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\": \"babilong_qa1__7fd5d334__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1511.3957484413688, \"nll_base\": 7.320788839588994, \"eval_ref\": \"babilong_qa1__7fd5d334__s000__tb65536__v000\", \"oracle_delta\": 0.1908456553583555... | {
"index": "babilong_qa1__7fd5d334__s000__tb65536__v000",
"uid": "babilong_qa1__7fd5d334__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\": \"babilong_qa20__3b4d4d68__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 501.59173728315625, \"nll_base\": 6.217786516462054, \"eval_ref\": \"babilong_qa20__3b4d4d68__s000__tb131072__v000\", \"oracle_delta\": 0.34727783203... | {
"index": "babilong_qa20__3b4d4d68__s000__tb131072__v000",
"uid": "babilong_qa20__3b4d4d68__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\": \"babilong_qa2__dc624378__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1378.8941367151647, \"nll_base\": 7.229037106692136, \"eval_ref\": \"babilong_qa2__dc624378__s000__tb65536__v001\", \"oracle_delta\": 0.1845069298377399... | {
"index": "babilong_qa2__dc624378__s000__tb65536__v001",
"uid": "babilong_qa2__dc624378__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\": \"babilong_qa3__83166444__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1739.3754604635067, \"nll_base\": 7.461281397009409, \"eval_ref\": \"babilong_qa3__83166444__s000__tb131072__v001\", \"oracle_delta\": 0.4292901151923... | {
"index": "babilong_qa3__83166444__s000__tb131072__v001",
"uid": "babilong_qa3__83166444__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\": \"babilong_qa3__83166444__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1739.3754604635067, \"nll_base\": 7.461281397009409, \"eval_ref\": \"babilong_qa3__83166444__s000__tb65536__v000\", \"oracle_delta\": 0.1999232794648859... | {
"index": "babilong_qa3__83166444__s000__tb65536__v000",
"uid": "babilong_qa3__83166444__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\": \"babilong_qa5__6e2cf7ed__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 5265.768456452834, \"nll_base\": 8.568982369559151, \"eval_ref\": \"babilong_qa5__6e2cf7ed__s000__tb131072__v000\", \"oracle_delta\": 1.21522696358816... | {
"index": "babilong_qa5__6e2cf7ed__s000__tb131072__v000",
"uid": "babilong_qa5__6e2cf7ed__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\": \"babilong_qa5__6e2cf7ed__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 5265.768456452834, \"nll_base\": 8.568982369559151, \"eval_ref\": \"babilong_qa5__6e2cf7ed__s000__tb262144__v001\", \"oracle_delta\": 1.27538016183035... | {
"index": "babilong_qa5__6e2cf7ed__s000__tb262144__v001",
"uid": "babilong_qa5__6e2cf7ed__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\": \"babilong_qa6__ed4707a5__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 5884.522574973091, \"nll_base\": 8.680080890655518, \"eval_ref\": \"babilong_qa6__ed4707a5__s000__tb131072__v000\", \"oracle_delta\": 1.53137254714965... | {
"index": "babilong_qa6__ed4707a5__s000__tb131072__v000",
"uid": "babilong_qa6__ed4707a5__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\": \"babilong_qa6__ed4707a5__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 5884.522574973091, \"nll_base\": 8.680080890655518, \"eval_ref\": \"babilong_qa6__ed4707a5__s000__tb262144__v001\", \"oracle_delta\": 1.31803035736083... | {
"index": "babilong_qa6__ed4707a5__s000__tb262144__v001",
"uid": "babilong_qa6__ed4707a5__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\": \"babilong_qa6__ed4707a5__s000__tb65536__v002\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 5884.522574973091, \"nll_base\": 8.680080890655518, \"eval_ref\": \"babilong_qa6__ed4707a5__s000__tb65536__v002\", \"oracle_delta\": 0.8860228061676025,... | {
"index": "babilong_qa6__ed4707a5__s000__tb65536__v002",
"uid": "babilong_qa6__ed4707a5__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\": \"babilong_qa7__66378dca__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 15682.941391434391, \"nll_base\": 9.66032886505127, \"eval_ref\": \"babilong_qa7__66378dca__s000__tb131072__v001\", \"oracle_delta\": 0.59184837341308... | {
"index": "babilong_qa7__66378dca__s000__tb131072__v001",
"uid": "babilong_qa7__66378dca__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\": \"babilong_qa7__66378dca__s000__tb262144__v002\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 15682.941391434391, \"nll_base\": 9.66032886505127, \"eval_ref\": \"babilong_qa7__66378dca__s000__tb262144__v002\", \"oracle_delta\": 0.52186059951782... | {
"index": "babilong_qa7__66378dca__s000__tb262144__v002",
"uid": "babilong_qa7__66378dca__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\": \"babilong_qa7__66378dca__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 15682.941391434391, \"nll_base\": 9.66032886505127, \"eval_ref\": \"babilong_qa7__66378dca__s000__tb65536__v000\", \"oracle_delta\": 0.3892245292663574,... | {
"index": "babilong_qa7__66378dca__s000__tb65536__v000",
"uid": "babilong_qa7__66378dca__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\": \"babilong_qa8__0122efef__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 7944.543745219848, \"nll_base\": 8.980240650666065, \"eval_ref\": \"babilong_qa8__0122efef__s000__tb131072__v000\", \"oracle_delta\": 0.88107025928986... | {
"index": "babilong_qa8__0122efef__s000__tb131072__v000",
"uid": "babilong_qa8__0122efef__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\": \"babilong_qa9__d28fd694__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3738.0319849402435, \"nll_base\": 8.226314544677734, \"eval_ref\": \"babilong_qa9__d28fd694__s000__tb262144__v000\", \"oracle_delta\": 1.4104328155517... | {
"index": "babilong_qa9__d28fd694__s000__tb262144__v000",
"uid": "babilong_qa9__d28fd694__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\": \"babilong_qa9__d28fd694__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3738.0319849402435, \"nll_base\": 8.226314544677734, \"eval_ref\": \"babilong_qa9__d28fd694__s000__tb65536__v001\", \"oracle_delta\": 0.6550471782684326... | {
"index": "babilong_qa9__d28fd694__s000__tb65536__v001",
"uid": "babilong_qa9__d28fd694__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\": \"bbh_cot_fewshot_boolean_expressions__bc93583e__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1572.9258134513643, \"nll_base\": 7.360692739486694, \"eval_ref\": \"bbh_cot_fewshot_boolean_expressions__bc93583e__s000__tb131... | {
"index": "bbh_cot_fewshot_boolean_expressions__bc93583e__s000__tb131072__v001",
"uid": "bbh_cot_fewshot_boolean_expressions__bc93583e__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\": \"bbh_cot_fewshot_boolean_expressions__bc93583e__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1572.9258134513643, \"nll_base\": 7.360692739486694, \"eval_ref\": \"bbh_cot_fewshot_boolean_expressions__bc93583e__s000__tb65536... | {
"index": "bbh_cot_fewshot_boolean_expressions__bc93583e__s000__tb65536__v000",
"uid": "bbh_cot_fewshot_boolean_expressions__bc93583e__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\": \"bbh_cot_fewshot_causal_judgement__1424a484__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3482.1013135715975, \"nll_base\": 8.155391216278076, \"eval_ref\": \"bbh_cot_fewshot_causal_judgement__1424a484__s000__tb131072__v... | {
"index": "bbh_cot_fewshot_causal_judgement__1424a484__s000__tb131072__v000",
"uid": "bbh_cot_fewshot_causal_judgement__1424a484__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\": \"bbh_cot_fewshot_navigate__fb0b232d__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2814.9201671700416, \"nll_base\": 7.9426891803741455, \"eval_ref\": \"bbh_cot_fewshot_navigate__fb0b232d__s000__tb131072__v000\", \"oracle... | {
"index": "bbh_cot_fewshot_navigate__fb0b232d__s000__tb131072__v000",
"uid": "bbh_cot_fewshot_navigate__fb0b232d__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\": \"bbh_cot_fewshot_reasoning_about_colored_objects__66e69415__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 108.21976787122996, \"nll_base\": 4.684164047241211, \"eval_ref\": \"bbh_cot_fewshot_reasoning_about_colored_object... | {
"index": "bbh_cot_fewshot_reasoning_about_colored_objects__66e69415__s000__tb131072__v000",
"uid": "bbh_cot_fewshot_reasoning_about_colored_objects__66e69415__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\": \"bbh_cot_fewshot_tracking_shuffled_objects_five_objects__f9664b68__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 70.86066503710413, \"nll_base\": 4.260715484619141, \"eval_ref\": \"bbh_cot_fewshot_tracking_shuffled_object... | {
"index": "bbh_cot_fewshot_tracking_shuffled_objects_five_objects__f9664b68__s000__tb131072__v001",
"uid": "bbh_cot_fewshot_tracking_shuffled_objects_five_objects__f9664b68__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\": \"bbh_cot_fewshot_tracking_shuffled_objects_five_objects__f9664b68__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 70.86066503710413, \"nll_base\": 4.260715484619141, \"eval_ref\": \"bbh_cot_fewshot_tracking_shuffled_objects_... | {
"index": "bbh_cot_fewshot_tracking_shuffled_objects_five_objects__f9664b68__s000__tb65536__v000",
"uid": "bbh_cot_fewshot_tracking_shuffled_objects_five_objects__f9664b68__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\": \"bbh_cot_fewshot_tracking_shuffled_objects_seven_objects__3961eddf__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 79.71258719370722, \"nll_base\": 4.378427505493164, \"eval_ref\": \"bbh_cot_fewshot_tracking_shuffled_objec... | {
"index": "bbh_cot_fewshot_tracking_shuffled_objects_seven_objects__3961eddf__s000__tb131072__v000",
"uid": "bbh_cot_fewshot_tracking_shuffled_objects_seven_objects__3961eddf__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\": \"bbh_cot_fewshot_tracking_shuffled_objects_three_objects__10a8aa0a__s000__tb131072__v002\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 58.84353843530539, \"nll_base\": 4.0748820304870605, \"eval_ref\": \"bbh_cot_fewshot_tracking_shuffled_obje... | {
"index": "bbh_cot_fewshot_tracking_shuffled_objects_three_objects__10a8aa0a__s000__tb131072__v002",
"uid": "bbh_cot_fewshot_tracking_shuffled_objects_three_objects__10a8aa0a__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\": \"bbh_cot_fewshot_tracking_shuffled_objects_three_objects__10a8aa0a__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 58.84353843530539, \"nll_base\": 4.0748820304870605, \"eval_ref\": \"bbh_cot_fewshot_tracking_shuffled_object... | {
"index": "bbh_cot_fewshot_tracking_shuffled_objects_three_objects__10a8aa0a__s000__tb65536__v001",
"uid": "bbh_cot_fewshot_tracking_shuffled_objects_three_objects__10a8aa0a__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\": \"bbh_cot_fewshot_web_of_lies__96f24ae7__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 13525.750056704346, \"nll_base\": 9.51235055923462, \"eval_ref\": \"bbh_cot_fewshot_web_of_lies__96f24ae7__s000__tb131072__v001\", \"or... | {
"index": "bbh_cot_fewshot_web_of_lies__96f24ae7__s000__tb131072__v001",
"uid": "bbh_cot_fewshot_web_of_lies__96f24ae7__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\": \"bbh_cot_fewshot_web_of_lies__96f24ae7__s000__tb262144__v002\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 13525.750056704346, \"nll_base\": 9.51235055923462, \"eval_ref\": \"bbh_cot_fewshot_web_of_lies__96f24ae7__s000__tb262144__v002\", \"or... | {
"index": "bbh_cot_fewshot_web_of_lies__96f24ae7__s000__tb262144__v002",
"uid": "bbh_cot_fewshot_web_of_lies__96f24ae7__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\": \"bbh_cot_fewshot_web_of_lies__96f24ae7__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 13525.750056704346, \"nll_base\": 9.51235055923462, \"eval_ref\": \"bbh_cot_fewshot_web_of_lies__96f24ae7__s000__tb65536__v000\", \"oracl... | {
"index": "bbh_cot_fewshot_web_of_lies__96f24ae7__s000__tb65536__v000",
"uid": "bbh_cot_fewshot_web_of_lies__96f24ae7__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\": \"bbh_cot_zeroshot_boolean_expressions__2710e764__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1643.9644974342693, \"nll_base\": 7.404865980148315, \"eval_ref\": \"bbh_cot_zeroshot_boolean_expressions__2710e764__s000__tb1... | {
"index": "bbh_cot_zeroshot_boolean_expressions__2710e764__s000__tb131072__v000",
"uid": "bbh_cot_zeroshot_boolean_expressions__2710e764__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\": \"bbh_cot_zeroshot_boolean_expressions__2710e764__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1643.9644974342693, \"nll_base\": 7.404865980148315, \"eval_ref\": \"bbh_cot_zeroshot_boolean_expressions__2710e764__s000__tb2... | {
"index": "bbh_cot_zeroshot_boolean_expressions__2710e764__s000__tb262144__v001",
"uid": "bbh_cot_zeroshot_boolean_expressions__2710e764__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\": \"bbh_cot_zeroshot_causal_judgement__c7bbf1d4__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2933.492424636864, \"nll_base\": 7.9839489459991455, \"eval_ref\": \"bbh_cot_zeroshot_causal_judgement__c7bbf1d4__s000__tb131072_... | {
"index": "bbh_cot_zeroshot_causal_judgement__c7bbf1d4__s000__tb131072__v001",
"uid": "bbh_cot_zeroshot_causal_judgement__c7bbf1d4__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\": \"bbh_cot_zeroshot_causal_judgement__c7bbf1d4__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2933.492424636864, \"nll_base\": 7.9839489459991455, \"eval_ref\": \"bbh_cot_zeroshot_causal_judgement__c7bbf1d4__s000__tb65536__v0... | {
"index": "bbh_cot_zeroshot_causal_judgement__c7bbf1d4__s000__tb65536__v000",
"uid": "bbh_cot_zeroshot_causal_judgement__c7bbf1d4__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\": \"bbh_cot_zeroshot_logical_deduction_five_objects__ecc48e6f__s000__tb131072__v002\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 69.81258702802866, \"nll_base\": 4.245814323425293, \"eval_ref\": \"bbh_cot_zeroshot_logical_deduction_five_objects... | {
"index": "bbh_cot_zeroshot_logical_deduction_five_objects__ecc48e6f__s000__tb131072__v002",
"uid": "bbh_cot_zeroshot_logical_deduction_five_objects__ecc48e6f__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\": \"bbh_cot_zeroshot_logical_deduction_seven_objects__4c16021b__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 80.56607566240746, \"nll_base\": 4.389077663421631, \"eval_ref\": \"bbh_cot_zeroshot_logical_deduction_seven_objec... | {
"index": "bbh_cot_zeroshot_logical_deduction_seven_objects__4c16021b__s000__tb131072__v001",
"uid": "bbh_cot_zeroshot_logical_deduction_seven_objects__4c16021b__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\": \"bbh_cot_zeroshot_object_counting__f27fbd0f__s000__tb131072__v002\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1902.251781446547, \"nll_base\": 7.550793611550633, \"eval_ref\": \"bbh_cot_zeroshot_object_counting__f27fbd0f__s000__tb131072__v0... | {
"index": "bbh_cot_zeroshot_object_counting__f27fbd0f__s000__tb131072__v002",
"uid": "bbh_cot_zeroshot_object_counting__f27fbd0f__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\": \"bbh_cot_zeroshot_object_counting__f27fbd0f__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1902.251781446547, \"nll_base\": 7.550793611550633, \"eval_ref\": \"bbh_cot_zeroshot_object_counting__f27fbd0f__s000__tb262144__v0... | {
"index": "bbh_cot_zeroshot_object_counting__f27fbd0f__s000__tb262144__v000",
"uid": "bbh_cot_zeroshot_object_counting__f27fbd0f__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\": \"bbh_cot_zeroshot_reasoning_about_colored_objects__0d1bea2e__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 79.2336056218465, \"nll_base\": 4.372400522232056, \"eval_ref\": \"bbh_cot_zeroshot_reasoning_about_colored_object... | {
"index": "bbh_cot_zeroshot_reasoning_about_colored_objects__0d1bea2e__s000__tb131072__v000",
"uid": "bbh_cot_zeroshot_reasoning_about_colored_objects__0d1bea2e__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\": \"bbh_cot_zeroshot_tracking_shuffled_objects_five_objects__d67458fc__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 56.68438140394459, \"nll_base\": 4.037498712539673, \"eval_ref\": \"bbh_cot_zeroshot_tracking_shuffled_obje... | {
"index": "bbh_cot_zeroshot_tracking_shuffled_objects_five_objects__d67458fc__s000__tb131072__v001",
"uid": "bbh_cot_zeroshot_tracking_shuffled_objects_five_objects__d67458fc__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\": \"bbh_cot_zeroshot_tracking_shuffled_objects_five_objects__d67458fc__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 56.68438140394459, \"nll_base\": 4.037498712539673, \"eval_ref\": \"bbh_cot_zeroshot_tracking_shuffled_object... | {
"index": "bbh_cot_zeroshot_tracking_shuffled_objects_five_objects__d67458fc__s000__tb65536__v000",
"uid": "bbh_cot_zeroshot_tracking_shuffled_objects_five_objects__d67458fc__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\": \"bbh_cot_zeroshot_tracking_shuffled_objects_seven_objects__e6a8a228__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 70.86968727393196, \"nll_base\": 4.260842800140381, \"eval_ref\": \"bbh_cot_zeroshot_tracking_shuffled_objec... | {
"index": "bbh_cot_zeroshot_tracking_shuffled_objects_seven_objects__e6a8a228__s000__tb65536__v001",
"uid": "bbh_cot_zeroshot_tracking_shuffled_objects_seven_objects__e6a8a228__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\": \"bbh_cot_zeroshot_tracking_shuffled_objects_three_objects__f70fd602__s000__tb131072__v002\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 41.74408342403444, \"nll_base\": 3.7315577268600464, \"eval_ref\": \"bbh_cot_zeroshot_tracking_shuffled_ob... | {
"index": "bbh_cot_zeroshot_tracking_shuffled_objects_three_objects__f70fd602__s000__tb131072__v002",
"uid": "bbh_cot_zeroshot_tracking_shuffled_objects_three_objects__f70fd602__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\": \"bbh_cot_zeroshot_web_of_lies__c92ad13b__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 14936.85678871292, \"nll_base\": 9.611587047576904, \"eval_ref\": \"bbh_cot_zeroshot_web_of_lies__c92ad13b__s000__tb131072__v001\", \"... | {
"index": "bbh_cot_zeroshot_web_of_lies__c92ad13b__s000__tb131072__v001",
"uid": "bbh_cot_zeroshot_web_of_lies__c92ad13b__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\": \"bbh_cot_zeroshot_web_of_lies__c92ad13b__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 14936.85678871292, \"nll_base\": 9.611587047576904, \"eval_ref\": \"bbh_cot_zeroshot_web_of_lies__c92ad13b__s000__tb65536__v000\", \"ora... | {
"index": "bbh_cot_zeroshot_web_of_lies__c92ad13b__s000__tb65536__v000",
"uid": "bbh_cot_zeroshot_web_of_lies__c92ad13b__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\": \"bbh_fewshot_boolean_expressions__548ac767__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 935.3523557111789, \"nll_base\": 6.840923309326172, \"eval_ref\": \"bbh_fewshot_boolean_expressions__548ac767__s000__tb262144__v000... | {
"index": "bbh_fewshot_boolean_expressions__548ac767__s000__tb262144__v000",
"uid": "bbh_fewshot_boolean_expressions__548ac767__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\": \"bbh_fewshot_boolean_expressions__548ac767__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 935.3523557111789, \"nll_base\": 6.840923309326172, \"eval_ref\": \"bbh_fewshot_boolean_expressions__548ac767__s000__tb65536__v001\",... | {
"index": "bbh_fewshot_boolean_expressions__548ac767__s000__tb65536__v001",
"uid": "bbh_fewshot_boolean_expressions__548ac767__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\": \"bbh_fewshot_causal_judgement__9c9d9704__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1270.3566963824037, \"nll_base\": 7.147053003311157, \"eval_ref\": \"bbh_fewshot_causal_judgement__9c9d9704__s000__tb131072__v000\", \... | {
"index": "bbh_fewshot_causal_judgement__9c9d9704__s000__tb131072__v000",
"uid": "bbh_fewshot_causal_judgement__9c9d9704__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\": \"bbh_fewshot_date_understanding__0d9bbbe6__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 75.99251376801877, \"nll_base\": 4.330634832382202, \"eval_ref\": \"bbh_fewshot_date_understanding__0d9bbbe6__s000__tb131072__v001\"... | {
"index": "bbh_fewshot_date_understanding__0d9bbbe6__s000__tb131072__v001",
"uid": "bbh_fewshot_date_understanding__0d9bbbe6__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\": \"bbh_fewshot_date_understanding__0d9bbbe6__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 75.99251376801877, \"nll_base\": 4.330634832382202, \"eval_ref\": \"bbh_fewshot_date_understanding__0d9bbbe6__s000__tb65536__v000\", \... | {
"index": "bbh_fewshot_date_understanding__0d9bbbe6__s000__tb65536__v000",
"uid": "bbh_fewshot_date_understanding__0d9bbbe6__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\": \"bbh_fewshot_logical_deduction_five_objects__05c555bf__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 70.97846682310457, \"nll_base\": 4.262376546859741, \"eval_ref\": \"bbh_fewshot_logical_deduction_five_objects__05c555bf... | {
"index": "bbh_fewshot_logical_deduction_five_objects__05c555bf__s000__tb131072__v001",
"uid": "bbh_fewshot_logical_deduction_five_objects__05c555bf__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\": \"bbh_fewshot_logical_deduction_seven_objects__054174b4__s000__tb131072__v002\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 84.17017023274583, \"nll_base\": 4.432840585708618, \"eval_ref\": \"bbh_fewshot_logical_deduction_seven_objects__054174... | {
"index": "bbh_fewshot_logical_deduction_seven_objects__054174b4__s000__tb131072__v002",
"uid": "bbh_fewshot_logical_deduction_seven_objects__054174b4__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\": \"bbh_fewshot_logical_deduction_three_objects__de5d13ac__s000__tb131072__v002\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 52.081413668709864, \"nll_base\": 3.952808141708374, \"eval_ref\": \"bbh_fewshot_logical_deduction_three_objects__de5d1... | {
"index": "bbh_fewshot_logical_deduction_three_objects__de5d13ac__s000__tb131072__v002",
"uid": "bbh_fewshot_logical_deduction_three_objects__de5d13ac__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\": \"bbh_fewshot_movie_recommendation__7dae1f8e__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 52.44193433649737, \"nll_base\": 3.9597065448760986, \"eval_ref\": \"bbh_fewshot_movie_recommendation__7dae1f8e__s000__tb262144__v... | {
"index": "bbh_fewshot_movie_recommendation__7dae1f8e__s000__tb262144__v000",
"uid": "bbh_fewshot_movie_recommendation__7dae1f8e__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\": \"bbh_fewshot_movie_recommendation__7dae1f8e__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 52.44193433649737, \"nll_base\": 3.9597065448760986, \"eval_ref\": \"bbh_fewshot_movie_recommendation__7dae1f8e__s000__tb65536__v001... | {
"index": "bbh_fewshot_movie_recommendation__7dae1f8e__s000__tb65536__v001",
"uid": "bbh_fewshot_movie_recommendation__7dae1f8e__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\": \"bbh_fewshot_navigate__23589417__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1422.34638595681, \"nll_base\": 7.260063171386719, \"eval_ref\": \"bbh_fewshot_navigate__23589417__s000__tb131072__v001\", \"oracle_delta\": 0... | {
"index": "bbh_fewshot_navigate__23589417__s000__tb131072__v001",
"uid": "bbh_fewshot_navigate__23589417__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\": \"bbh_fewshot_navigate__23589417__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1422.34638595681, \"nll_base\": 7.260063171386719, \"eval_ref\": \"bbh_fewshot_navigate__23589417__s000__tb65536__v000\", \"oracle_delta\": 0.22... | {
"index": "bbh_fewshot_navigate__23589417__s000__tb65536__v000",
"uid": "bbh_fewshot_navigate__23589417__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\": \"bbh_fewshot_object_counting__400b13dc__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1038.6608206976591, \"nll_base\": 6.945687489966824, \"eval_ref\": \"bbh_fewshot_object_counting__400b13dc__s000__tb131072__v001\", \"o... | {
"index": "bbh_fewshot_object_counting__400b13dc__s000__tb131072__v001",
"uid": "bbh_fewshot_object_counting__400b13dc__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\": \"bbh_fewshot_object_counting__400b13dc__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1038.6608206976591, \"nll_base\": 6.945687489966824, \"eval_ref\": \"bbh_fewshot_object_counting__400b13dc__s000__tb65536__v000\", \"orac... | {
"index": "bbh_fewshot_object_counting__400b13dc__s000__tb65536__v000",
"uid": "bbh_fewshot_object_counting__400b13dc__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\": \"bbh_fewshot_reasoning_about_colored_objects__020084b0__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 100.67941276732245, \"nll_base\": 4.611941337585449, \"eval_ref\": \"bbh_fewshot_reasoning_about_colored_objects__02008... | {
"index": "bbh_fewshot_reasoning_about_colored_objects__020084b0__s000__tb131072__v000",
"uid": "bbh_fewshot_reasoning_about_colored_objects__020084b0__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\": \"bbh_fewshot_tracking_shuffled_objects_five_objects__8908202c__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 46.70009436341746, \"nll_base\": 3.8437461853027344, \"eval_ref\": \"bbh_fewshot_tracking_shuffled_objects_five_... | {
"index": "bbh_fewshot_tracking_shuffled_objects_five_objects__8908202c__s000__tb131072__v001",
"uid": "bbh_fewshot_tracking_shuffled_objects_five_objects__8908202c__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\": \"bbh_fewshot_tracking_shuffled_objects_five_objects__8908202c__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 46.70009436341746, \"nll_base\": 3.8437461853027344, \"eval_ref\": \"bbh_fewshot_tracking_shuffled_objects_five_ob... | {
"index": "bbh_fewshot_tracking_shuffled_objects_five_objects__8908202c__s000__tb65536__v000",
"uid": "bbh_fewshot_tracking_shuffled_objects_five_objects__8908202c__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\": \"bbh_fewshot_tracking_shuffled_objects_three_objects__e33fc02e__s000__tb131072__v002\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 56.14090518965972, \"nll_base\": 4.027864694595337, \"eval_ref\": \"bbh_fewshot_tracking_shuffled_objects_three... | {
"index": "bbh_fewshot_tracking_shuffled_objects_three_objects__e33fc02e__s000__tb131072__v002",
"uid": "bbh_fewshot_tracking_shuffled_objects_three_objects__e33fc02e__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\": \"bbh_fewshot_tracking_shuffled_objects_three_objects__e33fc02e__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 56.14090518965972, \"nll_base\": 4.027864694595337, \"eval_ref\": \"bbh_fewshot_tracking_shuffled_objects_three_o... | {
"index": "bbh_fewshot_tracking_shuffled_objects_three_objects__e33fc02e__s000__tb65536__v001",
"uid": "bbh_fewshot_tracking_shuffled_objects_three_objects__e33fc02e__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\": \"bbh_fewshot_web_of_lies__d7b20a26__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2282.5802951908786, \"nll_base\": 7.733061790466309, \"eval_ref\": \"bbh_fewshot_web_of_lies__d7b20a26__s000__tb131072__v001\", \"oracle_de... | {
"index": "bbh_fewshot_web_of_lies__d7b20a26__s000__tb131072__v001",
"uid": "bbh_fewshot_web_of_lies__d7b20a26__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\": \"bbh_fewshot_web_of_lies__d7b20a26__s000__tb262144__v002\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2282.5802951908786, \"nll_base\": 7.733061790466309, \"eval_ref\": \"bbh_fewshot_web_of_lies__d7b20a26__s000__tb262144__v002\", \"oracle_de... | {
"index": "bbh_fewshot_web_of_lies__d7b20a26__s000__tb262144__v002",
"uid": "bbh_fewshot_web_of_lies__d7b20a26__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\": \"bbh_fewshot_web_of_lies__d7b20a26__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2282.5802951908786, \"nll_base\": 7.733061790466309, \"eval_ref\": \"bbh_fewshot_web_of_lies__d7b20a26__s000__tb65536__v000\", \"oracle_delta... | {
"index": "bbh_fewshot_web_of_lies__d7b20a26__s000__tb65536__v000",
"uid": "bbh_fewshot_web_of_lies__d7b20a26__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\": \"bbh_zeroshot_boolean_expressions__d016b0eb__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1085.3736395762214, \"nll_base\": 6.989679574966431, \"eval_ref\": \"bbh_zeroshot_boolean_expressions__d016b0eb__s000__tb131072__v... | {
"index": "bbh_zeroshot_boolean_expressions__d016b0eb__s000__tb131072__v001",
"uid": "bbh_zeroshot_boolean_expressions__d016b0eb__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\": \"bbh_zeroshot_boolean_expressions__d016b0eb__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1085.3736395762214, \"nll_base\": 6.989679574966431, \"eval_ref\": \"bbh_zeroshot_boolean_expressions__d016b0eb__s000__tb65536__v000... | {
"index": "bbh_zeroshot_boolean_expressions__d016b0eb__s000__tb65536__v000",
"uid": "bbh_zeroshot_boolean_expressions__d016b0eb__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\": \"bbh_zeroshot_causal_judgement__2a0f02e3__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1401.2026647505866, \"nll_base\": 7.245086193084717, \"eval_ref\": \"bbh_zeroshot_causal_judgement__2a0f02e3__s000__tb131072__v000\",... | {
"index": "bbh_zeroshot_causal_judgement__2a0f02e3__s000__tb131072__v000",
"uid": "bbh_zeroshot_causal_judgement__2a0f02e3__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\": \"bbh_zeroshot_date_understanding__c529ef4f__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 70.57609262954203, \"nll_base\": 4.2566914558410645, \"eval_ref\": \"bbh_zeroshot_date_understanding__c529ef4f__s000__tb262144__v00... | {
"index": "bbh_zeroshot_date_understanding__c529ef4f__s000__tb262144__v000",
"uid": "bbh_zeroshot_date_understanding__c529ef4f__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\": \"bbh_zeroshot_date_understanding__c529ef4f__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 70.57609262954203, \"nll_base\": 4.2566914558410645, \"eval_ref\": \"bbh_zeroshot_date_understanding__c529ef4f__s000__tb65536__v001\"... | {
"index": "bbh_zeroshot_date_understanding__c529ef4f__s000__tb65536__v001",
"uid": "bbh_zeroshot_date_understanding__c529ef4f__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\": \"bbh_zeroshot_movie_recommendation__810e8025__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 50.030571746195925, \"nll_base\": 3.912634253501892, \"eval_ref\": \"bbh_zeroshot_movie_recommendation__810e8025__s000__tb262144_... | {
"index": "bbh_zeroshot_movie_recommendation__810e8025__s000__tb262144__v000",
"uid": "bbh_zeroshot_movie_recommendation__810e8025__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\": \"bbh_zeroshot_movie_recommendation__810e8025__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 50.030571746195925, \"nll_base\": 3.912634253501892, \"eval_ref\": \"bbh_zeroshot_movie_recommendation__810e8025__s000__tb65536__v0... | {
"index": "bbh_zeroshot_movie_recommendation__810e8025__s000__tb65536__v001",
"uid": "bbh_zeroshot_movie_recommendation__810e8025__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\": \"bbh_zeroshot_navigate__ee3521c7__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1452.2861943075338, \"nll_base\": 7.2808942794799805, \"eval_ref\": \"bbh_zeroshot_navigate__ee3521c7__s000__tb65536__v001\", \"oracle_delta\":... | {
"index": "bbh_zeroshot_navigate__ee3521c7__s000__tb65536__v001",
"uid": "bbh_zeroshot_navigate__ee3521c7__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\": \"bbh_zeroshot_object_counting__6029a678__s000__tb131072__v002\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1030.1551231235364, \"nll_base\": 6.937464674858198, \"eval_ref\": \"bbh_zeroshot_object_counting__6029a678__s000__tb131072__v002\", \... | {
"index": "bbh_zeroshot_object_counting__6029a678__s000__tb131072__v002",
"uid": "bbh_zeroshot_object_counting__6029a678__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\": \"bbh_zeroshot_tracking_shuffled_objects_five_objects__04ddcd21__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 50.968499043151105, \"nll_base\": 3.931207776069641, \"eval_ref\": \"bbh_zeroshot_tracking_shuffled_objects_fiv... | {
"index": "bbh_zeroshot_tracking_shuffled_objects_five_objects__04ddcd21__s000__tb131072__v000",
"uid": "bbh_zeroshot_tracking_shuffled_objects_five_objects__04ddcd21__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\": \"bbh_zeroshot_tracking_shuffled_objects_five_objects__04ddcd21__s000__tb65536__v002\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 50.968499043151105, \"nll_base\": 3.931207776069641, \"eval_ref\": \"bbh_zeroshot_tracking_shuffled_objects_five_... | {
"index": "bbh_zeroshot_tracking_shuffled_objects_five_objects__04ddcd21__s000__tb65536__v002",
"uid": "bbh_zeroshot_tracking_shuffled_objects_five_objects__04ddcd21__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\": \"bbh_zeroshot_tracking_shuffled_objects_seven_objects__4b2ca405__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 67.26457080577899, \"nll_base\": 4.208633661270142, \"eval_ref\": \"bbh_zeroshot_tracking_shuffled_objects_sev... | {
"index": "bbh_zeroshot_tracking_shuffled_objects_seven_objects__4b2ca405__s000__tb131072__v001",
"uid": "bbh_zeroshot_tracking_shuffled_objects_seven_objects__4b2ca405__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\": \"bbh_zeroshot_tracking_shuffled_objects_three_objects__19b90ffa__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 56.75627061672151, \"nll_base\": 4.038766145706177, \"eval_ref\": \"bbh_zeroshot_tracking_shuffled_objects_thr... | {
"index": "bbh_zeroshot_tracking_shuffled_objects_three_objects__19b90ffa__s000__tb131072__v001",
"uid": "bbh_zeroshot_tracking_shuffled_objects_three_objects__19b90ffa__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\": \"bbh_zeroshot_tracking_shuffled_objects_three_objects__19b90ffa__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 56.75627061672151, \"nll_base\": 4.038766145706177, \"eval_ref\": \"bbh_zeroshot_tracking_shuffled_objects_three... | {
"index": "bbh_zeroshot_tracking_shuffled_objects_three_objects__19b90ffa__s000__tb65536__v000",
"uid": "bbh_zeroshot_tracking_shuffled_objects_three_objects__19b90ffa__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\": \"bbh_zeroshot_web_of_lies__7142a317__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2349.5091160523907, \"nll_base\": 7.7619616985321045, \"eval_ref\": \"bbh_zeroshot_web_of_lies__7142a317__s000__tb262144__v000\", \"oracle... | {
"index": "bbh_zeroshot_web_of_lies__7142a317__s000__tb262144__v000",
"uid": "bbh_zeroshot_web_of_lies__7142a317__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\": \"bbh_zeroshot_web_of_lies__7142a317__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2349.5091160523907, \"nll_base\": 7.7619616985321045, \"eval_ref\": \"bbh_zeroshot_web_of_lies__7142a317__s000__tb65536__v001\", \"oracle_de... | {
"index": "bbh_zeroshot_web_of_lies__7142a317__s000__tb65536__v001",
"uid": "bbh_zeroshot_web_of_lies__7142a317__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\": \"bbq_ambig__9ff79e01__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 523.8625164232577, \"nll_base\": 6.2612292766571045, \"eval_ref\": \"bbq_ambig__9ff79e01__s000__tb131072__v000\", \"oracle_delta\": 0.2766683101654053, \... | {
"index": "bbq_ambig__9ff79e01__s000__tb131072__v000",
"uid": "bbq_ambig__9ff79e01__s000__tb131072__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"bbq_generate__d40b6981__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 94.44076792757797, \"nll_base\": 4.54797284356479, \"eval_ref\": \"bbq_generate__d40b6981__s000__tb131072__v000\", \"oracle_delta\": 0.286459824134563... | {
"index": "bbq_generate__d40b6981__s000__tb131072__v000",
"uid": "bbq_generate__d40b6981__s000__tb131072__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"bbq_generate_ambig__bab292e9__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3991.5845050368607, \"nll_base\": 8.291943550109863, \"eval_ref\": \"bbq_generate_ambig__bab292e9__s000__tb65536__v000\", \"oracle_delta\": 0.1278... | {
"index": "bbq_generate_ambig__bab292e9__s000__tb65536__v000",
"uid": "bbq_generate_ambig__bab292e9__s000__tb65536__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"belebele_afr_prompt_2__2a1947bc__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2633.4850100009376, \"nll_base\": 7.876063346862793, \"eval_ref\": \"belebele_afr_prompt_2__2a1947bc__s000__tb262144__v000\", \"oracle_delta\... | {
"index": "belebele_afr_prompt_2__2a1947bc__s000__tb262144__v000",
"uid": "belebele_afr_prompt_2__2a1947bc__s000__tb262144__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"belebele_arb_Arab__eb1e0b7d__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2967.4835116546014, \"nll_base\": 7.995469570159912, \"eval_ref\": \"belebele_arb_Arab__eb1e0b7d__s000__tb65536__v000\", \"oracle_delta\": 0.165464... | {
"index": "belebele_arb_Arab__eb1e0b7d__s000__tb65536__v000",
"uid": "belebele_arb_Arab__eb1e0b7d__s000__tb65536__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"belebele_arz_prompt_4__ca540ddd__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 428.4632554379486, \"nll_base\": 6.060204982757568, \"eval_ref\": \"belebele_arz_prompt_4__ca540ddd__s000__tb65536__v000\", \"oracle_delta\": 0... | {
"index": "belebele_arz_prompt_4__ca540ddd__s000__tb65536__v000",
"uid": "belebele_arz_prompt_4__ca540ddd__s000__tb65536__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"belebele_bam_prompt_1__d877666b__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3902.4825141164292, \"nll_base\": 8.269368171691895, \"eval_ref\": \"belebele_bam_prompt_1__d877666b__s000__tb65536__v000\", \"oracle_delta\": ... | {
"index": "belebele_bam_prompt_1__d877666b__s000__tb65536__v000",
"uid": "belebele_bam_prompt_1__d877666b__s000__tb65536__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"belebele_cat_Latn__63b6b72c__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4131.551933650057, \"nll_base\": 8.326408386230469, \"eval_ref\": \"belebele_cat_Latn__63b6b72c__s000__tb131072__v000\", \"oracle_delta\": 1.5115... | {
"index": "belebele_cat_Latn__63b6b72c__s000__tb131072__v000",
"uid": "belebele_cat_Latn__63b6b72c__s000__tb131072__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"belebele_fuv_prompt_5__32117c4c__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1227.8248143909207, \"nll_base\": 7.112999439239502, \"eval_ref\": \"belebele_fuv_prompt_5__32117c4c__s000__tb65536__v000\", \"oracle_delta\": ... | {
"index": "belebele_fuv_prompt_5__32117c4c__s000__tb65536__v000",
"uid": "belebele_fuv_prompt_5__32117c4c__s000__tb65536__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"belebele_gaz_prompt_3__cdb6d15c__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4595.981538372337, \"nll_base\": 8.432937622070312, \"eval_ref\": \"belebele_gaz_prompt_3__cdb6d15c__s000__tb65536__v000\", \"oracle_delta\": 0... | {
"index": "belebele_gaz_prompt_3__cdb6d15c__s000__tb65536__v000",
"uid": "belebele_gaz_prompt_3__cdb6d15c__s000__tb65536__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"belebele_hau_prompt_1__9e317e10__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3224.765990254506, \"nll_base\": 8.078615665435791, \"eval_ref\": \"belebele_hau_prompt_1__9e317e10__s000__tb262144__v000\", \"oracle_delta\"... | {
"index": "belebele_hau_prompt_1__9e317e10__s000__tb262144__v000",
"uid": "belebele_hau_prompt_1__9e317e10__s000__tb262144__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"belebele_hau_prompt_3__340ec6f7__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 7428.22392792898, \"nll_base\": 8.913042068481445, \"eval_ref\": \"belebele_hau_prompt_3__340ec6f7__s000__tb131072__v000\", \"oracle_delta\":... | {
"index": "belebele_hau_prompt_3__340ec6f7__s000__tb131072__v000",
"uid": "belebele_hau_prompt_3__340ec6f7__s000__tb131072__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"belebele_hin_Latn__6bf5a28e__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4692.975690937157, \"nll_base\": 8.453822135925293, \"eval_ref\": \"belebele_hin_Latn__6bf5a28e__s000__tb65536__v000\", \"oracle_delta\": 0.3192219... | {
"index": "belebele_hin_Latn__6bf5a28e__s000__tb65536__v000",
"uid": "belebele_hin_Latn__6bf5a28e__s000__tb65536__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"belebele_ibo_Latn__f45783cd__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1997.3850243537022, \"nll_base\": 7.5995941162109375, \"eval_ref\": \"belebele_ibo_Latn__f45783cd__s000__tb65536__v000\", \"oracle_delta\": 1.05011... | {
"index": "belebele_ibo_Latn__f45783cd__s000__tb65536__v000",
"uid": "belebele_ibo_Latn__f45783cd__s000__tb65536__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"belebele_kaz_Cyrl__66577a85__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2024.3225479246573, \"nll_base\": 7.612990379333496, \"eval_ref\": \"belebele_kaz_Cyrl__66577a85__s000__tb131072__v000\", \"oracle_delta\": 0.370... | {
"index": "belebele_kaz_Cyrl__66577a85__s000__tb131072__v000",
"uid": "belebele_kaz_Cyrl__66577a85__s000__tb131072__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"belebele_kea_prompt_5__b4c6307b__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 944.8627209035182, \"nll_base\": 6.85103964805603, \"eval_ref\": \"belebele_kea_prompt_5__b4c6307b__s000__tb131072__v000\", \"oracle_delta\":... | {
"index": "belebele_kea_prompt_5__b4c6307b__s000__tb131072__v000",
"uid": "belebele_kea_prompt_5__b4c6307b__s000__tb131072__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"belebele_kin_Latn__5c58cd66__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3760.6120334295624, \"nll_base\": 8.23233699798584, \"eval_ref\": \"belebele_kin_Latn__5c58cd66__s000__tb65536__v000\", \"oracle_delta\": 1.0881841... | {
"index": "belebele_kin_Latn__5c58cd66__s000__tb65536__v000",
"uid": "belebele_kin_Latn__5c58cd66__s000__tb65536__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"belebele_lao_Laoo__b7044a99__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 493.34375717489075, \"nll_base\": 6.201206207275391, \"eval_ref\": \"belebele_lao_Laoo__b7044a99__s000__tb131072__v000\", \"oracle_delta\": 0.353... | {
"index": "belebele_lao_Laoo__b7044a99__s000__tb131072__v000",
"uid": "belebele_lao_Laoo__b7044a99__s000__tb131072__v000",
"axis": "multilingual",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.