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\": \"belebele_lin_prompt_4__44e05a69__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 924.2242127112627, \"nll_base\": 6.828954696655273, \"eval_ref\": \"belebele_lin_prompt_4__44e05a69__s000__tb262144__v000\", \"oracle_delta\"... | {
"index": "belebele_lin_prompt_4__44e05a69__s000__tb262144__v000",
"uid": "belebele_lin_prompt_4__44e05a69__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_lug_prompt_4__a206f75e__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1071.5712043193441, \"nll_base\": 6.976881265640259, \"eval_ref\": \"belebele_lug_prompt_4__a206f75e__s000__tb131072__v000\", \"oracle_delta\... | {
"index": "belebele_lug_prompt_4__a206f75e__s000__tb131072__v000",
"uid": "belebele_lug_prompt_4__a206f75e__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_lug_prompt_5__a7eaba83__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 895.1711702870331, \"nll_base\": 6.797014951705933, \"eval_ref\": \"belebele_lug_prompt_5__a7eaba83__s000__tb65536__v000\", \"oracle_delta\": 0... | {
"index": "belebele_lug_prompt_5__a7eaba83__s000__tb65536__v000",
"uid": "belebele_lug_prompt_5__a7eaba83__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_mkd_Cyrl__8396ed54__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3382.490559201255, \"nll_base\": 8.126367568969727, \"eval_ref\": \"belebele_mkd_Cyrl__8396ed54__s000__tb65536__v000\", \"oracle_delta\": 0.3599829... | {
"index": "belebele_mkd_Cyrl__8396ed54__s000__tb65536__v000",
"uid": "belebele_mkd_Cyrl__8396ed54__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_nya_prompt_3__069cbe0c__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 5518.074016002768, \"nll_base\": 8.615784168243408, \"eval_ref\": \"belebele_nya_prompt_3__069cbe0c__s000__tb65536__v000\", \"oracle_delta\": 0... | {
"index": "belebele_nya_prompt_3__069cbe0c__s000__tb65536__v000",
"uid": "belebele_nya_prompt_3__069cbe0c__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_plt_prompt_4__f08d47bd__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 782.9394951758469, \"nll_base\": 6.663055419921875, \"eval_ref\": \"belebele_plt_prompt_4__f08d47bd__s000__tb65536__v000\", \"oracle_delta\": 0... | {
"index": "belebele_plt_prompt_4__f08d47bd__s000__tb65536__v000",
"uid": "belebele_plt_prompt_4__f08d47bd__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_pol_Latn__ef872840__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3802.5612720805566, \"nll_base\": 8.243430137634277, \"eval_ref\": \"belebele_pol_Latn__ef872840__s000__tb131072__v000\", \"oracle_delta\": 1.929... | {
"index": "belebele_pol_Latn__ef872840__s000__tb131072__v000",
"uid": "belebele_pol_Latn__ef872840__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_por_prompt_3__11c55665__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 10893.45069703235, \"nll_base\": 9.29591703414917, \"eval_ref\": \"belebele_por_prompt_3__11c55665__s000__tb131072__v000\", \"oracle_delta\":... | {
"index": "belebele_por_prompt_3__11c55665__s000__tb131072__v000",
"uid": "belebele_por_prompt_3__11c55665__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_ron_Latn__9ad6a601__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2446.773557468351, \"nll_base\": 7.802525520324707, \"eval_ref\": \"belebele_ron_Latn__9ad6a601__s000__tb262144__v000\", \"oracle_delta\": 2.6342... | {
"index": "belebele_ron_Latn__9ad6a601__s000__tb262144__v000",
"uid": "belebele_ron_Latn__9ad6a601__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_sot_prompt_1__150eaecc__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3899.3631182814265, \"nll_base\": 8.268568515777588, \"eval_ref\": \"belebele_sot_prompt_1__150eaecc__s000__tb65536__v000\", \"oracle_delta\": ... | {
"index": "belebele_sot_prompt_1__150eaecc__s000__tb65536__v000",
"uid": "belebele_sot_prompt_1__150eaecc__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_sot_prompt_2__c45c7f22__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1782.8824894299198, \"nll_base\": 7.485986709594727, \"eval_ref\": \"belebele_sot_prompt_2__c45c7f22__s000__tb65536__v000\", \"oracle_delta\": ... | {
"index": "belebele_sot_prompt_2__c45c7f22__s000__tb65536__v000",
"uid": "belebele_sot_prompt_2__c45c7f22__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_sot_prompt_5__962938aa__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 911.6391089109238, \"nll_base\": 6.815244197845459, \"eval_ref\": \"belebele_sot_prompt_5__962938aa__s000__tb65536__v000\", \"oracle_delta\": 0... | {
"index": "belebele_sot_prompt_5__962938aa__s000__tb65536__v000",
"uid": "belebele_sot_prompt_5__962938aa__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_swe_Latn__99a7c912__s000__tb131072__v002\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4177.30033319053, \"nll_base\": 8.337420463562012, \"eval_ref\": \"belebele_swe_Latn__99a7c912__s000__tb131072__v002\", \"oracle_delta\": 1.91816... | {
"index": "belebele_swe_Latn__99a7c912__s000__tb131072__v002",
"uid": "belebele_swe_Latn__99a7c912__s000__tb131072__v002",
"axis": "code",
"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_swe_Latn__99a7c912__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 4177.30033319053, \"nll_base\": 8.337420463562012, \"eval_ref\": \"belebele_swe_Latn__99a7c912__s000__tb262144__v000\", \"oracle_delta\": 2.52028... | {
"index": "belebele_swe_Latn__99a7c912__s000__tb262144__v000",
"uid": "belebele_swe_Latn__99a7c912__s000__tb262144__v000",
"axis": "code",
"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_swe_Latn__99a7c912__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4177.30033319053, \"nll_base\": 8.337420463562012, \"eval_ref\": \"belebele_swe_Latn__99a7c912__s000__tb65536__v001\", \"oracle_delta\": 0.44490265... | {
"index": "belebele_swe_Latn__99a7c912__s000__tb65536__v001",
"uid": "belebele_swe_Latn__99a7c912__s000__tb65536__v001",
"axis": "code",
"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_swe_Latn__99a7c912__s001__tb131072__v005\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3519.7469296884915, \"nll_base\": 8.166144371032715, \"eval_ref\": \"belebele_swe_Latn__99a7c912__s001__tb131072__v005\", \"oracle_delta\": 1.907... | {
"index": "belebele_swe_Latn__99a7c912__s001__tb131072__v005",
"uid": "belebele_swe_Latn__99a7c912__s001__tb131072__v005",
"axis": "code",
"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_swe_Latn__99a7c912__s001__tb262144__v003\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3519.7469296884915, \"nll_base\": 8.166144371032715, \"eval_ref\": \"belebele_swe_Latn__99a7c912__s001__tb262144__v003\", \"oracle_delta\": 2.504... | {
"index": "belebele_swe_Latn__99a7c912__s001__tb262144__v003",
"uid": "belebele_swe_Latn__99a7c912__s001__tb262144__v003",
"axis": "code",
"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_swe_Latn__99a7c912__s001__tb65536__v004\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3519.7469296884915, \"nll_base\": 8.166144371032715, \"eval_ref\": \"belebele_swe_Latn__99a7c912__s001__tb65536__v004\", \"oracle_delta\": 0.576682... | {
"index": "belebele_swe_Latn__99a7c912__s001__tb65536__v004",
"uid": "belebele_swe_Latn__99a7c912__s001__tb65536__v004",
"axis": "code",
"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_swe_Latn__99a7c912__s002__tb131072__v008\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4479.901157945432, \"nll_base\": 8.407356262207031, \"eval_ref\": \"belebele_swe_Latn__99a7c912__s002__tb131072__v008\", \"oracle_delta\": 1.9235... | {
"index": "belebele_swe_Latn__99a7c912__s002__tb131072__v008",
"uid": "belebele_swe_Latn__99a7c912__s002__tb131072__v008",
"axis": "code",
"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_swe_Latn__99a7c912__s002__tb262144__v006\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 4479.901157945432, \"nll_base\": 8.407356262207031, \"eval_ref\": \"belebele_swe_Latn__99a7c912__s002__tb262144__v006\", \"oracle_delta\": 2.5793... | {
"index": "belebele_swe_Latn__99a7c912__s002__tb262144__v006",
"uid": "belebele_swe_Latn__99a7c912__s002__tb262144__v006",
"axis": "code",
"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_swe_Latn__99a7c912__s002__tb65536__v007\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4479.901157945432, \"nll_base\": 8.407356262207031, \"eval_ref\": \"belebele_swe_Latn__99a7c912__s002__tb65536__v007\", \"oracle_delta\": 0.5393013... | {
"index": "belebele_swe_Latn__99a7c912__s002__tb65536__v007",
"uid": "belebele_swe_Latn__99a7c912__s002__tb65536__v007",
"axis": "code",
"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_swe_Latn__99a7c912__s003__tb131072__v011\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4691.279756271586, \"nll_base\": 8.453460693359375, \"eval_ref\": \"belebele_swe_Latn__99a7c912__s003__tb131072__v011\", \"oracle_delta\": 1.9012... | {
"index": "belebele_swe_Latn__99a7c912__s003__tb131072__v011",
"uid": "belebele_swe_Latn__99a7c912__s003__tb131072__v011",
"axis": "code",
"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_swe_Latn__99a7c912__s003__tb262144__v009\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 4691.279756271586, \"nll_base\": 8.453460693359375, \"eval_ref\": \"belebele_swe_Latn__99a7c912__s003__tb262144__v009\", \"oracle_delta\": 2.4735... | {
"index": "belebele_swe_Latn__99a7c912__s003__tb262144__v009",
"uid": "belebele_swe_Latn__99a7c912__s003__tb262144__v009",
"axis": "code",
"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_swe_Latn__99a7c912__s003__tb65536__v010\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4691.279756271586, \"nll_base\": 8.453460693359375, \"eval_ref\": \"belebele_swe_Latn__99a7c912__s003__tb65536__v010\", \"oracle_delta\": 0.6348664... | {
"index": "belebele_swe_Latn__99a7c912__s003__tb65536__v010",
"uid": "belebele_swe_Latn__99a7c912__s003__tb65536__v010",
"axis": "code",
"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_swe_Latn__99a7c912__s004__tb131072__v014\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4100.719607484626, \"nll_base\": 8.318917751312256, \"eval_ref\": \"belebele_swe_Latn__99a7c912__s004__tb131072__v014\", \"oracle_delta\": 1.9912... | {
"index": "belebele_swe_Latn__99a7c912__s004__tb131072__v014",
"uid": "belebele_swe_Latn__99a7c912__s004__tb131072__v014",
"axis": "code",
"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_swe_Latn__99a7c912__s004__tb262144__v012\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 4100.719607484626, \"nll_base\": 8.318917751312256, \"eval_ref\": \"belebele_swe_Latn__99a7c912__s004__tb262144__v012\", \"oracle_delta\": 2.5367... | {
"index": "belebele_swe_Latn__99a7c912__s004__tb262144__v012",
"uid": "belebele_swe_Latn__99a7c912__s004__tb262144__v012",
"axis": "code",
"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_swe_Latn__99a7c912__s004__tb65536__v013\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4100.719607484626, \"nll_base\": 8.318917751312256, \"eval_ref\": \"belebele_swe_Latn__99a7c912__s004__tb65536__v013\", \"oracle_delta\": 0.5457544... | {
"index": "belebele_swe_Latn__99a7c912__s004__tb65536__v013",
"uid": "belebele_swe_Latn__99a7c912__s004__tb65536__v013",
"axis": "code",
"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_swe_Latn__99a7c912__s005__tb131072__v017\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4102.088598785943, \"nll_base\": 8.319251537322998, \"eval_ref\": \"belebele_swe_Latn__99a7c912__s005__tb131072__v017\", \"oracle_delta\": 1.8382... | {
"index": "belebele_swe_Latn__99a7c912__s005__tb131072__v017",
"uid": "belebele_swe_Latn__99a7c912__s005__tb131072__v017",
"axis": "code",
"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_swe_Latn__99a7c912__s005__tb262144__v015\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 4102.088598785943, \"nll_base\": 8.319251537322998, \"eval_ref\": \"belebele_swe_Latn__99a7c912__s005__tb262144__v015\", \"oracle_delta\": 2.5118... | {
"index": "belebele_swe_Latn__99a7c912__s005__tb262144__v015",
"uid": "belebele_swe_Latn__99a7c912__s005__tb262144__v015",
"axis": "code",
"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_swe_Latn__99a7c912__s005__tb65536__v016\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4102.088598785943, \"nll_base\": 8.319251537322998, \"eval_ref\": \"belebele_swe_Latn__99a7c912__s005__tb65536__v016\", \"oracle_delta\": 0.4654691... | {
"index": "belebele_swe_Latn__99a7c912__s005__tb65536__v016",
"uid": "belebele_swe_Latn__99a7c912__s005__tb65536__v016",
"axis": "code",
"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_swe_Latn__99a7c912__s006__tb131072__v020\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4159.905603745719, \"nll_base\": 8.333247661590576, \"eval_ref\": \"belebele_swe_Latn__99a7c912__s006__tb131072__v020\", \"oracle_delta\": 1.9158... | {
"index": "belebele_swe_Latn__99a7c912__s006__tb131072__v020",
"uid": "belebele_swe_Latn__99a7c912__s006__tb131072__v020",
"axis": "code",
"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_swe_Latn__99a7c912__s006__tb262144__v018\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 4159.905603745719, \"nll_base\": 8.333247661590576, \"eval_ref\": \"belebele_swe_Latn__99a7c912__s006__tb262144__v018\", \"oracle_delta\": 2.5804... | {
"index": "belebele_swe_Latn__99a7c912__s006__tb262144__v018",
"uid": "belebele_swe_Latn__99a7c912__s006__tb262144__v018",
"axis": "code",
"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_swe_Latn__99a7c912__s006__tb65536__v019\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4159.905603745719, \"nll_base\": 8.333247661590576, \"eval_ref\": \"belebele_swe_Latn__99a7c912__s006__tb65536__v019\", \"oracle_delta\": 0.4740722... | {
"index": "belebele_swe_Latn__99a7c912__s006__tb65536__v019",
"uid": "belebele_swe_Latn__99a7c912__s006__tb65536__v019",
"axis": "code",
"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_swe_Latn__99a7c912__s007__tb131072__v023\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4300.932600563905, \"nll_base\": 8.366587162017822, \"eval_ref\": \"belebele_swe_Latn__99a7c912__s007__tb131072__v023\", \"oracle_delta\": 1.8117... | {
"index": "belebele_swe_Latn__99a7c912__s007__tb131072__v023",
"uid": "belebele_swe_Latn__99a7c912__s007__tb131072__v023",
"axis": "code",
"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_swe_Latn__99a7c912__s007__tb262144__v021\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 4300.932600563905, \"nll_base\": 8.366587162017822, \"eval_ref\": \"belebele_swe_Latn__99a7c912__s007__tb262144__v021\", \"oracle_delta\": 2.5754... | {
"index": "belebele_swe_Latn__99a7c912__s007__tb262144__v021",
"uid": "belebele_swe_Latn__99a7c912__s007__tb262144__v021",
"axis": "code",
"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_swe_Latn__99a7c912__s007__tb65536__v022\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4300.932600563905, \"nll_base\": 8.366587162017822, \"eval_ref\": \"belebele_swe_Latn__99a7c912__s007__tb65536__v022\", \"oracle_delta\": 0.4703433... | {
"index": "belebele_swe_Latn__99a7c912__s007__tb65536__v022",
"uid": "belebele_swe_Latn__99a7c912__s007__tb65536__v022",
"axis": "code",
"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_swe_Latn__99a7c912__s008__tb131072__v026\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4037.4441632873713, \"nll_base\": 8.303367137908936, \"eval_ref\": \"belebele_swe_Latn__99a7c912__s008__tb131072__v026\", \"oracle_delta\": 1.989... | {
"index": "belebele_swe_Latn__99a7c912__s008__tb131072__v026",
"uid": "belebele_swe_Latn__99a7c912__s008__tb131072__v026",
"axis": "code",
"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_swe_Latn__99a7c912__s008__tb262144__v024\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 4037.4441632873713, \"nll_base\": 8.303367137908936, \"eval_ref\": \"belebele_swe_Latn__99a7c912__s008__tb262144__v024\", \"oracle_delta\": 2.568... | {
"index": "belebele_swe_Latn__99a7c912__s008__tb262144__v024",
"uid": "belebele_swe_Latn__99a7c912__s008__tb262144__v024",
"axis": "code",
"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_swe_Latn__99a7c912__s008__tb65536__v025\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4037.4441632873713, \"nll_base\": 8.303367137908936, \"eval_ref\": \"belebele_swe_Latn__99a7c912__s008__tb65536__v025\", \"oracle_delta\": 0.581380... | {
"index": "belebele_swe_Latn__99a7c912__s008__tb65536__v025",
"uid": "belebele_swe_Latn__99a7c912__s008__tb65536__v025",
"axis": "code",
"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_swe_Latn__99a7c912__s009__tb131072__v029\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3937.6312567060518, \"nll_base\": 8.278334617614746, \"eval_ref\": \"belebele_swe_Latn__99a7c912__s009__tb131072__v029\", \"oracle_delta\": 1.831... | {
"index": "belebele_swe_Latn__99a7c912__s009__tb131072__v029",
"uid": "belebele_swe_Latn__99a7c912__s009__tb131072__v029",
"axis": "code",
"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_swe_Latn__99a7c912__s009__tb262144__v027\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3937.6312567060518, \"nll_base\": 8.278334617614746, \"eval_ref\": \"belebele_swe_Latn__99a7c912__s009__tb262144__v027\", \"oracle_delta\": 2.491... | {
"index": "belebele_swe_Latn__99a7c912__s009__tb262144__v027",
"uid": "belebele_swe_Latn__99a7c912__s009__tb262144__v027",
"axis": "code",
"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_swe_Latn__99a7c912__s009__tb65536__v028\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3937.6312567060518, \"nll_base\": 8.278334617614746, \"eval_ref\": \"belebele_swe_Latn__99a7c912__s009__tb65536__v028\", \"oracle_delta\": 0.604516... | {
"index": "belebele_swe_Latn__99a7c912__s009__tb65536__v028",
"uid": "belebele_swe_Latn__99a7c912__s009__tb65536__v028",
"axis": "code",
"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_swe_Latn__99a7c912__s010__tb131072__v032\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4160.859823600315, \"nll_base\": 8.333477020263672, \"eval_ref\": \"belebele_swe_Latn__99a7c912__s010__tb131072__v032\", \"oracle_delta\": 1.8057... | {
"index": "belebele_swe_Latn__99a7c912__s010__tb131072__v032",
"uid": "belebele_swe_Latn__99a7c912__s010__tb131072__v032",
"axis": "code",
"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_swe_Latn__99a7c912__s010__tb262144__v030\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 4160.859823600315, \"nll_base\": 8.333477020263672, \"eval_ref\": \"belebele_swe_Latn__99a7c912__s010__tb262144__v030\", \"oracle_delta\": 2.6613... | {
"index": "belebele_swe_Latn__99a7c912__s010__tb262144__v030",
"uid": "belebele_swe_Latn__99a7c912__s010__tb262144__v030",
"axis": "code",
"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_swe_Latn__99a7c912__s010__tb65536__v031\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4160.859823600315, \"nll_base\": 8.333477020263672, \"eval_ref\": \"belebele_swe_Latn__99a7c912__s010__tb65536__v031\", \"oracle_delta\": 0.6447603... | {
"index": "belebele_swe_Latn__99a7c912__s010__tb65536__v031",
"uid": "belebele_swe_Latn__99a7c912__s010__tb65536__v031",
"axis": "code",
"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_swe_Latn__99a7c912__s011__tb131072__v035\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4069.100528822412, \"nll_base\": 8.311177253723145, \"eval_ref\": \"belebele_swe_Latn__99a7c912__s011__tb131072__v035\", \"oracle_delta\": 2.0327... | {
"index": "belebele_swe_Latn__99a7c912__s011__tb131072__v035",
"uid": "belebele_swe_Latn__99a7c912__s011__tb131072__v035",
"axis": "code",
"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_swe_Latn__99a7c912__s011__tb262144__v033\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 4069.100528822412, \"nll_base\": 8.311177253723145, \"eval_ref\": \"belebele_swe_Latn__99a7c912__s011__tb262144__v033\", \"oracle_delta\": 2.6689... | {
"index": "belebele_swe_Latn__99a7c912__s011__tb262144__v033",
"uid": "belebele_swe_Latn__99a7c912__s011__tb262144__v033",
"axis": "code",
"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_swe_Latn__99a7c912__s011__tb65536__v034\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4069.100528822412, \"nll_base\": 8.311177253723145, \"eval_ref\": \"belebele_swe_Latn__99a7c912__s011__tb65536__v034\", \"oracle_delta\": 0.7707824... | {
"index": "belebele_swe_Latn__99a7c912__s011__tb65536__v034",
"uid": "belebele_swe_Latn__99a7c912__s011__tb65536__v034",
"axis": "code",
"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_swe_Latn__99a7c912__s012__tb131072__v038\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4425.412234221412, \"nll_base\": 8.395118713378906, \"eval_ref\": \"belebele_swe_Latn__99a7c912__s012__tb131072__v038\", \"oracle_delta\": 1.8764... | {
"index": "belebele_swe_Latn__99a7c912__s012__tb131072__v038",
"uid": "belebele_swe_Latn__99a7c912__s012__tb131072__v038",
"axis": "code",
"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_swe_Latn__99a7c912__s012__tb262144__v036\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 4425.412234221412, \"nll_base\": 8.395118713378906, \"eval_ref\": \"belebele_swe_Latn__99a7c912__s012__tb262144__v036\", \"oracle_delta\": 2.5494... | {
"index": "belebele_swe_Latn__99a7c912__s012__tb262144__v036",
"uid": "belebele_swe_Latn__99a7c912__s012__tb262144__v036",
"axis": "code",
"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_swe_Latn__99a7c912__s012__tb65536__v037\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4425.412234221412, \"nll_base\": 8.395118713378906, \"eval_ref\": \"belebele_swe_Latn__99a7c912__s012__tb65536__v037\", \"oracle_delta\": 0.4826874... | {
"index": "belebele_swe_Latn__99a7c912__s012__tb65536__v037",
"uid": "belebele_swe_Latn__99a7c912__s012__tb65536__v037",
"axis": "code",
"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_swe_Latn__99a7c912__s013__tb131072__v041\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4758.476526032614, \"nll_base\": 8.467682838439941, \"eval_ref\": \"belebele_swe_Latn__99a7c912__s013__tb131072__v041\", \"oracle_delta\": 1.9795... | {
"index": "belebele_swe_Latn__99a7c912__s013__tb131072__v041",
"uid": "belebele_swe_Latn__99a7c912__s013__tb131072__v041",
"axis": "code",
"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_swe_Latn__99a7c912__s013__tb262144__v039\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 4758.476526032614, \"nll_base\": 8.467682838439941, \"eval_ref\": \"belebele_swe_Latn__99a7c912__s013__tb262144__v039\", \"oracle_delta\": 2.8181... | {
"index": "belebele_swe_Latn__99a7c912__s013__tb262144__v039",
"uid": "belebele_swe_Latn__99a7c912__s013__tb262144__v039",
"axis": "code",
"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_swe_Latn__99a7c912__s013__tb65536__v040\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4758.476526032614, \"nll_base\": 8.467682838439941, \"eval_ref\": \"belebele_swe_Latn__99a7c912__s013__tb65536__v040\", \"oracle_delta\": 0.5572638... | {
"index": "belebele_swe_Latn__99a7c912__s013__tb65536__v040",
"uid": "belebele_swe_Latn__99a7c912__s013__tb65536__v040",
"axis": "code",
"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_swe_Latn__99a7c912__s014__tb262144__v042\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3789.378621829852, \"nll_base\": 8.239957332611084, \"eval_ref\": \"belebele_swe_Latn__99a7c912__s014__tb262144__v042\", \"oracle_delta\": 2.5673... | {
"index": "belebele_swe_Latn__99a7c912__s014__tb262144__v042",
"uid": "belebele_swe_Latn__99a7c912__s014__tb262144__v042",
"axis": "code",
"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_tgk_Cyrl__6405d54e__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2304.2066787402464, \"nll_base\": 7.742491722106934, \"eval_ref\": \"belebele_tgk_Cyrl__6405d54e__s000__tb65536__v000\", \"oracle_delta\": 0.296438... | {
"index": "belebele_tgk_Cyrl__6405d54e__s000__tb65536__v000",
"uid": "belebele_tgk_Cyrl__6405d54e__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_war_Latn__80d61ff6__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3916.969994259423, \"nll_base\": 8.273073673248291, \"eval_ref\": \"belebele_war_Latn__80d61ff6__s000__tb131072__v000\", \"oracle_delta\": 1.9100... | {
"index": "belebele_war_Latn__80d61ff6__s000__tb131072__v000",
"uid": "belebele_war_Latn__80d61ff6__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_xho_prompt_2__1ffb6ef4__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1764.375078753647, \"nll_base\": 7.4755518436431885, \"eval_ref\": \"belebele_xho_prompt_2__1ffb6ef4__s000__tb65536__v000\", \"oracle_delta\": ... | {
"index": "belebele_xho_prompt_2__1ffb6ef4__s000__tb65536__v000",
"uid": "belebele_xho_prompt_2__1ffb6ef4__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_xho_prompt_4__c512863f__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 799.0088489381742, \"nll_base\": 6.6833720207214355, \"eval_ref\": \"belebele_xho_prompt_4__c512863f__s000__tb65536__v000\", \"oracle_delta\": ... | {
"index": "belebele_xho_prompt_4__c512863f__s000__tb65536__v000",
"uid": "belebele_xho_prompt_4__c512863f__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_zsm_Latn__df6c414c__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 5729.481941305819, \"nll_base\": 8.653380393981934, \"eval_ref\": \"belebele_zsm_Latn__df6c414c__s000__tb262144__v000\", \"oracle_delta\": 3.1348... | {
"index": "belebele_zsm_Latn__df6c414c__s000__tb262144__v000",
"uid": "belebele_zsm_Latn__df6c414c__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_zul_prompt_3__82289161__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 5434.226386623108, \"nll_base\": 8.600472450256348, \"eval_ref\": \"belebele_zul_prompt_3__82289161__s000__tb131072__v000\", \"oracle_delta\"... | {
"index": "belebele_zul_prompt_3__82289161__s000__tb131072__v000",
"uid": "belebele_zul_prompt_3__82289161__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\": \"bertaqa_en_mt_llama-2-70b__a01d57d3__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 5817.469116247445, \"nll_base\": 8.668620586395264, \"eval_ref\": \"bertaqa_en_mt_llama-2-70b__a01d57d3__s000__tb131072__v000\", \"oracle... | {
"index": "bertaqa_en_mt_llama-2-70b__a01d57d3__s000__tb131072__v000",
"uid": "bertaqa_en_mt_llama-2-70b__a01d57d3__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\": \"bertaqa_eu__838bcbe1__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 5270.3339849583945, \"nll_base\": 8.569849014282227, \"eval_ref\": \"bertaqa_eu__838bcbe1__s000__tb262144__v000\", \"oracle_delta\": 1.9446969032287598,... | {
"index": "bertaqa_eu__838bcbe1__s000__tb262144__v000",
"uid": "bertaqa_eu__838bcbe1__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\": \"bhs__basque__IO__S_IO_DO_V_AUX__cbe5fcc1__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 18.069527273049165, \"nll_base\": 2.894226943390279, \"eval_ref\": \"bhs__basque__IO__S_IO_DO_V_AUX__cbe5fcc1__s000__tb65536__v000\", ... | {
"index": "bhs__basque__IO__S_IO_DO_V_AUX__cbe5fcc1__s000__tb65536__v000",
"uid": "bhs__basque__IO__S_IO_DO_V_AUX__cbe5fcc1__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\": \"bhs__basque__S__IO_S_V_AUX__b66b754a__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 16.585846770741156, \"nll_base\": 2.808549727516613, \"eval_ref\": \"bhs__basque__S__IO_S_V_AUX__b66b754a__s000__tb65536__v000\", \"oracle... | {
"index": "bhs__basque__S__IO_S_V_AUX__b66b754a__s000__tb65536__v000",
"uid": "bhs__basque__S__IO_S_V_AUX__b66b754a__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\": \"bhs__basque__S__S_DO_V_AUX__e9e9c7e0__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 23.213857216304838, \"nll_base\": 3.14474939416956, \"eval_ref\": \"bhs__basque__S__S_DO_V_AUX__e9e9c7e0__s000__tb131072__v000\", \"orac... | {
"index": "bhs__basque__S__S_DO_V_AUX__e9e9c7e0__s000__tb131072__v000",
"uid": "bhs__basque__S__S_DO_V_AUX__e9e9c7e0__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\": \"bhs__hindi__S_O_V__d5904d4d__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2.414640855131176, \"nll_base\": 0.8815505618140811, \"eval_ref\": \"bhs__hindi__S_O_V__d5904d4d__s000__tb65536__v000\", \"oracle_delta\": 0.044559... | {
"index": "bhs__hindi__S_O_V__d5904d4d__s000__tb65536__v000",
"uid": "bhs__hindi__S_O_V__d5904d4d__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\": \"bhs__hindi__S_PossPRN_PossN_O_V__3bca20a9__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2.2479072991209517, \"nll_base\": 0.8099996941353343, \"eval_ref\": \"bhs__hindi__S_PossPRN_PossN_O_V__3bca20a9__s000__tb65536__v000\... | {
"index": "bhs__hindi__S_PossPRN_PossN_O_V__3bca20a9__s000__tb65536__v000",
"uid": "bhs__hindi__S_PossPRN_PossN_O_V__3bca20a9__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\": \"bhs__hindi__S_ne_PossPRN_O_V__320fb62e__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2.463382422664056, \"nll_base\": 0.9015353741032062, \"eval_ref\": \"bhs__hindi__S_ne_PossPRN_O_V__320fb62e__s000__tb65536__v000\", \"or... | {
"index": "bhs__hindi__S_ne_PossPRN_O_V__320fb62e__s000__tb65536__v000",
"uid": "bhs__hindi__S_ne_PossPRN_O_V__320fb62e__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\": \"bhs__hindi__S_ne_PossPRN_PossN_O_V__1939727e__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2.5198939530320454, \"nll_base\": 0.9242168185077136, \"eval_ref\": \"bhs__hindi__S_ne_PossPRN_PossN_O_V__1939727e__s000__tb2621... | {
"index": "bhs__hindi__S_ne_PossPRN_PossN_O_V__1939727e__s000__tb262144__v000",
"uid": "bhs__hindi__S_ne_PossPRN_PossN_O_V__1939727e__s000__tb262144__v000",
"axis": "instruction",
"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\": \"bhs__swahili__N_of_Poss_D_V__e32d68a1__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 20.87851996314877, \"nll_base\": 3.0387208775111607, \"eval_ref\": \"bhs__swahili__N_of_Poss_D_V__e32d68a1__s000__tb65536__v000\", \"orac... | {
"index": "bhs__swahili__N_of_Poss_D_V__e32d68a1__s000__tb65536__v000",
"uid": "bhs__swahili__N_of_Poss_D_V__e32d68a1__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\": \"bhs__swahili__N_of_Poss_D_ni_A__d92a0c9f__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 19.61595215447027, \"nll_base\": 2.9763431206434787, \"eval_ref\": \"bhs__swahili__N_of_Poss_D_ni_A__d92a0c9f__s000__tb65536__v000\", ... | {
"index": "bhs__swahili__N_of_Poss_D_ni_A__d92a0c9f__s000__tb65536__v000",
"uid": "bhs__swahili__N_of_Poss_D_ni_A__d92a0c9f__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\": \"bigbench_abstract_narrative_understanding_multiple_choice__c6692f56__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 6.722662532255379, \"nll_base\": 1.9054842862215908, \"eval_ref\": \"bigbench_abstract_narrative_understa... | {
"index": "bigbench_abstract_narrative_understanding_multiple_choice__c6692f56__s000__tb131072__v000",
"uid": "bigbench_abstract_narrative_understanding_multiple_choice__c6692f56__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_anachronisms_generate_until__4b35ea69__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1329.0649646385914, \"nll_base\": 7.192230939865112, \"eval_ref\": \"bigbench_anachronisms_generate_until__4b35ea69__s000__tb1... | {
"index": "bigbench_anachronisms_generate_until__4b35ea69__s000__tb131072__v001",
"uid": "bigbench_anachronisms_generate_until__4b35ea69__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_anachronisms_generate_until__4b35ea69__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1329.0649646385914, \"nll_base\": 7.192230939865112, \"eval_ref\": \"bigbench_anachronisms_generate_until__4b35ea69__s000__tb655... | {
"index": "bigbench_anachronisms_generate_until__4b35ea69__s000__tb65536__v000",
"uid": "bigbench_anachronisms_generate_until__4b35ea69__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_anachronisms_multiple_choice__3a594412__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 300.22820694347104, \"nll_base\": 5.704542875289917, \"eval_ref\": \"bigbench_anachronisms_multiple_choice__3a594412__s000__t... | {
"index": "bigbench_anachronisms_multiple_choice__3a594412__s000__tb131072__v001",
"uid": "bigbench_anachronisms_multiple_choice__3a594412__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_anachronisms_multiple_choice__3a594412__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 300.22820694347104, \"nll_base\": 5.704542875289917, \"eval_ref\": \"bigbench_anachronisms_multiple_choice__3a594412__s000__tb6... | {
"index": "bigbench_anachronisms_multiple_choice__3a594412__s000__tb65536__v000",
"uid": "bigbench_anachronisms_multiple_choice__3a594412__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_analogical_similarity_generate_until__bc7f2128__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 73.93416418867574, \"nll_base\": 4.303175024084143, \"eval_ref\": \"bigbench_analogical_similarity_generate_until__bc... | {
"index": "bigbench_analogical_similarity_generate_until__bc7f2128__s000__tb131072__v001",
"uid": "bigbench_analogical_similarity_generate_until__bc7f2128__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_analogical_similarity_multiple_choice__a764f080__s000__tb131072__v002\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 72.41681424281732, \"nll_base\": 4.282438513350813, \"eval_ref\": \"bigbench_analogical_similarity_multiple_choice__... | {
"index": "bigbench_analogical_similarity_multiple_choice__a764f080__s000__tb131072__v002",
"uid": "bigbench_analogical_similarity_multiple_choice__a764f080__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_analytic_entailment_generate_until__f801d31d__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 129.76937353454574, \"nll_base\": 4.865758825231482, \"eval_ref\": \"bigbench_analytic_entailment_generate_until__f801d... | {
"index": "bigbench_analytic_entailment_generate_until__f801d31d__s000__tb131072__v000",
"uid": "bigbench_analytic_entailment_generate_until__f801d31d__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_analytic_entailment_generate_until__f801d31d__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 129.76937353454574, \"nll_base\": 4.865758825231482, \"eval_ref\": \"bigbench_analytic_entailment_generate_until__f801d... | {
"index": "bigbench_analytic_entailment_generate_until__f801d31d__s000__tb262144__v001",
"uid": "bigbench_analytic_entailment_generate_until__f801d31d__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_analytic_entailment_generate_until__f801d31d__s000__tb65536__v002\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 129.76937353454574, \"nll_base\": 4.865758825231482, \"eval_ref\": \"bigbench_analytic_entailment_generate_until__f801d31... | {
"index": "bigbench_analytic_entailment_generate_until__f801d31d__s000__tb65536__v002",
"uid": "bigbench_analytic_entailment_generate_until__f801d31d__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_analytic_entailment_multiple_choice__db9e832d__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 12.511369628812126, \"nll_base\": 2.5266378012048194, \"eval_ref\": \"bigbench_analytic_entailment_multiple_choice__db9e... | {
"index": "bigbench_analytic_entailment_multiple_choice__db9e832d__s000__tb65536__v001",
"uid": "bigbench_analytic_entailment_multiple_choice__db9e832d__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_arithmetic_generate_until__9b60e6cd__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 37.3445077304827, \"nll_base\": 3.6201858520507812, \"eval_ref\": \"bigbench_arithmetic_generate_until__9b60e6cd__s000__tb131072... | {
"index": "bigbench_arithmetic_generate_until__9b60e6cd__s000__tb131072__v000",
"uid": "bigbench_arithmetic_generate_until__9b60e6cd__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_arithmetic_generate_until__9b60e6cd__s000__tb65536__v002\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 37.3445077304827, \"nll_base\": 3.6201858520507812, \"eval_ref\": \"bigbench_arithmetic_generate_until__9b60e6cd__s000__tb65536__v... | {
"index": "bigbench_arithmetic_generate_until__9b60e6cd__s000__tb65536__v002",
"uid": "bigbench_arithmetic_generate_until__9b60e6cd__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_arithmetic_multiple_choice__afaba70f__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 24.372697617331625, \"nll_base\": 3.193463555697737, \"eval_ref\": \"bigbench_arithmetic_multiple_choice__afaba70f__s000__tb131... | {
"index": "bigbench_arithmetic_multiple_choice__afaba70f__s000__tb131072__v001",
"uid": "bigbench_arithmetic_multiple_choice__afaba70f__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_arithmetic_multiple_choice__afaba70f__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 24.372697617331625, \"nll_base\": 3.193463555697737, \"eval_ref\": \"bigbench_arithmetic_multiple_choice__afaba70f__s000__tb65536... | {
"index": "bigbench_arithmetic_multiple_choice__afaba70f__s000__tb65536__v000",
"uid": "bigbench_arithmetic_multiple_choice__afaba70f__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_ascii_word_recognition_generate_until__fcb10938__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 29645.850881157006, \"nll_base\": 10.297077458079269, \"eval_ref\": \"bigbench_ascii_word_recognition_generate_until... | {
"index": "bigbench_ascii_word_recognition_generate_until__fcb10938__s000__tb131072__v000",
"uid": "bigbench_ascii_word_recognition_generate_until__fcb10938__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_auto_debugging_generate_until__27b86e44__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 11.429915793442868, \"nll_base\": 2.4362341106267267, \"eval_ref\": \"bigbench_auto_debugging_generate_until__27b86e44__s000... | {
"index": "bigbench_auto_debugging_generate_until__27b86e44__s000__tb131072__v000",
"uid": "bigbench_auto_debugging_generate_until__27b86e44__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_auto_debugging_generate_until__27b86e44__s000__tb65536__v002\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 11.429915793442868, \"nll_base\": 2.4362341106267267, \"eval_ref\": \"bigbench_auto_debugging_generate_until__27b86e44__s000__... | {
"index": "bigbench_auto_debugging_generate_until__27b86e44__s000__tb65536__v002",
"uid": "bigbench_auto_debugging_generate_until__27b86e44__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_bbq_lite_json_generate_until__3e324ef7__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 91.09645484595102, \"nll_base\": 4.51191888853561, \"eval_ref\": \"bigbench_bbq_lite_json_generate_until__3e324ef7__s000__tb1... | {
"index": "bigbench_bbq_lite_json_generate_until__3e324ef7__s000__tb131072__v001",
"uid": "bigbench_bbq_lite_json_generate_until__3e324ef7__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_bbq_lite_json_generate_until__3e324ef7__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 91.09645484595102, \"nll_base\": 4.51191888853561, \"eval_ref\": \"bigbench_bbq_lite_json_generate_until__3e324ef7__s000__tb655... | {
"index": "bigbench_bbq_lite_json_generate_until__3e324ef7__s000__tb65536__v000",
"uid": "bigbench_bbq_lite_json_generate_until__3e324ef7__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_bbq_lite_json_multiple_choice__95e5652f__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 73.39681187171641, \"nll_base\": 4.295880499686904, \"eval_ref\": \"bigbench_bbq_lite_json_multiple_choice__95e5652f__s000__... | {
"index": "bigbench_bbq_lite_json_multiple_choice__95e5652f__s000__tb131072__v000",
"uid": "bigbench_bbq_lite_json_multiple_choice__95e5652f__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_bridging_anaphora_resolution_barqa_generate_until__0b29cef0__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 175.6393275624905, \"nll_base\": 5.1684326171875, \"eval_ref\": \"bigbench_bridging_anaphora_resolution_... | {
"index": "bigbench_bridging_anaphora_resolution_barqa_generate_until__0b29cef0__s000__tb131072__v001",
"uid": "bigbench_bridging_anaphora_resolution_barqa_generate_until__0b29cef0__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_bridging_anaphora_resolution_barqa_generate_until__0b29cef0__s000__tb262144__v002\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 175.6393275624905, \"nll_base\": 5.1684326171875, \"eval_ref\": \"bigbench_bridging_anaphora_resolution_... | {
"index": "bigbench_bridging_anaphora_resolution_barqa_generate_until__0b29cef0__s000__tb262144__v002",
"uid": "bigbench_bridging_anaphora_resolution_barqa_generate_until__0b29cef0__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_bridging_anaphora_resolution_barqa_generate_until__0b29cef0__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 175.6393275624905, \"nll_base\": 5.1684326171875, \"eval_ref\": \"bigbench_bridging_anaphora_resolution_ba... | {
"index": "bigbench_bridging_anaphora_resolution_barqa_generate_until__0b29cef0__s000__tb65536__v000",
"uid": "bigbench_bridging_anaphora_resolution_barqa_generate_until__0b29cef0__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_causal_judgment_multiple_choice__6587b9d6__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 636.1940979036901, \"nll_base\": 6.455503702163696, \"eval_ref\": \"bigbench_causal_judgment_multiple_choice__6587b9d6__s0... | {
"index": "bigbench_causal_judgment_multiple_choice__6587b9d6__s000__tb131072__v001",
"uid": "bigbench_causal_judgment_multiple_choice__6587b9d6__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_causal_judgment_multiple_choice__6587b9d6__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 636.1940979036901, \"nll_base\": 6.455503702163696, \"eval_ref\": \"bigbench_causal_judgment_multiple_choice__6587b9d6__s000... | {
"index": "bigbench_causal_judgment_multiple_choice__6587b9d6__s000__tb65536__v000",
"uid": "bigbench_causal_judgment_multiple_choice__6587b9d6__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_cause_and_effect_multiple_choice__5cd5317e__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3.1278831375072143, \"nll_base\": 1.1403564618531927, \"eval_ref\": \"bigbench_cause_and_effect_multiple_choice__5cd5317e... | {
"index": "bigbench_cause_and_effect_multiple_choice__5cd5317e__s000__tb262144__v000",
"uid": "bigbench_cause_and_effect_multiple_choice__5cd5317e__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_cause_and_effect_multiple_choice__5cd5317e__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3.1278831375072143, \"nll_base\": 1.1403564618531927, \"eval_ref\": \"bigbench_cause_and_effect_multiple_choice__5cd5317e__... | {
"index": "bigbench_cause_and_effect_multiple_choice__5cd5317e__s000__tb65536__v001",
"uid": "bigbench_cause_and_effect_multiple_choice__5cd5317e__s000__tb65536__v001",
"axis": "reasoning",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.