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\": \"mmlu_prox_lite_it_law__903ebf89__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2495.018754811707, \"nll_base\": 7.822051525115967, \"eval_ref\": \"mmlu_prox_lite_it_law__903ebf89__s000__tb131072__v000\", \"oracle_delta\"... | {
"index": "mmlu_prox_lite_it_law__903ebf89__s000__tb131072__v000",
"uid": "mmlu_prox_lite_it_law__903ebf89__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\": \"mmlu_prox_lite_it_other__5feff858__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2909.3754138071913, \"nll_base\": 7.975693702697754, \"eval_ref\": \"mmlu_prox_lite_it_other__5feff858__s000__tb131072__v000\", \"oracle_de... | {
"index": "mmlu_prox_lite_it_other__5feff858__s000__tb131072__v000",
"uid": "mmlu_prox_lite_it_other__5feff858__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\": \"mmlu_prox_lite_ja_health__6eaf8ffa__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4982.251213954744, \"nll_base\": 8.513637118869358, \"eval_ref\": \"mmlu_prox_lite_ja_health__6eaf8ffa__s000__tb65536__v000\", \"oracle_delt... | {
"index": "mmlu_prox_lite_ja_health__6eaf8ffa__s000__tb65536__v000",
"uid": "mmlu_prox_lite_ja_health__6eaf8ffa__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\": \"mmlu_prox_lite_ja_law__8523e1dc__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4408.386206026031, \"nll_base\": 8.391263961791992, \"eval_ref\": \"mmlu_prox_lite_ja_law__8523e1dc__s000__tb131072__v000\", \"oracle_delta\"... | {
"index": "mmlu_prox_lite_ja_law__8523e1dc__s000__tb131072__v000",
"uid": "mmlu_prox_lite_ja_law__8523e1dc__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\": \"mmlu_prox_lite_ja_physics__a1199dbd__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4837.5042911435985, \"nll_base\": 8.484154224395752, \"eval_ref\": \"mmlu_prox_lite_ja_physics__a1199dbd__s000__tb65536__v000\", \"oracle_d... | {
"index": "mmlu_prox_lite_ja_physics__a1199dbd__s000__tb65536__v000",
"uid": "mmlu_prox_lite_ja_physics__a1199dbd__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\": \"mmlu_prox_lite_ko_biology__e73faefb__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1894.6330355984799, \"nll_base\": 7.546780450003488, \"eval_ref\": \"mmlu_prox_lite_ko_biology__e73faefb__s000__tb65536__v000\", \"oracle_d... | {
"index": "mmlu_prox_lite_ko_biology__e73faefb__s000__tb65536__v000",
"uid": "mmlu_prox_lite_ko_biology__e73faefb__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\": \"mmlu_prox_lite_ko_chemistry__16bc0f8a__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3398.1581264033443, \"nll_base\": 8.130988836288452, \"eval_ref\": \"mmlu_prox_lite_ko_chemistry__16bc0f8a__s000__tb65536__v000\", \"orac... | {
"index": "mmlu_prox_lite_ko_chemistry__16bc0f8a__s000__tb65536__v000",
"uid": "mmlu_prox_lite_ko_chemistry__16bc0f8a__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\": \"mmlu_prox_lite_ko_economics__61cfcd72__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3973.8355072947365, \"nll_base\": 8.287487030029297, \"eval_ref\": \"mmlu_prox_lite_ko_economics__61cfcd72__s000__tb65536__v000\", \"orac... | {
"index": "mmlu_prox_lite_ko_economics__61cfcd72__s000__tb65536__v000",
"uid": "mmlu_prox_lite_ko_economics__61cfcd72__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\": \"mmlu_prox_lite_ko_engineering__06bedd93__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2614.602604615068, \"nll_base\": 7.86886739730835, \"eval_ref\": \"mmlu_prox_lite_ko_engineering__06bedd93__s000__tb65536__v000\", \"or... | {
"index": "mmlu_prox_lite_ko_engineering__06bedd93__s000__tb65536__v000",
"uid": "mmlu_prox_lite_ko_engineering__06bedd93__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\": \"mmlu_prox_lite_ko_history__96cc3164__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3153.663873759741, \"nll_base\": 8.056320190429688, \"eval_ref\": \"mmlu_prox_lite_ko_history__96cc3164__s000__tb131072__v000\", \"oracle... | {
"index": "mmlu_prox_lite_ko_history__96cc3164__s000__tb131072__v000",
"uid": "mmlu_prox_lite_ko_history__96cc3164__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\": \"mmlu_prox_lite_mr_biology__3d7e7646__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2095.340950862923, \"nll_base\": 7.647471564156668, \"eval_ref\": \"mmlu_prox_lite_mr_biology__3d7e7646__s000__tb65536__v000\", \"oracle_de... | {
"index": "mmlu_prox_lite_mr_biology__3d7e7646__s000__tb65536__v000",
"uid": "mmlu_prox_lite_mr_biology__3d7e7646__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\": \"mmlu_prox_lite_mr_business__4c28eca5__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4440.216790106975, \"nll_base\": 8.398458480834961, \"eval_ref\": \"mmlu_prox_lite_mr_business__4c28eca5__s000__tb65536__v000\", \"oracle_... | {
"index": "mmlu_prox_lite_mr_business__4c28eca5__s000__tb65536__v000",
"uid": "mmlu_prox_lite_mr_business__4c28eca5__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\": \"mmlu_prox_lite_mr_chemistry__e32334c2__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2667.4451966820025, \"nll_base\": 7.888876438140869, \"eval_ref\": \"mmlu_prox_lite_mr_chemistry__e32334c2__s000__tb65536__v000\", \"orac... | {
"index": "mmlu_prox_lite_mr_chemistry__e32334c2__s000__tb65536__v000",
"uid": "mmlu_prox_lite_mr_chemistry__e32334c2__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\": \"mmlu_prox_lite_mr_computer_science__d5046f11__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4645.228702123877, \"nll_base\": 8.443595886230469, \"eval_ref\": \"mmlu_prox_lite_mr_computer_science__d5046f11__s000__tb131072... | {
"index": "mmlu_prox_lite_mr_computer_science__d5046f11__s000__tb131072__v000",
"uid": "mmlu_prox_lite_mr_computer_science__d5046f11__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\": \"mmlu_prox_lite_mr_engineering__872601cb__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3654.6665875071717, \"nll_base\": 8.203760147094727, \"eval_ref\": \"mmlu_prox_lite_mr_engineering__872601cb__s000__tb262144__v000\",... | {
"index": "mmlu_prox_lite_mr_engineering__872601cb__s000__tb262144__v000",
"uid": "mmlu_prox_lite_mr_engineering__872601cb__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\": \"mmlu_prox_lite_mr_health__103750bc__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2480.0369450751496, \"nll_base\": 7.816028736255787, \"eval_ref\": \"mmlu_prox_lite_mr_health__103750bc__s000__tb65536__v000\", \"oracle_del... | {
"index": "mmlu_prox_lite_mr_health__103750bc__s000__tb65536__v000",
"uid": "mmlu_prox_lite_mr_health__103750bc__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\": \"mmlu_prox_lite_mr_physics__d5638c9b__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4517.027210881163, \"nll_base\": 8.415609359741211, \"eval_ref\": \"mmlu_prox_lite_mr_physics__d5638c9b__s000__tb65536__v000\", \"oracle_de... | {
"index": "mmlu_prox_lite_mr_physics__d5638c9b__s000__tb65536__v000",
"uid": "mmlu_prox_lite_mr_physics__d5638c9b__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\": \"mmlu_prox_lite_mr_psychology__f90aeaa8__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3244.590585657108, \"nll_base\": 8.084744453430176, \"eval_ref\": \"mmlu_prox_lite_mr_psychology__f90aeaa8__s000__tb131072__v000\", \"... | {
"index": "mmlu_prox_lite_mr_psychology__f90aeaa8__s000__tb131072__v000",
"uid": "mmlu_prox_lite_mr_psychology__f90aeaa8__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\": \"mmlu_prox_lite_ne_biology__d4b89949__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2824.993293811353, \"nll_base\": 7.946261269705636, \"eval_ref\": \"mmlu_prox_lite_ne_biology__d4b89949__s000__tb131072__v000\", \"oracle... | {
"index": "mmlu_prox_lite_ne_biology__d4b89949__s000__tb131072__v000",
"uid": "mmlu_prox_lite_ne_biology__d4b89949__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\": \"mmlu_prox_lite_ne_business__ebe5b4f9__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 6484.198208388134, \"nll_base\": 8.77712345123291, \"eval_ref\": \"mmlu_prox_lite_ne_business__ebe5b4f9__s000__tb131072__v000\", \"oracl... | {
"index": "mmlu_prox_lite_ne_business__ebe5b4f9__s000__tb131072__v000",
"uid": "mmlu_prox_lite_ne_business__ebe5b4f9__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\": \"mmlu_prox_lite_ne_chemistry__bc194379__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 4899.029074662622, \"nll_base\": 8.496792316436768, \"eval_ref\": \"mmlu_prox_lite_ne_chemistry__bc194379__s000__tb262144__v000\", \"or... | {
"index": "mmlu_prox_lite_ne_chemistry__bc194379__s000__tb262144__v000",
"uid": "mmlu_prox_lite_ne_chemistry__bc194379__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\": \"mmlu_prox_lite_ne_computer_science__c5aa0835__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 8613.116666369478, \"nll_base\": 9.061041514078775, \"eval_ref\": \"mmlu_prox_lite_ne_computer_science__c5aa0835__s000__tb65536__v... | {
"index": "mmlu_prox_lite_ne_computer_science__c5aa0835__s000__tb65536__v000",
"uid": "mmlu_prox_lite_ne_computer_science__c5aa0835__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\": \"mmlu_prox_lite_ne_engineering__7e2d98d7__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 4566.035771913566, \"nll_base\": 8.426400661468506, \"eval_ref\": \"mmlu_prox_lite_ne_engineering__7e2d98d7__s000__tb262144__v000\", ... | {
"index": "mmlu_prox_lite_ne_engineering__7e2d98d7__s000__tb262144__v000",
"uid": "mmlu_prox_lite_ne_engineering__7e2d98d7__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\": \"mmlu_prox_lite_ne_health__1e4e1992__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4218.496610704866, \"nll_base\": 8.347234090169271, \"eval_ref\": \"mmlu_prox_lite_ne_health__1e4e1992__s000__tb131072__v000\", \"oracle_d... | {
"index": "mmlu_prox_lite_ne_health__1e4e1992__s000__tb131072__v000",
"uid": "mmlu_prox_lite_ne_health__1e4e1992__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\": \"mmlu_prox_lite_ne_history__25ded481__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2177.6475708160187, \"nll_base\": 7.686000477183949, \"eval_ref\": \"mmlu_prox_lite_ne_history__25ded481__s000__tb65536__v000\", \"oracle_d... | {
"index": "mmlu_prox_lite_ne_history__25ded481__s000__tb65536__v000",
"uid": "mmlu_prox_lite_ne_history__25ded481__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\": \"mmlu_prox_lite_ne_physics__6a9d8c37__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 7816.406475270286, \"nll_base\": 8.963980197906494, \"eval_ref\": \"mmlu_prox_lite_ne_physics__6a9d8c37__s000__tb262144__v000\", \"oracle... | {
"index": "mmlu_prox_lite_ne_physics__6a9d8c37__s000__tb262144__v000",
"uid": "mmlu_prox_lite_ne_physics__6a9d8c37__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\": \"mmlu_prox_lite_ne_psychology__2cb9cde8__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3882.5499386944166, \"nll_base\": 8.264247417449951, \"eval_ref\": \"mmlu_prox_lite_ne_psychology__2cb9cde8__s000__tb131072__v000\", \... | {
"index": "mmlu_prox_lite_ne_psychology__2cb9cde8__s000__tb131072__v000",
"uid": "mmlu_prox_lite_ne_psychology__2cb9cde8__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\": \"mmlu_prox_lite_pt_biology__7a629846__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3940.479277721826, \"nll_base\": 8.27905763898577, \"eval_ref\": \"mmlu_prox_lite_pt_biology__7a629846__s000__tb131072__v000\", \"oracle_... | {
"index": "mmlu_prox_lite_pt_biology__7a629846__s000__tb131072__v000",
"uid": "mmlu_prox_lite_pt_biology__7a629846__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\": \"mmlu_prox_lite_pt_computer_science__93ba03f3__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 4636.489090823165, \"nll_base\": 8.441712697347006, \"eval_ref\": \"mmlu_prox_lite_pt_computer_science__93ba03f3__s000__tb262144... | {
"index": "mmlu_prox_lite_pt_computer_science__93ba03f3__s000__tb262144__v000",
"uid": "mmlu_prox_lite_pt_computer_science__93ba03f3__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\": \"mmlu_prox_lite_pt_economics__7505ad40__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 7923.393183255867, \"nll_base\": 8.977574825286865, \"eval_ref\": \"mmlu_prox_lite_pt_economics__7505ad40__s000__tb131072__v000\", \"or... | {
"index": "mmlu_prox_lite_pt_economics__7505ad40__s000__tb131072__v000",
"uid": "mmlu_prox_lite_pt_economics__7505ad40__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\": \"mmlu_prox_lite_pt_health__37a4903a__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 4574.491908389984, \"nll_base\": 8.428250913266782, \"eval_ref\": \"mmlu_prox_lite_pt_health__37a4903a__s000__tb262144__v000\", \"oracle_d... | {
"index": "mmlu_prox_lite_pt_health__37a4903a__s000__tb262144__v000",
"uid": "mmlu_prox_lite_pt_health__37a4903a__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\": \"mmlu_prox_lite_pt_law__df7f34f9__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4292.04523223122, \"nll_base\": 8.364518642425537, \"eval_ref\": \"mmlu_prox_lite_pt_law__df7f34f9__s000__tb131072__v000\", \"oracle_delta\":... | {
"index": "mmlu_prox_lite_pt_law__df7f34f9__s000__tb131072__v000",
"uid": "mmlu_prox_lite_pt_law__df7f34f9__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\": \"mmlu_prox_lite_pt_math__74ab3a96__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 7132.512819788647, \"nll_base\": 8.872418880462646, \"eval_ref\": \"mmlu_prox_lite_pt_math__74ab3a96__s000__tb131072__v000\", \"oracle_delta... | {
"index": "mmlu_prox_lite_pt_math__74ab3a96__s000__tb131072__v000",
"uid": "mmlu_prox_lite_pt_math__74ab3a96__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\": \"mmlu_prox_lite_ru_business__aed5f568__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4204.742741812708, \"nll_base\": 8.343968391418457, \"eval_ref\": \"mmlu_prox_lite_ru_business__aed5f568__s000__tb65536__v000\", \"oracle_... | {
"index": "mmlu_prox_lite_ru_business__aed5f568__s000__tb65536__v000",
"uid": "mmlu_prox_lite_ru_business__aed5f568__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\": \"mmlu_prox_lite_ru_law__a03f9fa0__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2023.126928876565, \"nll_base\": 7.612399578094482, \"eval_ref\": \"mmlu_prox_lite_ru_law__a03f9fa0__s000__tb65536__v000\", \"oracle_delta\": 0... | {
"index": "mmlu_prox_lite_ru_law__a03f9fa0__s000__tb65536__v000",
"uid": "mmlu_prox_lite_ru_law__a03f9fa0__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\": \"mmlu_prox_lite_ru_psychology__5a918dd6__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3633.966617902609, \"nll_base\": 8.198080062866211, \"eval_ref\": \"mmlu_prox_lite_ru_psychology__5a918dd6__s000__tb65536__v000\", \"ora... | {
"index": "mmlu_prox_lite_ru_psychology__5a918dd6__s000__tb65536__v000",
"uid": "mmlu_prox_lite_ru_psychology__5a918dd6__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\": \"mmlu_prox_lite_sr_biology__2949d48a__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2351.860671204092, \"nll_base\": 7.762962068830218, \"eval_ref\": \"mmlu_prox_lite_sr_biology__2949d48a__s000__tb65536__v000\", \"oracle_de... | {
"index": "mmlu_prox_lite_sr_biology__2949d48a__s000__tb65536__v000",
"uid": "mmlu_prox_lite_sr_biology__2949d48a__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\": \"mmlu_prox_lite_sr_business__34541a29__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3405.0930768611, \"nll_base\": 8.13302755355835, \"eval_ref\": \"mmlu_prox_lite_sr_business__34541a29__s000__tb262144__v000\", \"oracle_... | {
"index": "mmlu_prox_lite_sr_business__34541a29__s000__tb262144__v000",
"uid": "mmlu_prox_lite_sr_business__34541a29__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\": \"mmlu_prox_lite_sr_computer_science__cf0e4bbc__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 5308.020369698426, \"nll_base\": 8.576974232991537, \"eval_ref\": \"mmlu_prox_lite_sr_computer_science__cf0e4bbc__s000__tb65536__v... | {
"index": "mmlu_prox_lite_sr_computer_science__cf0e4bbc__s000__tb65536__v000",
"uid": "mmlu_prox_lite_sr_computer_science__cf0e4bbc__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\": \"mmlu_prox_lite_sr_health__57e1b9b1__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2938.702466063737, \"nll_base\": 7.985723424840857, \"eval_ref\": \"mmlu_prox_lite_sr_health__57e1b9b1__s000__tb65536__v000\", \"oracle_delt... | {
"index": "mmlu_prox_lite_sr_health__57e1b9b1__s000__tb65536__v000",
"uid": "mmlu_prox_lite_sr_health__57e1b9b1__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\": \"mmlu_prox_lite_sr_history__57d7e67d__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1747.255562286788, \"nll_base\": 7.465801585804332, \"eval_ref\": \"mmlu_prox_lite_sr_history__57d7e67d__s000__tb262144__v000\", \"oracle... | {
"index": "mmlu_prox_lite_sr_history__57d7e67d__s000__tb262144__v000",
"uid": "mmlu_prox_lite_sr_history__57d7e67d__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\": \"mmlu_prox_lite_sw_biology__1468c45a__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3075.9960895929266, \"nll_base\": 8.031384059361049, \"eval_ref\": \"mmlu_prox_lite_sw_biology__1468c45a__s000__tb262144__v000\", \"oracl... | {
"index": "mmlu_prox_lite_sw_biology__1468c45a__s000__tb262144__v000",
"uid": "mmlu_prox_lite_sw_biology__1468c45a__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\": \"mmlu_prox_lite_sw_business__035ef88b__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3797.3953439610655, \"nll_base\": 8.24207067489624, \"eval_ref\": \"mmlu_prox_lite_sw_business__035ef88b__s000__tb131072__v000\", \"orac... | {
"index": "mmlu_prox_lite_sw_business__035ef88b__s000__tb131072__v000",
"uid": "mmlu_prox_lite_sw_business__035ef88b__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\": \"mmlu_prox_lite_sw_economics__105e02b8__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4547.612091853231, \"nll_base\": 8.422357559204102, \"eval_ref\": \"mmlu_prox_lite_sw_economics__105e02b8__s000__tb65536__v000\", \"oracl... | {
"index": "mmlu_prox_lite_sw_economics__105e02b8__s000__tb65536__v000",
"uid": "mmlu_prox_lite_sw_economics__105e02b8__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\": \"mmlu_prox_lite_sw_engineering__f4a9652c__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3675.7754136122126, \"nll_base\": 8.209519386291504, \"eval_ref\": \"mmlu_prox_lite_sw_engineering__f4a9652c__s000__tb262144__v000\",... | {
"index": "mmlu_prox_lite_sw_engineering__f4a9652c__s000__tb262144__v000",
"uid": "mmlu_prox_lite_sw_engineering__f4a9652c__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\": \"mmlu_prox_lite_sw_history__32e56501__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3962.7917172803996, \"nll_base\": 8.284704034978693, \"eval_ref\": \"mmlu_prox_lite_sw_history__32e56501__s000__tb262144__v000\", \"oracl... | {
"index": "mmlu_prox_lite_sw_history__32e56501__s000__tb262144__v000",
"uid": "mmlu_prox_lite_sw_history__32e56501__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\": \"mmlu_prox_lite_sw_math__0145598e__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 5271.243802451253, \"nll_base\": 8.570021629333496, \"eval_ref\": \"mmlu_prox_lite_sw_math__0145598e__s000__tb65536__v000\", \"oracle_delta\":... | {
"index": "mmlu_prox_lite_sw_math__0145598e__s000__tb65536__v000",
"uid": "mmlu_prox_lite_sw_math__0145598e__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\": \"mmlu_prox_lite_sw_other__812e618d__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 4322.069099713664, \"nll_base\": 8.371489524841309, \"eval_ref\": \"mmlu_prox_lite_sw_other__812e618d__s000__tb262144__v000\", \"oracle_del... | {
"index": "mmlu_prox_lite_sw_other__812e618d__s000__tb262144__v000",
"uid": "mmlu_prox_lite_sw_other__812e618d__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\": \"mmlu_prox_lite_sw_physics__fa79c0af__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4607.152195250942, \"nll_base\": 8.435365200042725, \"eval_ref\": \"mmlu_prox_lite_sw_physics__fa79c0af__s000__tb65536__v000\", \"oracle_de... | {
"index": "mmlu_prox_lite_sw_physics__fa79c0af__s000__tb65536__v000",
"uid": "mmlu_prox_lite_sw_physics__fa79c0af__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\": \"mmlu_prox_lite_sw_psychology__944116c9__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3546.454569127945, \"nll_base\": 8.173703670501709, \"eval_ref\": \"mmlu_prox_lite_sw_psychology__944116c9__s000__tb262144__v000\", \"... | {
"index": "mmlu_prox_lite_sw_psychology__944116c9__s000__tb262144__v000",
"uid": "mmlu_prox_lite_sw_psychology__944116c9__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\": \"mmlu_prox_lite_te_biology__1c906c01__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1475.8259707535087, \"nll_base\": 7.296973092215402, \"eval_ref\": \"mmlu_prox_lite_te_biology__1c906c01__s000__tb131072__v000\", \"oracl... | {
"index": "mmlu_prox_lite_te_biology__1c906c01__s000__tb131072__v000",
"uid": "mmlu_prox_lite_te_biology__1c906c01__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\": \"mmlu_prox_lite_th_computer_science__ea6a4e5f__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3938.721672558531, \"nll_base\": 8.278611501057943, \"eval_ref\": \"mmlu_prox_lite_th_computer_science__ea6a4e5f__s000__tb65536__v... | {
"index": "mmlu_prox_lite_th_computer_science__ea6a4e5f__s000__tb65536__v000",
"uid": "mmlu_prox_lite_th_computer_science__ea6a4e5f__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\": \"mmlu_prox_lite_uk_business__6068403c__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 5774.641859475751, \"nll_base\": 8.661231517791748, \"eval_ref\": \"mmlu_prox_lite_uk_business__6068403c__s000__tb65536__v000\", \"oracle_... | {
"index": "mmlu_prox_lite_uk_business__6068403c__s000__tb65536__v000",
"uid": "mmlu_prox_lite_uk_business__6068403c__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\": \"mmlu_prox_lite_uk_chemistry__64374ce3__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3705.409602560815, \"nll_base\": 8.217549085617065, \"eval_ref\": \"mmlu_prox_lite_uk_chemistry__64374ce3__s000__tb131072__v000\", \"or... | {
"index": "mmlu_prox_lite_uk_chemistry__64374ce3__s000__tb131072__v000",
"uid": "mmlu_prox_lite_uk_chemistry__64374ce3__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\": \"mmlu_prox_lite_uk_computer_science__983cca88__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 8042.315600585555, \"nll_base\": 8.992472330729166, \"eval_ref\": \"mmlu_prox_lite_uk_computer_science__983cca88__s000__tb65536__v... | {
"index": "mmlu_prox_lite_uk_computer_science__983cca88__s000__tb65536__v000",
"uid": "mmlu_prox_lite_uk_computer_science__983cca88__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\": \"mmlu_prox_lite_uk_law__87f9a8ef__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2537.0784518719192, \"nll_base\": 7.838768482208252, \"eval_ref\": \"mmlu_prox_lite_uk_law__87f9a8ef__s000__tb131072__v000\", \"oracle_delta\... | {
"index": "mmlu_prox_lite_uk_law__87f9a8ef__s000__tb131072__v000",
"uid": "mmlu_prox_lite_uk_law__87f9a8ef__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\": \"mmlu_prox_lite_uk_physics__2863b934__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 5502.652958206962, \"nll_base\": 8.612985610961914, \"eval_ref\": \"mmlu_prox_lite_uk_physics__2863b934__s000__tb131072__v000\", \"oracle... | {
"index": "mmlu_prox_lite_uk_physics__2863b934__s000__tb131072__v000",
"uid": "mmlu_prox_lite_uk_physics__2863b934__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\": \"mmlu_prox_lite_ur_physics__f63f0ba1__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 5597.666081088207, \"nll_base\": 8.630105018615723, \"eval_ref\": \"mmlu_prox_lite_ur_physics__f63f0ba1__s000__tb131072__v000\", \"oracle... | {
"index": "mmlu_prox_lite_ur_physics__f63f0ba1__s000__tb131072__v000",
"uid": "mmlu_prox_lite_ur_physics__f63f0ba1__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\": \"mmlu_prox_lite_vi_computer_science__f008be6c__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4289.676597756248, \"nll_base\": 8.363966623942057, \"eval_ref\": \"mmlu_prox_lite_vi_computer_science__f008be6c__s000__tb131072... | {
"index": "mmlu_prox_lite_vi_computer_science__f008be6c__s000__tb131072__v000",
"uid": "mmlu_prox_lite_vi_computer_science__f008be6c__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\": \"mmlu_prox_lite_vi_economics__8dc84d90__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2725.4243228708424, \"nll_base\": 7.910379409790039, \"eval_ref\": \"mmlu_prox_lite_vi_economics__8dc84d90__s000__tb131072__v000\", \"o... | {
"index": "mmlu_prox_lite_vi_economics__8dc84d90__s000__tb131072__v000",
"uid": "mmlu_prox_lite_vi_economics__8dc84d90__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\": \"mmlu_prox_lite_vi_engineering__3abb6a16__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2869.028676609305, \"nll_base\": 7.961728811264038, \"eval_ref\": \"mmlu_prox_lite_vi_engineering__3abb6a16__s000__tb65536__v000\", \"o... | {
"index": "mmlu_prox_lite_vi_engineering__3abb6a16__s000__tb65536__v000",
"uid": "mmlu_prox_lite_vi_engineering__3abb6a16__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\": \"mmlu_prox_lite_vi_health__66b94944__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2660.1913282276378, \"nll_base\": 7.886153327094184, \"eval_ref\": \"mmlu_prox_lite_vi_health__66b94944__s000__tb65536__v000\", \"oracle_del... | {
"index": "mmlu_prox_lite_vi_health__66b94944__s000__tb65536__v000",
"uid": "mmlu_prox_lite_vi_health__66b94944__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\": \"mmlu_prox_lite_vi_history__6b7b7907__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1389.8604013820004, \"nll_base\": 7.23695859042081, \"eval_ref\": \"mmlu_prox_lite_vi_history__6b7b7907__s000__tb131072__v000\", \"oracle... | {
"index": "mmlu_prox_lite_vi_history__6b7b7907__s000__tb131072__v000",
"uid": "mmlu_prox_lite_vi_history__6b7b7907__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\": \"mmlu_prox_lite_vi_philosophy__2d767b32__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3586.47692010073, \"nll_base\": 8.184925640330595, \"eval_ref\": \"mmlu_prox_lite_vi_philosophy__2d767b32__s000__tb131072__v000\", \"o... | {
"index": "mmlu_prox_lite_vi_philosophy__2d767b32__s000__tb131072__v000",
"uid": "mmlu_prox_lite_vi_philosophy__2d767b32__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\": \"mmlu_prox_lite_vi_physics__5cdbd145__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3316.9344093000013, \"nll_base\": 8.106796264648438, \"eval_ref\": \"mmlu_prox_lite_vi_physics__5cdbd145__s000__tb65536__v000\", \"oracle_d... | {
"index": "mmlu_prox_lite_vi_physics__5cdbd145__s000__tb65536__v000",
"uid": "mmlu_prox_lite_vi_physics__5cdbd145__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\": \"mmlu_prox_lite_wo_business__a6ae604d__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4994.903455821175, \"nll_base\": 8.516173362731934, \"eval_ref\": \"mmlu_prox_lite_wo_business__a6ae604d__s000__tb131072__v000\", \"orac... | {
"index": "mmlu_prox_lite_wo_business__a6ae604d__s000__tb131072__v000",
"uid": "mmlu_prox_lite_wo_business__a6ae604d__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\": \"mmlu_prox_lite_wo_chemistry__b7b93185__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4083.0849507889084, \"nll_base\": 8.314608097076416, \"eval_ref\": \"mmlu_prox_lite_wo_chemistry__b7b93185__s000__tb65536__v000\", \"orac... | {
"index": "mmlu_prox_lite_wo_chemistry__b7b93185__s000__tb65536__v000",
"uid": "mmlu_prox_lite_wo_chemistry__b7b93185__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\": \"mmlu_prox_lite_wo_engineering__59cd6638__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3551.4569301676156, \"nll_base\": 8.175113201141357, \"eval_ref\": \"mmlu_prox_lite_wo_engineering__59cd6638__s000__tb131072__v000\",... | {
"index": "mmlu_prox_lite_wo_engineering__59cd6638__s000__tb131072__v000",
"uid": "mmlu_prox_lite_wo_engineering__59cd6638__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\": \"mmlu_prox_lite_wo_history__9d2e91ac__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2955.8056048541434, \"nll_base\": 7.991526517001065, \"eval_ref\": \"mmlu_prox_lite_wo_history__9d2e91ac__s000__tb262144__v000\", \"oracl... | {
"index": "mmlu_prox_lite_wo_history__9d2e91ac__s000__tb262144__v000",
"uid": "mmlu_prox_lite_wo_history__9d2e91ac__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\": \"mmlu_prox_lite_yo_biology__7f73e8ba__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2100.756405207625, \"nll_base\": 7.650052751813616, \"eval_ref\": \"mmlu_prox_lite_yo_biology__7f73e8ba__s000__tb262144__v000\", \"oracle... | {
"index": "mmlu_prox_lite_yo_biology__7f73e8ba__s000__tb262144__v000",
"uid": "mmlu_prox_lite_yo_biology__7f73e8ba__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\": \"mmlu_prox_lite_yo_computer_science__b1018af1__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2178.1094171482487, \"nll_base\": 7.686212539672852, \"eval_ref\": \"mmlu_prox_lite_yo_computer_science__b1018af1__s000__tb65536__... | {
"index": "mmlu_prox_lite_yo_computer_science__b1018af1__s000__tb65536__v000",
"uid": "mmlu_prox_lite_yo_computer_science__b1018af1__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\": \"mmlu_prox_lite_yo_health__3a96c44d__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2242.619425627732, \"nll_base\": 7.715399848090278, \"eval_ref\": \"mmlu_prox_lite_yo_health__3a96c44d__s000__tb65536__v000\", \"oracle_delt... | {
"index": "mmlu_prox_lite_yo_health__3a96c44d__s000__tb65536__v000",
"uid": "mmlu_prox_lite_yo_health__3a96c44d__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\": \"mmlu_prox_lite_yo_history__c5cd838e__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1674.984206814819, \"nll_base\": 7.423559015447443, \"eval_ref\": \"mmlu_prox_lite_yo_history__c5cd838e__s000__tb65536__v000\", \"oracle_de... | {
"index": "mmlu_prox_lite_yo_history__c5cd838e__s000__tb65536__v000",
"uid": "mmlu_prox_lite_yo_history__c5cd838e__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\": \"mmlu_prox_lite_yo_law__1e78d97c__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1704.8081915117034, \"nll_base\": 7.4412078857421875, \"eval_ref\": \"mmlu_prox_lite_yo_law__1e78d97c__s000__tb262144__v000\", \"oracle_delta... | {
"index": "mmlu_prox_lite_yo_law__1e78d97c__s000__tb262144__v000",
"uid": "mmlu_prox_lite_yo_law__1e78d97c__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\": \"mmlu_prox_lite_yo_philosophy__1d3f596a__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1921.35547241545, \"nll_base\": 7.560786191154929, \"eval_ref\": \"mmlu_prox_lite_yo_philosophy__1d3f596a__s000__tb65536__v000\", \"orac... | {
"index": "mmlu_prox_lite_yo_philosophy__1d3f596a__s000__tb65536__v000",
"uid": "mmlu_prox_lite_yo_philosophy__1d3f596a__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\": \"mmlu_prox_lite_yo_physics__cf0ab248__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3421.500747795057, \"nll_base\": 8.137834548950195, \"eval_ref\": \"mmlu_prox_lite_yo_physics__cf0ab248__s000__tb131072__v000\", \"oracle... | {
"index": "mmlu_prox_lite_yo_physics__cf0ab248__s000__tb131072__v000",
"uid": "mmlu_prox_lite_yo_physics__cf0ab248__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\": \"mmlu_prox_lite_yo_psychology__7b0f727e__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2771.456615589155, \"nll_base\": 7.927128314971924, \"eval_ref\": \"mmlu_prox_lite_yo_psychology__7b0f727e__s000__tb131072__v000\", \"... | {
"index": "mmlu_prox_lite_yo_psychology__7b0f727e__s000__tb131072__v000",
"uid": "mmlu_prox_lite_yo_psychology__7b0f727e__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\": \"mmlu_prox_lite_zh_business__087d5f8e__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 12417.57936967514, \"nll_base\": 9.426868438720703, \"eval_ref\": \"mmlu_prox_lite_zh_business__087d5f8e__s000__tb65536__v000\", \"oracle_... | {
"index": "mmlu_prox_lite_zh_business__087d5f8e__s000__tb65536__v000",
"uid": "mmlu_prox_lite_zh_business__087d5f8e__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\": \"mmlu_prox_lite_zh_philosophy__8dac874b__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 11754.22782029922, \"nll_base\": 9.371968269348145, \"eval_ref\": \"mmlu_prox_lite_zh_philosophy__8dac874b__s000__tb65536__v000\", \"ora... | {
"index": "mmlu_prox_lite_zh_philosophy__8dac874b__s000__tb65536__v000",
"uid": "mmlu_prox_lite_zh_philosophy__8dac874b__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\": \"mmlu_prox_lite_zu_biology__205bc11d__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2607.6626823927295, \"nll_base\": 7.86620957510812, \"eval_ref\": \"mmlu_prox_lite_zu_biology__205bc11d__s000__tb131072__v000\", \"oracle... | {
"index": "mmlu_prox_lite_zu_biology__205bc11d__s000__tb131072__v000",
"uid": "mmlu_prox_lite_zu_biology__205bc11d__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\": \"mmlu_prox_lite_zu_chemistry__f434879b__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4241.8957542391745, \"nll_base\": 8.352765560150146, \"eval_ref\": \"mmlu_prox_lite_zu_chemistry__f434879b__s000__tb131072__v000\", \"o... | {
"index": "mmlu_prox_lite_zu_chemistry__f434879b__s000__tb131072__v000",
"uid": "mmlu_prox_lite_zu_chemistry__f434879b__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\": \"mmlu_prox_lite_zu_computer_science__8703d858__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4436.963047922372, \"nll_base\": 8.397725423177084, \"eval_ref\": \"mmlu_prox_lite_zu_computer_science__8703d858__s000__tb65536__v... | {
"index": "mmlu_prox_lite_zu_computer_science__8703d858__s000__tb65536__v000",
"uid": "mmlu_prox_lite_zu_computer_science__8703d858__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\": \"mmlu_prox_lite_zu_law__27af0477__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2318.726925012411, \"nll_base\": 7.748773574829102, \"eval_ref\": \"mmlu_prox_lite_zu_law__27af0477__s000__tb65536__v000\", \"oracle_delta\": 0... | {
"index": "mmlu_prox_lite_zu_law__27af0477__s000__tb65536__v000",
"uid": "mmlu_prox_lite_zu_law__27af0477__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\": \"mmlu_prox_lite_zu_other__b809601e__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3959.5719343502697, \"nll_base\": 8.283891201019287, \"eval_ref\": \"mmlu_prox_lite_zu_other__b809601e__s000__tb65536__v000\", \"oracle_delta... | {
"index": "mmlu_prox_lite_zu_other__b809601e__s000__tb65536__v000",
"uid": "mmlu_prox_lite_zu_other__b809601e__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\": \"mmlu_prox_mr_economics__3d7ef691__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3569.1756957421526, \"nll_base\": 8.180089950561523, \"eval_ref\": \"mmlu_prox_mr_economics__3d7ef691__s000__tb65536__v000\", \"oracle_delta\"... | {
"index": "mmlu_prox_mr_economics__3d7ef691__s000__tb65536__v000",
"uid": "mmlu_prox_mr_economics__3d7ef691__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\": \"mmlu_prox_mr_philosophy__afe7c9b5__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4219.167175973771, \"nll_base\": 8.347393035888672, \"eval_ref\": \"mmlu_prox_mr_philosophy__afe7c9b5__s000__tb65536__v000\", \"oracle_delta\... | {
"index": "mmlu_prox_mr_philosophy__afe7c9b5__s000__tb65536__v000",
"uid": "mmlu_prox_mr_philosophy__afe7c9b5__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\": \"mmlu_prox_mr_physics__f74b9eba__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4743.253125767554, \"nll_base\": 8.464478492736816, \"eval_ref\": \"mmlu_prox_mr_physics__f74b9eba__s000__tb65536__v000\", \"oracle_delta\": 0.2... | {
"index": "mmlu_prox_mr_physics__f74b9eba__s000__tb65536__v000",
"uid": "mmlu_prox_mr_physics__f74b9eba__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\": \"mmlu_prox_mr_psychology__9750554a__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4121.253404807095, \"nll_base\": 8.323912620544434, \"eval_ref\": \"mmlu_prox_mr_psychology__9750554a__s000__tb131072__v000\", \"oracle_del... | {
"index": "mmlu_prox_mr_psychology__9750554a__s000__tb131072__v000",
"uid": "mmlu_prox_mr_psychology__9750554a__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\": \"mmlu_prox_ne_biology__668de2da__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2837.063470549088, \"nll_base\": 7.950524806976318, \"eval_ref\": \"mmlu_prox_ne_biology__668de2da__s000__tb65536__v000\", \"oracle_delta\": 0.2... | {
"index": "mmlu_prox_ne_biology__668de2da__s000__tb65536__v000",
"uid": "mmlu_prox_ne_biology__668de2da__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\": \"mmlu_prox_ne_business__8be3267d__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 5498.500940185788, \"nll_base\": 8.612230777740479, \"eval_ref\": \"mmlu_prox_ne_business__8be3267d__s000__tb262144__v000\", \"oracle_delta\"... | {
"index": "mmlu_prox_ne_business__8be3267d__s000__tb262144__v000",
"uid": "mmlu_prox_ne_business__8be3267d__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\": \"mmlu_prox_ne_engineering__bf0848a8__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 6281.698347498589, \"nll_base\": 8.74539566040039, \"eval_ref\": \"mmlu_prox_ne_engineering__bf0848a8__s000__tb131072__v000\", \"oracle_de... | {
"index": "mmlu_prox_ne_engineering__bf0848a8__s000__tb131072__v000",
"uid": "mmlu_prox_ne_engineering__bf0848a8__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\": \"mmlu_prox_ne_health__c7d1befa__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4012.9244477434086, \"nll_base\": 8.29727554321289, \"eval_ref\": \"mmlu_prox_ne_health__c7d1befa__s000__tb65536__v000\", \"oracle_delta\": 0.286... | {
"index": "mmlu_prox_ne_health__c7d1befa__s000__tb65536__v000",
"uid": "mmlu_prox_ne_health__c7d1befa__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\": \"mmlu_prox_ne_math__2567a778__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 8851.900753241647, \"nll_base\": 9.088387489318848, \"eval_ref\": \"mmlu_prox_ne_math__2567a778__s000__tb65536__v000\", \"oracle_delta\": 0.2483167... | {
"index": "mmlu_prox_ne_math__2567a778__s000__tb65536__v000",
"uid": "mmlu_prox_ne_math__2567a778__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\": \"mmlu_prox_ne_other__b2148371__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 6073.179162523428, \"nll_base\": 8.711637496948242, \"eval_ref\": \"mmlu_prox_ne_other__b2148371__s000__tb262144__v000\", \"oracle_delta\": 0.93... | {
"index": "mmlu_prox_ne_other__b2148371__s000__tb262144__v000",
"uid": "mmlu_prox_ne_other__b2148371__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\": \"mmlu_prox_ne_physics__a739450c__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 7363.128073205482, \"nll_base\": 8.904240131378174, \"eval_ref\": \"mmlu_prox_ne_physics__a739450c__s000__tb65536__v000\", \"oracle_delta\": 0.3... | {
"index": "mmlu_prox_ne_physics__a739450c__s000__tb65536__v000",
"uid": "mmlu_prox_ne_physics__a739450c__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\": \"mmlu_prox_pt_biology__025580df__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4831.231825608687, \"nll_base\": 8.482856750488281, \"eval_ref\": \"mmlu_prox_pt_biology__025580df__s000__tb131072__v000\", \"oracle_delta\": ... | {
"index": "mmlu_prox_pt_biology__025580df__s000__tb131072__v000",
"uid": "mmlu_prox_pt_biology__025580df__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\": \"mmlu_prox_pt_business__e73d9255__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 6895.392692001112, \"nll_base\": 8.838608741760254, \"eval_ref\": \"mmlu_prox_pt_business__e73d9255__s000__tb262144__v000\", \"oracle_delta\"... | {
"index": "mmlu_prox_pt_business__e73d9255__s000__tb262144__v000",
"uid": "mmlu_prox_pt_business__e73d9255__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\": \"mmlu_prox_pt_chemistry__4bcd59df__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 5579.561047524513, \"nll_base\": 8.62686538696289, \"eval_ref\": \"mmlu_prox_pt_chemistry__4bcd59df__s000__tb131072__v000\", \"oracle_delta\... | {
"index": "mmlu_prox_pt_chemistry__4bcd59df__s000__tb131072__v000",
"uid": "mmlu_prox_pt_chemistry__4bcd59df__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\": \"mmlu_prox_pt_computer_science__31719ef3__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2829.0809244391144, \"nll_base\": 7.947707176208496, \"eval_ref\": \"mmlu_prox_pt_computer_science__31719ef3__s000__tb65536__v000\", \"... | {
"index": "mmlu_prox_pt_computer_science__31719ef3__s000__tb65536__v000",
"uid": "mmlu_prox_pt_computer_science__31719ef3__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\": \"mmlu_prox_pt_engineering__50f96d8a__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4161.020534962428, \"nll_base\": 8.333515644073486, \"eval_ref\": \"mmlu_prox_pt_engineering__50f96d8a__s000__tb131072__v000\", \"oracle_d... | {
"index": "mmlu_prox_pt_engineering__50f96d8a__s000__tb131072__v000",
"uid": "mmlu_prox_pt_engineering__50f96d8a__s000__tb131072__v000",
"axis": "instruction",
"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.