_english_proverbs_multiple_choi...
{ "index": "bigbench_swahili_english_proverbs_multiple_choice__29123686__s000__tb65536__v001", "uid": "bigbench_swahili_english_proverbs_multiple_choice__29123686__s000__tb65536__v001", "axis": "reasoning", "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\": \"bigbench_swedish_to_german_proverbs_generate_until__bce78ef6__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 5.102629296000922, \"nll_base\": 1.6297559551164216, \"eval_ref\": \"bigbench_swedish_to_german_proverbs_generat...
{ "index": "bigbench_swedish_to_german_proverbs_generate_until__bce78ef6__s000__tb131072__v001", "uid": "bigbench_swedish_to_german_proverbs_generate_until__bce78ef6__s000__tb131072__v001", "axis": "reasoning", "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\": \"bigbench_swedish_to_german_proverbs_generate_until__bce78ef6__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 5.102629296000922, \"nll_base\": 1.6297559551164216, \"eval_ref\": \"bigbench_swedish_to_german_proverbs_generate_...
{ "index": "bigbench_swedish_to_german_proverbs_generate_until__bce78ef6__s000__tb65536__v000", "uid": "bigbench_swedish_to_german_proverbs_generate_until__bce78ef6__s000__tb65536__v000", "axis": "reasoning", "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\": \"bigbench_swedish_to_german_proverbs_multiple_choice__7bb2824b__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 5.427353163583393, \"nll_base\": 1.6914515682511186, \"eval_ref\": \"bigbench_swedish_to_german_proverbs_multip...
{ "index": "bigbench_swedish_to_german_proverbs_multiple_choice__7bb2824b__s000__tb131072__v000", "uid": "bigbench_swedish_to_german_proverbs_multiple_choice__7bb2824b__s000__tb131072__v000", "axis": "reasoning", "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\": \"bigbench_swedish_to_german_proverbs_multiple_choice__7bb2824b__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 5.427353163583393, \"nll_base\": 1.6914515682511186, \"eval_ref\": \"bigbench_swedish_to_german_proverbs_multip...
{ "index": "bigbench_swedish_to_german_proverbs_multiple_choice__7bb2824b__s000__tb262144__v001", "uid": "bigbench_swedish_to_german_proverbs_multiple_choice__7bb2824b__s000__tb262144__v001", "axis": "reasoning", "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\": \"bigbench_symbol_interpretation_generate_until__5bbef10a__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4.034900777563687, \"nll_base\": 1.3949817109768576, \"eval_ref\": \"bigbench_symbol_interpretation_generate_until__5...
{ "index": "bigbench_symbol_interpretation_generate_until__5bbef10a__s000__tb131072__v001", "uid": "bigbench_symbol_interpretation_generate_until__5bbef10a__s000__tb131072__v001", "axis": "reasoning", "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\": \"bigbench_temporal_sequences_generate_until__7317cb85__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 8.94044640998438, \"nll_base\": 2.190585521941489, \"eval_ref\": \"bigbench_temporal_sequences_generate_until__7317cb85_...
{ "index": "bigbench_temporal_sequences_generate_until__7317cb85__s000__tb131072__v001", "uid": "bigbench_temporal_sequences_generate_until__7317cb85__s000__tb131072__v001", "axis": "reasoning", "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\": \"bigbench_temporal_sequences_generate_until__7317cb85__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 8.94044640998438, \"nll_base\": 2.190585521941489, \"eval_ref\": \"bigbench_temporal_sequences_generate_until__7317cb85__s...
{ "index": "bigbench_temporal_sequences_generate_until__7317cb85__s000__tb65536__v000", "uid": "bigbench_temporal_sequences_generate_until__7317cb85__s000__tb65536__v000", "axis": "reasoning", "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\": \"bigbench_temporal_sequences_multiple_choice__0797b992__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 8.7932540323063, \"nll_base\": 2.173984840267994, \"eval_ref\": \"bigbench_temporal_sequences_multiple_choice__0797b992...
{ "index": "bigbench_temporal_sequences_multiple_choice__0797b992__s000__tb131072__v001", "uid": "bigbench_temporal_sequences_multiple_choice__0797b992__s000__tb131072__v001", "axis": "reasoning", "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\": \"bigbench_temporal_sequences_multiple_choice__0797b992__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 8.7932540323063, \"nll_base\": 2.173984840267994, \"eval_ref\": \"bigbench_temporal_sequences_multiple_choice__0797b992__...
{ "index": "bigbench_temporal_sequences_multiple_choice__0797b992__s000__tb65536__v000", "uid": "bigbench_temporal_sequences_multiple_choice__0797b992__s000__tb65536__v000", "axis": "reasoning", "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\": \"bigbench_tense_generate_until__131f05e6__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 6.682410463304869, \"nll_base\": 1.899478770258775, \"eval_ref\": \"bigbench_tense_generate_until__131f05e6__s000__tb65536__v000\", \"o...
{ "index": "bigbench_tense_generate_until__131f05e6__s000__tb65536__v000", "uid": "bigbench_tense_generate_until__131f05e6__s000__tb65536__v000", "axis": "reasoning", "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\": \"bigbench_timedial_generate_until__e122df07__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 126.61381893306515, \"nll_base\": 4.841141658042794, \"eval_ref\": \"bigbench_timedial_generate_until__e122df07__s000__tb65536__v001...
{ "index": "bigbench_timedial_generate_until__e122df07__s000__tb65536__v001", "uid": "bigbench_timedial_generate_until__e122df07__s000__tb65536__v001", "axis": "reasoning", "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\": \"bigbench_timedial_multiple_choice__fd56da5a__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 79.4469236719531, \"nll_base\": 4.375089171930408, \"eval_ref\": \"bigbench_timedial_multiple_choice__fd56da5a__s000__tb65536__v001...
{ "index": "bigbench_timedial_multiple_choice__fd56da5a__s000__tb65536__v001", "uid": "bigbench_timedial_multiple_choice__fd56da5a__s000__tb65536__v001", "axis": "reasoning", "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\": \"bigbench_topical_chat_generate_until__ad0eb5ea__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 40.76277740522559, \"nll_base\": 3.7077693464949326, \"eval_ref\": \"bigbench_topical_chat_generate_until__ad0eb5ea__s000__tb1...
{ "index": "bigbench_topical_chat_generate_until__ad0eb5ea__s000__tb131072__v001", "uid": "bigbench_topical_chat_generate_until__ad0eb5ea__s000__tb131072__v001", "axis": "reasoning", "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\": \"bigbench_understanding_fables_generate_until__9bbcbd76__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 8.610541750262193, \"nll_base\": 2.1529872375127126, \"eval_ref\": \"bigbench_understanding_fables_generate_until__9bb...
{ "index": "bigbench_understanding_fables_generate_until__9bbcbd76__s000__tb131072__v001", "uid": "bigbench_understanding_fables_generate_until__9bbcbd76__s000__tb131072__v001", "axis": "reasoning", "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\": \"bigbench_understanding_fables_generate_until__9bbcbd76__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 8.610541750262193, \"nll_base\": 2.1529872375127126, \"eval_ref\": \"bigbench_understanding_fables_generate_until__9bbcb...
{ "index": "bigbench_understanding_fables_generate_until__9bbcbd76__s000__tb65536__v000", "uid": "bigbench_understanding_fables_generate_until__9bbcbd76__s000__tb65536__v000", "axis": "reasoning", "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\": \"bigbench_understanding_fables_multiple_choice__144360ba__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 8.028812990752645, \"nll_base\": 2.0830366952078685, \"eval_ref\": \"bigbench_understanding_fables_multiple_choice__1...
{ "index": "bigbench_understanding_fables_multiple_choice__144360ba__s000__tb131072__v001", "uid": "bigbench_understanding_fables_multiple_choice__144360ba__s000__tb131072__v001", "axis": "reasoning", "budget": 2048, "token_budget": 131072, "variant": "filter" }