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\": \"sib_fuv_prompt_2__9269b525__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1484.139625280144, \"nll_base\": 7.302590506417411, \"eval_ref\": \"sib_fuv_prompt_2__9269b525__s000__tb65536__v000\", \"oracle_delta\": 0.780377757... | {
"index": "sib_fuv_prompt_2__9269b525__s000__tb65536__v000",
"uid": "sib_fuv_prompt_2__9269b525__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\": \"sib_gaz_prompt_2__1dad271c__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1255.0347219384705, \"nll_base\": 7.1349185180664065, \"eval_ref\": \"sib_gaz_prompt_2__1dad271c__s000__tb262144__v000\", \"oracle_delta\": 1.5703... | {
"index": "sib_gaz_prompt_2__1dad271c__s000__tb262144__v000",
"uid": "sib_gaz_prompt_2__1dad271c__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\": \"sib_ibo_prompt_2__4ef8e4d1__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1398.6215192599668, \"nll_base\": 7.243242401437661, \"eval_ref\": \"sib_ibo_prompt_2__4ef8e4d1__s000__tb65536__v000\", \"oracle_delta\": 0.73500517... | {
"index": "sib_ibo_prompt_2__4ef8e4d1__s000__tb65536__v000",
"uid": "sib_ibo_prompt_2__4ef8e4d1__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\": \"sib_kab_prompt_1__10a4afe4__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 475.74609344539476, \"nll_base\": 6.164884294782366, \"eval_ref\": \"sib_kab_prompt_1__10a4afe4__s000__tb131072__v000\", \"oracle_delta\": 0.37063... | {
"index": "sib_kab_prompt_1__10a4afe4__s000__tb131072__v000",
"uid": "sib_kab_prompt_1__10a4afe4__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\": \"sib_kab_prompt_2__851e4751__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1936.8790132345032, \"nll_base\": 7.5688332005550985, \"eval_ref\": \"sib_kab_prompt_2__851e4751__s000__tb65536__v000\", \"oracle_delta\": 0.6085397... | {
"index": "sib_kab_prompt_2__851e4751__s000__tb65536__v000",
"uid": "sib_kab_prompt_2__851e4751__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\": \"sib_kbp_prompt_1__4eeb78eb__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 599.4603238700603, \"nll_base\": 6.396029790242513, \"eval_ref\": \"sib_kbp_prompt_1__4eeb78eb__s000__tb131072__v000\", \"oracle_delta\": 0.608695... | {
"index": "sib_kbp_prompt_1__4eeb78eb__s000__tb131072__v000",
"uid": "sib_kbp_prompt_1__4eeb78eb__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\": \"sib_kin_prompt_4__5ee7c5ea__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1959.0469992553756, \"nll_base\": 7.580213409109214, \"eval_ref\": \"sib_kin_prompt_4__5ee7c5ea__s000__tb131072__v000\", \"oracle_delta\": 0.87886... | {
"index": "sib_kin_prompt_4__5ee7c5ea__s000__tb131072__v000",
"uid": "sib_kin_prompt_4__5ee7c5ea__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\": \"sib_kmb_prompt_2__89255424__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2021.6840522204488, \"nll_base\": 7.611686132287466, \"eval_ref\": \"sib_kmb_prompt_2__89255424__s000__tb65536__v000\", \"oracle_delta\": 0.65162576... | {
"index": "sib_kmb_prompt_2__89255424__s000__tb65536__v000",
"uid": "sib_kmb_prompt_2__89255424__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\": \"sib_knc_prompt_2__520d9fe0__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2157.818984360813, \"nll_base\": 7.676853261095412, \"eval_ref\": \"sib_knc_prompt_2__520d9fe0__s000__tb65536__v000\", \"oracle_delta\": 0.628797815... | {
"index": "sib_knc_prompt_2__520d9fe0__s000__tb65536__v000",
"uid": "sib_knc_prompt_2__520d9fe0__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\": \"sib_lin_prompt_3__24bb7482__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 648.0568760787149, \"nll_base\": 6.473978464226974, \"eval_ref\": \"sib_lin_prompt_3__24bb7482__s000__tb131072__v000\", \"oracle_delta\": 0.635241... | {
"index": "sib_lin_prompt_3__24bb7482__s000__tb131072__v000",
"uid": "sib_lin_prompt_3__24bb7482__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\": \"sib_lin_prompt_5__0e1248ec__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2173.616857361333, \"nll_base\": 7.68414781358507, \"eval_ref\": \"sib_lin_prompt_5__0e1248ec__s000__tb131072__v000\", \"oracle_delta\": 0.0385338... | {
"index": "sib_lin_prompt_5__0e1248ec__s000__tb131072__v000",
"uid": "sib_lin_prompt_5__0e1248ec__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\": \"sib_luo_prompt_1__456c255b__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 482.37627639157137, \"nll_base\": 6.178724465910921, \"eval_ref\": \"sib_luo_prompt_1__456c255b__s000__tb131072__v000\", \"oracle_delta\": 0.40444... | {
"index": "sib_luo_prompt_1__456c255b__s000__tb131072__v000",
"uid": "sib_luo_prompt_1__456c255b__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\": \"sib_nso_prompt_4__7710714e__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1367.7080603688644, \"nll_base\": 7.2208916692450495, \"eval_ref\": \"sib_nso_prompt_4__7710714e__s000__tb65536__v000\", \"oracle_delta\": 0.2606497... | {
"index": "sib_nso_prompt_4__7710714e__s000__tb65536__v000",
"uid": "sib_nso_prompt_4__7710714e__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\": \"sib_nus_prompt_3__76aef9c1__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 526.7432565155809, \"nll_base\": 6.266713250543654, \"eval_ref\": \"sib_nus_prompt_3__76aef9c1__s000__tb131072__v000\", \"oracle_delta\": 0.603290... | {
"index": "sib_nus_prompt_3__76aef9c1__s000__tb131072__v000",
"uid": "sib_nus_prompt_3__76aef9c1__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\": \"sib_plt_prompt_1__750f4d4f__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 364.5360642594722, \"nll_base\": 5.89862548828125, \"eval_ref\": \"sib_plt_prompt_1__750f4d4f__s000__tb65536__v000\", \"oracle_delta\": 0.2059939575... | {
"index": "sib_plt_prompt_1__750f4d4f__s000__tb65536__v000",
"uid": "sib_plt_prompt_1__750f4d4f__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\": \"sib_plt_prompt_4__b0682ac2__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1636.6780984680233, \"nll_base\": 7.400423916903409, \"eval_ref\": \"sib_plt_prompt_4__b0682ac2__s000__tb262144__v000\", \"oracle_delta\": 1.80549... | {
"index": "sib_plt_prompt_4__b0682ac2__s000__tb262144__v000",
"uid": "sib_plt_prompt_4__b0682ac2__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\": \"sib_taq_prompt_1__f7653350__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 385.26058634958065, \"nll_base\": 5.953919952990962, \"eval_ref\": \"sib_taq_prompt_1__f7653350__s000__tb131072__v000\", \"oracle_delta\": 0.31059... | {
"index": "sib_taq_prompt_1__f7653350__s000__tb131072__v000",
"uid": "sib_taq_prompt_1__f7653350__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\": \"sib_tso_prompt_1__4dba2586__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 536.9938180536105, \"nll_base\": 6.285986582438151, \"eval_ref\": \"sib_tso_prompt_1__4dba2586__s000__tb131072__v000\", \"oracle_delta\": 0.326053... | {
"index": "sib_tso_prompt_1__4dba2586__s000__tb131072__v000",
"uid": "sib_tso_prompt_1__4dba2586__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\": \"sib_xho_prompt_4__9ef4ff63__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1940.714089435005, \"nll_base\": 7.5708112716674805, \"eval_ref\": \"sib_xho_prompt_4__9ef4ff63__s000__tb65536__v000\", \"oracle_delta\": 0.32732105... | {
"index": "sib_xho_prompt_4__9ef4ff63__s000__tb65536__v000",
"uid": "sib_xho_prompt_4__9ef4ff63__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\": \"sib_zul_prompt_1__87807889__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1216.6044583810258, \"nll_base\": 7.103819026503452, \"eval_ref\": \"sib_zul_prompt_1__87807889__s000__tb131072__v000\", \"oracle_delta\": 0.40041... | {
"index": "sib_zul_prompt_1__87807889__s000__tb131072__v000",
"uid": "sib_zul_prompt_1__87807889__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\": \"sib_zul_prompt_5__3ab8301f__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 900.8105624146017, \"nll_base\": 6.803294982910156, \"eval_ref\": \"sib_zul_prompt_5__3ab8301f__s000__tb131072__v000\", \"oracle_delta\": 0.073913... | {
"index": "sib_zul_prompt_5__3ab8301f__s000__tb131072__v000",
"uid": "sib_zul_prompt_5__3ab8301f__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\": \"simple_cooccurrence_bias_gen__9f859ccc__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 31885.34111887771, \"nll_base\": 10.369901657104492, \"eval_ref\": \"simple_cooccurrence_bias_gen__9f859ccc__s000__tb65536__v000\", \"or... | {
"index": "simple_cooccurrence_bias_gen__9f859ccc__s000__tb65536__v000",
"uid": "simple_cooccurrence_bias_gen__9f859ccc__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\": \"siqa_ca__8416c6f5__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 6.464804074306474, \"nll_base\": 1.866372706166217, \"eval_ref\": \"siqa_ca__8416c6f5__s000__tb262144__v000\", \"oracle_delta\": 0.3101970704485115, \"rand... | {
"index": "siqa_ca__8416c6f5__s000__tb262144__v000",
"uid": "siqa_ca__8416c6f5__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\": \"super_glue-boolq-t5-prompt__3aa8bd7f__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2734.903035802433, \"nll_base\": 7.913851261138916, \"eval_ref\": \"super_glue-boolq-t5-prompt__3aa8bd7f__s000__tb131072__v000\", \"orac... | {
"index": "super_glue-boolq-t5-prompt__3aa8bd7f__s000__tb131072__v000",
"uid": "super_glue-boolq-t5-prompt__3aa8bd7f__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\": \"super_glue-boolq-t5-prompt__3aa8bd7f__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2734.903035802433, \"nll_base\": 7.913851261138916, \"eval_ref\": \"super_glue-boolq-t5-prompt__3aa8bd7f__s000__tb262144__v001\", \"orac... | {
"index": "super_glue-boolq-t5-prompt__3aa8bd7f__s000__tb262144__v001",
"uid": "super_glue-boolq-t5-prompt__3aa8bd7f__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\": \"super_glue-cb-t5-prompt__719eb24e__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 735.1333268835265, \"nll_base\": 6.6000518798828125, \"eval_ref\": \"super_glue-cb-t5-prompt__719eb24e__s000__tb131072__v001\", \"oracle_de... | {
"index": "super_glue-cb-t5-prompt__719eb24e__s000__tb131072__v001",
"uid": "super_glue-cb-t5-prompt__719eb24e__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\": \"super_glue-cb-t5-prompt__719eb24e__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 735.1333268835265, \"nll_base\": 6.6000518798828125, \"eval_ref\": \"super_glue-cb-t5-prompt__719eb24e__s000__tb65536__v000\", \"oracle_delta... | {
"index": "super_glue-cb-t5-prompt__719eb24e__s000__tb65536__v000",
"uid": "super_glue-cb-t5-prompt__719eb24e__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\": \"super_glue-copa-t5-prompt__c2b2d1b1__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 620.5634296543091, \"nll_base\": 6.430627822875977, \"eval_ref\": \"super_glue-copa-t5-prompt__c2b2d1b1__s000__tb131072__v000\", \"oracle... | {
"index": "super_glue-copa-t5-prompt__c2b2d1b1__s000__tb131072__v000",
"uid": "super_glue-copa-t5-prompt__c2b2d1b1__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\": \"super_glue-copa-t5-prompt__c2b2d1b1__s000__tb65536__v002\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 620.5634296543091, \"nll_base\": 6.430627822875977, \"eval_ref\": \"super_glue-copa-t5-prompt__c2b2d1b1__s000__tb65536__v002\", \"oracle_de... | {
"index": "super_glue-copa-t5-prompt__c2b2d1b1__s000__tb65536__v002",
"uid": "super_glue-copa-t5-prompt__c2b2d1b1__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\": \"super_glue-multirc-t5-prompt__fbf703de__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 679.18402060278, \"nll_base\": 6.5208921079282405, \"eval_ref\": \"super_glue-multirc-t5-prompt__fbf703de__s000__tb131072__v001\", \"o... | {
"index": "super_glue-multirc-t5-prompt__fbf703de__s000__tb131072__v001",
"uid": "super_glue-multirc-t5-prompt__fbf703de__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\": \"super_glue-multirc-t5-prompt__fbf703de__s000__tb262144__v002\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 679.18402060278, \"nll_base\": 6.5208921079282405, \"eval_ref\": \"super_glue-multirc-t5-prompt__fbf703de__s000__tb262144__v002\", \"o... | {
"index": "super_glue-multirc-t5-prompt__fbf703de__s000__tb262144__v002",
"uid": "super_glue-multirc-t5-prompt__fbf703de__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\": \"super_glue-multirc-t5-prompt__fbf703de__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 679.18402060278, \"nll_base\": 6.5208921079282405, \"eval_ref\": \"super_glue-multirc-t5-prompt__fbf703de__s000__tb65536__v000\", \"orac... | {
"index": "super_glue-multirc-t5-prompt__fbf703de__s000__tb65536__v000",
"uid": "super_glue-multirc-t5-prompt__fbf703de__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\": \"super_glue-record-t5-prompt__579a4fde__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 65.50482009438176, \"nll_base\": 4.182123729160854, \"eval_ref\": \"super_glue-record-t5-prompt__579a4fde__s000__tb131072__v001\", \"or... | {
"index": "super_glue-record-t5-prompt__579a4fde__s000__tb131072__v001",
"uid": "super_glue-record-t5-prompt__579a4fde__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\": \"tatoeba_eng_nno_p3__bb0bd0fe__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 179.7675779996031, \"nll_base\": 5.191664783194596, \"eval_ref\": \"tatoeba_eng_nno_p3__bb0bd0fe__s000__tb65536__v000\", \"oracle_delta\": 0.13729... | {
"index": "tatoeba_eng_nno_p3__bb0bd0fe__s000__tb65536__v000",
"uid": "tatoeba_eng_nno_p3__bb0bd0fe__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\": \"tinyTruthfulQA__17f34677__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3.5168246121749385, \"nll_base\": 1.257558483660063, \"eval_ref\": \"tinyTruthfulQA__17f34677__s000__tb262144__v000\", \"oracle_delta\": 0.281485114... | {
"index": "tinyTruthfulQA__17f34677__s000__tb262144__v000",
"uid": "tinyTruthfulQA__17f34677__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\": \"tinyTruthfulQA_mc1__be3e1598__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3.2840339423319778, \"nll_base\": 1.189072527204241, \"eval_ref\": \"tinyTruthfulQA_mc1__be3e1598__s000__tb131072__v000\", \"oracle_delta\": 0.1... | {
"index": "tinyTruthfulQA_mc1__be3e1598__s000__tb131072__v000",
"uid": "tinyTruthfulQA_mc1__be3e1598__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\": \"tmmluplus_culinary_skills__8b8e07e3__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1311.842001071448, \"nll_base\": 7.179187536239624, \"eval_ref\": \"tmmluplus_culinary_skills__8b8e07e3__s000__tb65536__v000\", \"oracle_de... | {
"index": "tmmluplus_culinary_skills__8b8e07e3__s000__tb65536__v000",
"uid": "tmmluplus_culinary_skills__8b8e07e3__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\": \"tmmluplus_financial_analysis__829a7ea6__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1240.8034766908463, \"nll_base\": 7.123514413833618, \"eval_ref\": \"tmmluplus_financial_analysis__829a7ea6__s000__tb131072__v000\", \... | {
"index": "tmmluplus_financial_analysis__829a7ea6__s000__tb131072__v000",
"uid": "tmmluplus_financial_analysis__829a7ea6__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\": \"tmmluplus_introduction_to_law__0dc50954__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 889.9390294126149, \"nll_base\": 6.7911529541015625, \"eval_ref\": \"tmmluplus_introduction_to_law__0dc50954__s000__tb131072__v000\",... | {
"index": "tmmluplus_introduction_to_law__0dc50954__s000__tb131072__v000",
"uid": "tmmluplus_introduction_to_law__0dc50954__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\": \"tmmluplus_junior_chemistry__eb8c8be1__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1137.7978646579163, \"nll_base\": 7.0368499755859375, \"eval_ref\": \"tmmluplus_junior_chemistry__eb8c8be1__s000__tb65536__v000\", \"oracl... | {
"index": "tmmluplus_junior_chemistry__eb8c8be1__s000__tb65536__v000",
"uid": "tmmluplus_junior_chemistry__eb8c8be1__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\": \"tmmluplus_junior_science_exam__4e0e00ee__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1193.014443866572, \"nll_base\": 7.084238529205322, \"eval_ref\": \"tmmluplus_junior_science_exam__4e0e00ee__s000__tb131072__v000\", ... | {
"index": "tmmluplus_junior_science_exam__4e0e00ee__s000__tb131072__v000",
"uid": "tmmluplus_junior_science_exam__4e0e00ee__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\": \"tmmluplus_politic_science__19fcbd00__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1541.0911171276855, \"nll_base\": 7.340245962142944, \"eval_ref\": \"tmmluplus_politic_science__19fcbd00__s000__tb131072__v000\", \"oracl... | {
"index": "tmmluplus_politic_science__19fcbd00__s000__tb131072__v000",
"uid": "tmmluplus_politic_science__19fcbd00__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\": \"tmmluplus_statistics_and_machine_learning__f2f7ca3c__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1334.2257417600963, \"nll_base\": 7.196106433868408, \"eval_ref\": \"tmmluplus_statistics_and_machine_learning__f2f7ca3c_... | {
"index": "tmmluplus_statistics_and_machine_learning__f2f7ca3c__s000__tb131072__v000",
"uid": "tmmluplus_statistics_and_machine_learning__f2f7ca3c__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\": \"tmmluplus_veterinary_pharmacology__56f29da2__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 912.375576439692, \"nll_base\": 6.816051721572876, \"eval_ref\": \"tmmluplus_veterinary_pharmacology__56f29da2__s000__tb65536__v000... | {
"index": "tmmluplus_veterinary_pharmacology__56f29da2__s000__tb65536__v000",
"uid": "tmmluplus_veterinary_pharmacology__56f29da2__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\": \"toksuite_chinese_colloquial__260cb378__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 11.80055010229286, \"nll_base\": 2.4681461492233727, \"eval_ref\": \"toksuite_chinese_colloquial__260cb378__s000__tb65536__v000\", \"orac... | {
"index": "toksuite_chinese_colloquial__260cb378__s000__tb65536__v000",
"uid": "toksuite_chinese_colloquial__260cb378__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\": \"toksuite_english_capitalization__fb968a24__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 28.73479152911895, \"nll_base\": 3.3581086369871183, \"eval_ref\": \"toksuite_english_capitalization__fb968a24__s000__tb131072__v00... | {
"index": "toksuite_english_capitalization__fb968a24__s000__tb131072__v000",
"uid": "toksuite_english_capitalization__fb968a24__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\": \"toksuite_english_compounds__2517ced0__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 73.34061790759796, \"nll_base\": 4.295114587854456, \"eval_ref\": \"toksuite_english_compounds__2517ced0__s000__tb131072__v000\", \"orac... | {
"index": "toksuite_english_compounds__2517ced0__s000__tb131072__v000",
"uid": "toksuite_english_compounds__2517ced0__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\": \"toksuite_english_date_formats__99b2584a__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2.595380315063888, \"nll_base\": 0.9537330627441406, \"eval_ref\": \"toksuite_english_date_formats__99b2584a__s000__tb131072__v000\",... | {
"index": "toksuite_english_date_formats__99b2584a__s000__tb131072__v000",
"uid": "toksuite_english_date_formats__99b2584a__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\": \"toksuite_english_historical_spelling__2bcc9bfa__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 60.55794395875528, \"nll_base\": 4.103600658069958, \"eval_ref\": \"toksuite_english_historical_spelling__2bcc9bfa__s000__tb6553... | {
"index": "toksuite_english_historical_spelling__2bcc9bfa__s000__tb65536__v000",
"uid": "toksuite_english_historical_spelling__2bcc9bfa__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\": \"toksuite_english_keyboard_proximity_errors__dc7f46b3__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 84.61546619351921, \"nll_base\": 4.4381170654296875, \"eval_ref\": \"toksuite_english_keyboard_proximity_errors__dc7f46b... | {
"index": "toksuite_english_keyboard_proximity_errors__dc7f46b3__s000__tb131072__v000",
"uid": "toksuite_english_keyboard_proximity_errors__dc7f46b3__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\": \"toksuite_english_orthographic_errors__7ef98905__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 168.2634301419321, \"nll_base\": 5.1255307878766745, \"eval_ref\": \"toksuite_english_orthographic_errors__7ef98905__s000__tb1... | {
"index": "toksuite_english_orthographic_errors__7ef98905__s000__tb131072__v000",
"uid": "toksuite_english_orthographic_errors__7ef98905__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\": \"toksuite_english_space_removal__d160bcef__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 60.56902869606252, \"nll_base\": 4.103783684807855, \"eval_ref\": \"toksuite_english_space_removal__d160bcef__s000__tb65536__v000\", \... | {
"index": "toksuite_english_space_removal__d160bcef__s000__tb65536__v000",
"uid": "toksuite_english_space_removal__d160bcef__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\": \"toksuite_english_web_search_query__ac2bcb3e__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 83.84731643167254, \"nll_base\": 4.4289974833643715, \"eval_ref\": \"toksuite_english_web_search_query__ac2bcb3e__s000__tb65536__v0... | {
"index": "toksuite_english_web_search_query__ac2bcb3e__s000__tb65536__v000",
"uid": "toksuite_english_web_search_query__ac2bcb3e__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\": \"toksuite_italian_code_language_script_switching__297645ec__s012__tb65536__v036\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 26.373777940055074, \"nll_base\": 3.2723702566964286, \"eval_ref\": \"toksuite_italian_code_language_script_switching... | {
"index": "toksuite_italian_code_language_script_switching__297645ec__s012__tb65536__v036",
"uid": "toksuite_italian_code_language_script_switching__297645ec__s012__tb65536__v036",
"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\": \"toksuite_stem_equivalent_expressions__3e97b5b0__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 260.67828755016114, \"nll_base\": 5.563287032277961, \"eval_ref\": \"toksuite_stem_equivalent_expressions__3e97b5b0__s000__tb2... | {
"index": "toksuite_stem_equivalent_expressions__3e97b5b0__s000__tb262144__v000",
"uid": "toksuite_stem_equivalent_expressions__3e97b5b0__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\": \"toksuite_stem_scripted_text__2a51778a__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1.837245883558939, \"nll_base\": 0.6082676478794643, \"eval_ref\": \"toksuite_stem_scripted_text__2a51778a__s000__tb131072__v000\", \"o... | {
"index": "toksuite_stem_scripted_text__2a51778a__s000__tb131072__v000",
"uid": "toksuite_stem_scripted_text__2a51778a__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\": \"toksuite_turkish_code_language_script_switching__62fe6f8f__s001__tb131072__v005\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 11.113663386003589, \"nll_base\": 2.4081752870141004, \"eval_ref\": \"toksuite_turkish_code_language_script_switchi... | {
"index": "toksuite_turkish_code_language_script_switching__62fe6f8f__s001__tb131072__v005",
"uid": "toksuite_turkish_code_language_script_switching__62fe6f8f__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\": \"toksuite_turkish_code_language_script_switching__62fe6f8f__s001__tb262144__v003\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 11.113663386003589, \"nll_base\": 2.4081752870141004, \"eval_ref\": \"toksuite_turkish_code_language_script_switchi... | {
"index": "toksuite_turkish_code_language_script_switching__62fe6f8f__s001__tb262144__v003",
"uid": "toksuite_turkish_code_language_script_switching__62fe6f8f__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\": \"toksuite_turkish_code_language_script_switching__62fe6f8f__s001__tb65536__v004\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 11.113663386003589, \"nll_base\": 2.4081752870141004, \"eval_ref\": \"toksuite_turkish_code_language_script_switching... | {
"index": "toksuite_turkish_code_language_script_switching__62fe6f8f__s001__tb65536__v004",
"uid": "toksuite_turkish_code_language_script_switching__62fe6f8f__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\": \"toksuite_turkish_code_language_script_switching__62fe6f8f__s003__tb131072__v011\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 11.07193392385557, \"nll_base\": 2.4044134309976406, \"eval_ref\": \"toksuite_turkish_code_language_script_switchin... | {
"index": "toksuite_turkish_code_language_script_switching__62fe6f8f__s003__tb131072__v011",
"uid": "toksuite_turkish_code_language_script_switching__62fe6f8f__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\": \"toksuite_turkish_code_language_script_switching__62fe6f8f__s003__tb262144__v009\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 11.07193392385557, \"nll_base\": 2.4044134309976406, \"eval_ref\": \"toksuite_turkish_code_language_script_switchin... | {
"index": "toksuite_turkish_code_language_script_switching__62fe6f8f__s003__tb262144__v009",
"uid": "toksuite_turkish_code_language_script_switching__62fe6f8f__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\": \"toksuite_turkish_code_language_script_switching__62fe6f8f__s003__tb65536__v010\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 11.07193392385557, \"nll_base\": 2.4044134309976406, \"eval_ref\": \"toksuite_turkish_code_language_script_switching_... | {
"index": "toksuite_turkish_code_language_script_switching__62fe6f8f__s003__tb65536__v010",
"uid": "toksuite_turkish_code_language_script_switching__62fe6f8f__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\": \"toksuite_turkish_code_language_script_switching__62fe6f8f__s004__tb262144__v012\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 14.354450577654871, \"nll_base\": 2.66406003890499, \"eval_ref\": \"toksuite_turkish_code_language_script_switching... | {
"index": "toksuite_turkish_code_language_script_switching__62fe6f8f__s004__tb262144__v012",
"uid": "toksuite_turkish_code_language_script_switching__62fe6f8f__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\": \"toksuite_turkish_code_language_script_switching__62fe6f8f__s006__tb65536__v019\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 12.86450527796002, \"nll_base\": 2.5544719901136173, \"eval_ref\": \"toksuite_turkish_code_language_script_switching_... | {
"index": "toksuite_turkish_code_language_script_switching__62fe6f8f__s006__tb65536__v019",
"uid": "toksuite_turkish_code_language_script_switching__62fe6f8f__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\": \"toksuite_turkish_code_language_script_switching__62fe6f8f__s008__tb65536__v025\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 11.988164321479058, \"nll_base\": 2.4839198565242264, \"eval_ref\": \"toksuite_turkish_code_language_script_switching... | {
"index": "toksuite_turkish_code_language_script_switching__62fe6f8f__s008__tb65536__v025",
"uid": "toksuite_turkish_code_language_script_switching__62fe6f8f__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\": \"toksuite_turkish_code_language_script_switching__62fe6f8f__s009__tb131072__v029\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 10.889080916236633, \"nll_base\": 2.3877605363434435, \"eval_ref\": \"toksuite_turkish_code_language_script_switchi... | {
"index": "toksuite_turkish_code_language_script_switching__62fe6f8f__s009__tb131072__v029",
"uid": "toksuite_turkish_code_language_script_switching__62fe6f8f__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\": \"toksuite_turkish_code_language_script_switching__62fe6f8f__s009__tb262144__v027\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 10.889080916236633, \"nll_base\": 2.3877605363434435, \"eval_ref\": \"toksuite_turkish_code_language_script_switchi... | {
"index": "toksuite_turkish_code_language_script_switching__62fe6f8f__s009__tb262144__v027",
"uid": "toksuite_turkish_code_language_script_switching__62fe6f8f__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\": \"toksuite_turkish_code_language_script_switching__62fe6f8f__s010__tb131072__v032\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 11.271983221020879, \"nll_base\": 2.42232028604141, \"eval_ref\": \"toksuite_turkish_code_language_script_switching... | {
"index": "toksuite_turkish_code_language_script_switching__62fe6f8f__s010__tb131072__v032",
"uid": "toksuite_turkish_code_language_script_switching__62fe6f8f__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\": \"toksuite_turkish_code_language_script_switching__62fe6f8f__s010__tb262144__v030\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 11.271983221020879, \"nll_base\": 2.42232028604141, \"eval_ref\": \"toksuite_turkish_code_language_script_switching... | {
"index": "toksuite_turkish_code_language_script_switching__62fe6f8f__s010__tb262144__v030",
"uid": "toksuite_turkish_code_language_script_switching__62fe6f8f__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\": \"toksuite_turkish_code_language_script_switching__62fe6f8f__s010__tb65536__v031\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 11.271983221020879, \"nll_base\": 2.42232028604141, \"eval_ref\": \"toksuite_turkish_code_language_script_switching__... | {
"index": "toksuite_turkish_code_language_script_switching__62fe6f8f__s010__tb65536__v031",
"uid": "toksuite_turkish_code_language_script_switching__62fe6f8f__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\": \"toksuite_turkish_code_language_script_switching__62fe6f8f__s011__tb262144__v033\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 11.19719689997214, \"nll_base\": 2.415663470188385, \"eval_ref\": \"toksuite_turkish_code_language_script_switching... | {
"index": "toksuite_turkish_code_language_script_switching__62fe6f8f__s011__tb262144__v033",
"uid": "toksuite_turkish_code_language_script_switching__62fe6f8f__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\": \"toksuite_turkish_code_language_script_switching__62fe6f8f__s014__tb262144__v042\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 12.735155988523495, \"nll_base\": 2.544366357160762, \"eval_ref\": \"toksuite_turkish_code_language_script_switchin... | {
"index": "toksuite_turkish_code_language_script_switching__62fe6f8f__s014__tb262144__v042",
"uid": "toksuite_turkish_code_language_script_switching__62fe6f8f__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\": \"toksuite_turkish_derivations__03bfe1b5__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 12.27961530873331, \"nll_base\": 2.507940595577925, \"eval_ref\": \"toksuite_turkish_derivations__03bfe1b5__s000__tb131072__v000\", \"... | {
"index": "toksuite_turkish_derivations__03bfe1b5__s000__tb131072__v000",
"uid": "toksuite_turkish_derivations__03bfe1b5__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\": \"toxigen__56a09ba9__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 998.4820459996595, \"nll_base\": 6.906236171722412, \"eval_ref\": \"toxigen__56a09ba9__s000__tb131072__v000\", \"oracle_delta\": 1.0341463088989258, \"rand... | {
"index": "toxigen__56a09ba9__s000__tb131072__v000",
"uid": "toxigen__56a09ba9__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\": \"translation_fr_dr_doda__0361b6e1__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 27.073712029209545, \"nll_base\": 3.298563221273901, \"eval_ref\": \"translation_fr_dr_doda__0361b6e1__s000__tb262144__v000\", \"oracle_delt... | {
"index": "translation_fr_dr_doda__0361b6e1__s000__tb262144__v000",
"uid": "translation_fr_dr_doda__0361b6e1__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\": \"translation_fr_dr_flores__a8d26e7e__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 11.695493954279687, \"nll_base\": 2.4592036355084437, \"eval_ref\": \"translation_fr_dr_flores__a8d26e7e__s000__tb262144__v000\", \"oracle... | {
"index": "translation_fr_dr_flores__a8d26e7e__s000__tb262144__v000",
"uid": "translation_fr_dr_flores__a8d26e7e__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\": \"triviaqa__569170be__s000__tb131072__v002\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 254.4838452304454, \"nll_base\": 5.539237357474662, \"eval_ref\": \"triviaqa__569170be__s000__tb131072__v002\", \"oracle_delta\": 0.16387461069467957, \"r... | {
"index": "triviaqa__569170be__s000__tb131072__v002",
"uid": "triviaqa__569170be__s000__tb131072__v002",
"axis": "qa",
"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\": \"triviaqa__569170be__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 254.4838452304454, \"nll_base\": 5.539237357474662, \"eval_ref\": \"triviaqa__569170be__s000__tb65536__v001\", \"oracle_delta\": 0.09437024915540526, \"rand... | {
"index": "triviaqa__569170be__s000__tb65536__v001",
"uid": "triviaqa__569170be__s000__tb65536__v001",
"axis": "qa",
"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\": \"triviaqa__569170be__s001__tb131072__v005\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 381.7356320642387, \"nll_base\": 5.944728306361607, \"eval_ref\": \"triviaqa__569170be__s001__tb131072__v005\", \"oracle_delta\": 0.3408343975360575, \"ra... | {
"index": "triviaqa__569170be__s001__tb131072__v005",
"uid": "triviaqa__569170be__s001__tb131072__v005",
"axis": "qa",
"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\": \"triviaqa__569170be__s001__tb262144__v003\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 381.7356320642387, \"nll_base\": 5.944728306361607, \"eval_ref\": \"triviaqa__569170be__s001__tb262144__v003\", \"oracle_delta\": 0.5549199030949516, \"ra... | {
"index": "triviaqa__569170be__s001__tb262144__v003",
"uid": "triviaqa__569170be__s001__tb262144__v003",
"axis": "qa",
"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\": \"triviaqa__569170be__s001__tb65536__v004\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 381.7356320642387, \"nll_base\": 5.944728306361607, \"eval_ref\": \"triviaqa__569170be__s001__tb65536__v004\", \"oracle_delta\": 0.1855341313959471, \"rando... | {
"index": "triviaqa__569170be__s001__tb65536__v004",
"uid": "triviaqa__569170be__s001__tb65536__v004",
"axis": "qa",
"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\": \"triviaqa__569170be__s002__tb262144__v006\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 390.3014168525606, \"nll_base\": 5.966919304342831, \"eval_ref\": \"triviaqa__569170be__s002__tb262144__v006\", \"oracle_delta\": 0.4507764030905328, \"ra... | {
"index": "triviaqa__569170be__s002__tb262144__v006",
"uid": "triviaqa__569170be__s002__tb262144__v006",
"axis": "qa",
"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\": \"triviaqa__569170be__s002__tb65536__v007\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 390.3014168525606, \"nll_base\": 5.966919304342831, \"eval_ref\": \"triviaqa__569170be__s002__tb65536__v007\", \"oracle_delta\": 0.13761309455422754, \"rand... | {
"index": "triviaqa__569170be__s002__tb65536__v007",
"uid": "triviaqa__569170be__s002__tb65536__v007",
"axis": "qa",
"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\": \"truthfulqa_mc2__6883bda0__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3.5028773628531216, \"nll_base\": 1.2535847344255577, \"eval_ref\": \"truthfulqa_mc2__6883bda0__s000__tb131072__v000\", \"oracle_delta\": 0.11843959... | {
"index": "truthfulqa_mc2__6883bda0__s000__tb131072__v000",
"uid": "truthfulqa_mc2__6883bda0__s000__tb131072__v000",
"axis": "knowledge",
"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\": \"truthfulqa_mc2__6883bda0__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3.5028773628531216, \"nll_base\": 1.2535847344255577, \"eval_ref\": \"truthfulqa_mc2__6883bda0__s000__tb262144__v001\", \"oracle_delta\": 0.25273701... | {
"index": "truthfulqa_mc2__6883bda0__s000__tb262144__v001",
"uid": "truthfulqa_mc2__6883bda0__s000__tb262144__v001",
"axis": "knowledge",
"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\": \"truthfulqa_mr_mc1__dee683cd__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1.2857120093814043, \"nll_base\": 0.2513126577982088, \"eval_ref\": \"truthfulqa_mr_mc1__dee683cd__s000__tb65536__v000\", \"oracle_delta\": 0.01359... | {
"index": "truthfulqa_mr_mc1__dee683cd__s000__tb65536__v000",
"uid": "truthfulqa_mr_mc1__dee683cd__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\": \"truthfulqa_ne_mc1__8bd050ed__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1.2883494776324929, \"nll_base\": 0.2533619244434674, \"eval_ref\": \"truthfulqa_ne_mc1__8bd050ed__s000__tb65536__v000\", \"oracle_delta\": 0.01229... | {
"index": "truthfulqa_ne_mc1__8bd050ed__s000__tb65536__v000",
"uid": "truthfulqa_ne_mc1__8bd050ed__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\": \"truthfulqa_ru_mc1__c82aaedd__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1.4270129481669227, \"nll_base\": 0.3555834121514473, \"eval_ref\": \"truthfulqa_ru_mc1__c82aaedd__s000__tb131072__v000\", \"oracle_delta\": 0.01... | {
"index": "truthfulqa_ru_mc1__c82aaedd__s000__tb131072__v000",
"uid": "truthfulqa_ru_mc1__c82aaedd__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\": \"uhura-arc-easy_ha_prompt_5__8aca8c90__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4113.237434701568, \"nll_base\": 8.32196569442749, \"eval_ref\": \"uhura-arc-easy_ha_prompt_5__8aca8c90__s000__tb65536__v000\", \"oracle_d... | {
"index": "uhura-arc-easy_ha_prompt_5__8aca8c90__s000__tb65536__v000",
"uid": "uhura-arc-easy_ha_prompt_5__8aca8c90__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\": \"wag__d47d8b0a__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 5.871728668271837, \"nll_base\": 1.770149082525231, \"eval_ref\": \"wag__d47d8b0a__s000__tb131072__v000\", \"oracle_delta\": 0.08743838299109319, \"random_delt... | {
"index": "wag__d47d8b0a__s000__tb131072__v000",
"uid": "wag__d47d8b0a__s000__tb131072__v000",
"axis": "qa",
"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\": \"wag__d47d8b0a__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 5.871728668271837, \"nll_base\": 1.770149082525231, \"eval_ref\": \"wag__d47d8b0a__s000__tb262144__v001\", \"oracle_delta\": 0.14742436390442615, \"random_delt... | {
"index": "wag__d47d8b0a__s000__tb262144__v001",
"uid": "wag__d47d8b0a__s000__tb262144__v001",
"axis": "qa",
"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\": \"wag__d47d8b0a__s000__tb65536__v002\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 5.871728668271837, \"nll_base\": 1.770149082525231, \"eval_ref\": \"wag__d47d8b0a__s000__tb65536__v002\", \"oracle_delta\": 0.03707517631322954, \"random_delta\"... | {
"index": "wag__d47d8b0a__s000__tb65536__v002",
"uid": "wag__d47d8b0a__s000__tb65536__v002",
"axis": "qa",
"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\": \"wag__d47d8b0a__s001__tb131072__v003\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 6.194481262211265, \"nll_base\": 1.8236587766617063, \"eval_ref\": \"wag__d47d8b0a__s001__tb131072__v003\", \"oracle_delta\": 0.083478534032428, \"random_delta... | {
"index": "wag__d47d8b0a__s001__tb131072__v003",
"uid": "wag__d47d8b0a__s001__tb131072__v003",
"axis": "qa",
"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\": \"wag__d47d8b0a__s001__tb262144__v004\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 6.194481262211265, \"nll_base\": 1.8236587766617063, \"eval_ref\": \"wag__d47d8b0a__s001__tb262144__v004\", \"oracle_delta\": 0.13147075592525415, \"random_del... | {
"index": "wag__d47d8b0a__s001__tb262144__v004",
"uid": "wag__d47d8b0a__s001__tb262144__v004",
"axis": "qa",
"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\": \"wag__d47d8b0a__s001__tb65536__v005\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 6.194481262211265, \"nll_base\": 1.8236587766617063, \"eval_ref\": \"wag__d47d8b0a__s001__tb65536__v005\", \"oracle_delta\": 0.030432080465649713, \"random_delta... | {
"index": "wag__d47d8b0a__s001__tb65536__v005",
"uid": "wag__d47d8b0a__s001__tb65536__v005",
"axis": "qa",
"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\": \"wag__d47d8b0a__s002__tb131072__v006\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 6.707092117912381, \"nll_base\": 1.9031654916158536, \"eval_ref\": \"wag__d47d8b0a__s002__tb131072__v006\", \"oracle_delta\": 0.058125875829680984, \"random_de... | {
"index": "wag__d47d8b0a__s002__tb131072__v006",
"uid": "wag__d47d8b0a__s002__tb131072__v006",
"axis": "qa",
"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\": \"wag__d47d8b0a__s002__tb262144__v007\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 6.707092117912381, \"nll_base\": 1.9031654916158536, \"eval_ref\": \"wag__d47d8b0a__s002__tb262144__v007\", \"oracle_delta\": 0.13788697777724845, \"random_del... | {
"index": "wag__d47d8b0a__s002__tb262144__v007",
"uid": "wag__d47d8b0a__s002__tb262144__v007",
"axis": "qa",
"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\": \"wag__d47d8b0a__s002__tb65536__v008\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 6.707092117912381, \"nll_base\": 1.9031654916158536, \"eval_ref\": \"wag__d47d8b0a__s002__tb65536__v008\", \"oracle_delta\": 0.00971376992822659, \"random_delta\... | {
"index": "wag__d47d8b0a__s002__tb65536__v008",
"uid": "wag__d47d8b0a__s002__tb65536__v008",
"axis": "qa",
"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\": \"winogrande__08141beb__s000__tb131072__v002\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 146.08593209217912, \"nll_base\": 4.984195024539263, \"eval_ref\": \"winogrande__08141beb__s000__tb131072__v002\", \"oracle_delta\": 0.02285584017761710... | {
"index": "winogrande__08141beb__s000__tb131072__v002",
"uid": "winogrande__08141beb__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\": \"wmdp_bio__a6d5d642__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 789.2002026878009, \"nll_base\": 6.671020030975342, \"eval_ref\": \"wmdp_bio__a6d5d642__s000__tb131072__v001\", \"oracle_delta\": 0.45104050636291504, \"r... | {
"index": "wmdp_bio__a6d5d642__s000__tb131072__v001",
"uid": "wmdp_bio__a6d5d642__s000__tb131072__v001",
"axis": "domain",
"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.