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_id_chemistry__65b8a484__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 8530.73413809371, \"nll_base\": 9.051430702209473, \"eval_ref\": \"mmlu_prox_id_chemistry__65b8a484__s000__tb262144__v000\", \"oracle_delta\... | {
"index": "mmlu_prox_id_chemistry__65b8a484__s000__tb262144__v000",
"uid": "mmlu_prox_id_chemistry__65b8a484__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_id_computer_science__0f966abe__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 6029.280101390105, \"nll_base\": 8.70438289642334, \"eval_ref\": \"mmlu_prox_id_computer_science__0f966abe__s000__tb65536__v000\", \"or... | {
"index": "mmlu_prox_id_computer_science__0f966abe__s000__tb65536__v000",
"uid": "mmlu_prox_id_computer_science__0f966abe__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_id_economics__675d66e3__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 8891.836006429166, \"nll_base\": 9.092888832092285, \"eval_ref\": \"mmlu_prox_id_economics__675d66e3__s000__tb131072__v000\", \"oracle_delta... | {
"index": "mmlu_prox_id_economics__675d66e3__s000__tb131072__v000",
"uid": "mmlu_prox_id_economics__675d66e3__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_id_health__d28d90d9__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 9576.135343932086, \"nll_base\": 9.16702938079834, \"eval_ref\": \"mmlu_prox_id_health__d28d90d9__s000__tb131072__v000\", \"oracle_delta\": 1.3... | {
"index": "mmlu_prox_id_health__d28d90d9__s000__tb131072__v000",
"uid": "mmlu_prox_id_health__d28d90d9__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_id_law__ac27bf2b__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3794.217016751941, \"nll_base\": 8.241233348846436, \"eval_ref\": \"mmlu_prox_id_law__ac27bf2b__s000__tb262144__v000\", \"oracle_delta\": 1.865232... | {
"index": "mmlu_prox_id_law__ac27bf2b__s000__tb262144__v000",
"uid": "mmlu_prox_id_law__ac27bf2b__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_id_philosophy__cd52b1c3__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 9449.373193664544, \"nll_base\": 9.153703689575195, \"eval_ref\": \"mmlu_prox_id_philosophy__cd52b1c3__s000__tb131072__v000\", \"oracle_del... | {
"index": "mmlu_prox_id_philosophy__cd52b1c3__s000__tb131072__v000",
"uid": "mmlu_prox_id_philosophy__cd52b1c3__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_id_physics__d630b10e__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 10542.005039243304, \"nll_base\": 9.263123035430908, \"eval_ref\": \"mmlu_prox_id_physics__d630b10e__s000__tb65536__v000\", \"oracle_delta\": 0.... | {
"index": "mmlu_prox_id_physics__d630b10e__s000__tb65536__v000",
"uid": "mmlu_prox_id_physics__d630b10e__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_id_psychology__81fd12ea__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 5045.876511462278, \"nll_base\": 8.526326656341553, \"eval_ref\": \"mmlu_prox_id_psychology__81fd12ea__s000__tb131072__v000\", \"oracle_del... | {
"index": "mmlu_prox_id_psychology__81fd12ea__s000__tb131072__v000",
"uid": "mmlu_prox_id_psychology__81fd12ea__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_it_business__fb2332f6__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3905.0606407559076, \"nll_base\": 8.270028591156006, \"eval_ref\": \"mmlu_prox_it_business__fb2332f6__s000__tb262144__v000\", \"oracle_delta\... | {
"index": "mmlu_prox_it_business__fb2332f6__s000__tb262144__v000",
"uid": "mmlu_prox_it_business__fb2332f6__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_it_engineering__8db36fed__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2292.99074686742, \"nll_base\": 7.737612247467041, \"eval_ref\": \"mmlu_prox_it_engineering__8db36fed__s000__tb131072__v000\", \"oracle_de... | {
"index": "mmlu_prox_it_engineering__8db36fed__s000__tb131072__v000",
"uid": "mmlu_prox_it_engineering__8db36fed__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_it_history__704219ac__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2230.467360348665, \"nll_base\": 7.709966421127319, \"eval_ref\": \"mmlu_prox_it_history__704219ac__s000__tb262144__v000\", \"oracle_delta\": ... | {
"index": "mmlu_prox_it_history__704219ac__s000__tb262144__v000",
"uid": "mmlu_prox_it_history__704219ac__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_it_other__8d716e5c__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3044.433144917244, \"nll_base\": 8.021070003509521, \"eval_ref\": \"mmlu_prox_it_other__8d716e5c__s000__tb262144__v000\", \"oracle_delta\": 2.06... | {
"index": "mmlu_prox_it_other__8d716e5c__s000__tb262144__v000",
"uid": "mmlu_prox_it_other__8d716e5c__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_ja_biology__2755bb62__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4401.11480043449, \"nll_base\": 8.389613151550293, \"eval_ref\": \"mmlu_prox_ja_biology__2755bb62__s000__tb131072__v000\", \"oracle_delta\": 0... | {
"index": "mmlu_prox_ja_biology__2755bb62__s000__tb131072__v000",
"uid": "mmlu_prox_ja_biology__2755bb62__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_ja_business__31180192__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 4792.590392133514, \"nll_base\": 8.474826335906982, \"eval_ref\": \"mmlu_prox_ja_business__31180192__s000__tb262144__v000\", \"oracle_delta\"... | {
"index": "mmlu_prox_ja_business__31180192__s000__tb262144__v000",
"uid": "mmlu_prox_ja_business__31180192__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_ja_engineering__5ec5e6d7__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3218.0670993351246, \"nll_base\": 8.076536178588867, \"eval_ref\": \"mmlu_prox_ja_engineering__5ec5e6d7__s000__tb65536__v000\", \"oracle_del... | {
"index": "mmlu_prox_ja_engineering__5ec5e6d7__s000__tb65536__v000",
"uid": "mmlu_prox_ja_engineering__5ec5e6d7__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_ja_psychology__11045183__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 6843.784672810134, \"nll_base\": 8.831096172332764, \"eval_ref\": \"mmlu_prox_ja_psychology__11045183__s000__tb65536__v000\", \"oracle_delta\... | {
"index": "mmlu_prox_ja_psychology__11045183__s000__tb65536__v000",
"uid": "mmlu_prox_ja_psychology__11045183__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_ko_biology__66a24b3a__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3604.4388510582908, \"nll_base\": 8.189921379089355, \"eval_ref\": \"mmlu_prox_ko_biology__66a24b3a__s000__tb65536__v000\", \"oracle_delta\": 0.... | {
"index": "mmlu_prox_ko_biology__66a24b3a__s000__tb65536__v000",
"uid": "mmlu_prox_ko_biology__66a24b3a__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_ko_business__2546dfdb__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4710.471900157497, \"nll_base\": 8.45754337310791, \"eval_ref\": \"mmlu_prox_ko_business__2546dfdb__s000__tb131072__v000\", \"oracle_delta\":... | {
"index": "mmlu_prox_ko_business__2546dfdb__s000__tb131072__v000",
"uid": "mmlu_prox_ko_business__2546dfdb__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_ko_engineering__ea69b0d7__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3306.7154065878776, \"nll_base\": 8.103710651397705, \"eval_ref\": \"mmlu_prox_ko_engineering__ea69b0d7__s000__tb131072__v000\", \"oracle_... | {
"index": "mmlu_prox_ko_engineering__ea69b0d7__s000__tb131072__v000",
"uid": "mmlu_prox_ko_engineering__ea69b0d7__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_ko_philosophy__5d1b0bef__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4990.791857131405, \"nll_base\": 8.515349864959717, \"eval_ref\": \"mmlu_prox_ko_philosophy__5d1b0bef__s000__tb65536__v000\", \"oracle_delta\... | {
"index": "mmlu_prox_ko_philosophy__5d1b0bef__s000__tb65536__v000",
"uid": "mmlu_prox_ko_philosophy__5d1b0bef__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_af_biology__d99152c8__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2074.965222231578, \"nll_base\": 7.637699672154018, \"eval_ref\": \"mmlu_prox_lite_af_biology__d99152c8__s000__tb131072__v000\", \"oracle... | {
"index": "mmlu_prox_lite_af_biology__d99152c8__s000__tb131072__v000",
"uid": "mmlu_prox_lite_af_biology__d99152c8__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_af_business__4577e3e1__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2633.1227529167572, \"nll_base\": 7.875925779342651, \"eval_ref\": \"mmlu_prox_lite_af_business__4577e3e1__s000__tb65536__v000\", \"oracle... | {
"index": "mmlu_prox_lite_af_business__4577e3e1__s000__tb65536__v000",
"uid": "mmlu_prox_lite_af_business__4577e3e1__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_af_chemistry__8341100c__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3300.743290808781, \"nll_base\": 8.101902961730957, \"eval_ref\": \"mmlu_prox_lite_af_chemistry__8341100c__s000__tb262144__v000\", \"or... | {
"index": "mmlu_prox_lite_af_chemistry__8341100c__s000__tb262144__v000",
"uid": "mmlu_prox_lite_af_chemistry__8341100c__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_af_history__ca9ffd7a__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1894.6965349861032, \"nll_base\": 7.54681396484375, \"eval_ref\": \"mmlu_prox_lite_af_history__ca9ffd7a__s000__tb131072__v000\", \"oracle... | {
"index": "mmlu_prox_lite_af_history__ca9ffd7a__s000__tb131072__v000",
"uid": "mmlu_prox_lite_af_history__ca9ffd7a__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_af_math__e89fb68b__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3535.149146774838, \"nll_base\": 8.17051076889038, \"eval_ref\": \"mmlu_prox_lite_af_math__e89fb68b__s000__tb65536__v000\", \"oracle_delta\": ... | {
"index": "mmlu_prox_lite_af_math__e89fb68b__s000__tb65536__v000",
"uid": "mmlu_prox_lite_af_math__e89fb68b__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_af_other__bc2d35b4__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2778.5623441031084, \"nll_base\": 7.929688930511475, \"eval_ref\": \"mmlu_prox_lite_af_other__bc2d35b4__s000__tb262144__v000\", \"oracle_de... | {
"index": "mmlu_prox_lite_af_other__bc2d35b4__s000__tb262144__v000",
"uid": "mmlu_prox_lite_af_other__bc2d35b4__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_af_psychology__04a30e23__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2714.1581901052054, \"nll_base\": 7.9062371253967285, \"eval_ref\": \"mmlu_prox_lite_af_psychology__04a30e23__s000__tb131072__v000\", ... | {
"index": "mmlu_prox_lite_af_psychology__04a30e23__s000__tb131072__v000",
"uid": "mmlu_prox_lite_af_psychology__04a30e23__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_ar_history__4a4c84a6__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2434.712211192909, \"nll_base\": 7.797583840110085, \"eval_ref\": \"mmlu_prox_lite_ar_history__4a4c84a6__s000__tb262144__v000\", \"oracle... | {
"index": "mmlu_prox_lite_ar_history__4a4c84a6__s000__tb262144__v000",
"uid": "mmlu_prox_lite_ar_history__4a4c84a6__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_ar_math__5dc21a20__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 5570.820039361154, \"nll_base\": 8.625297546386719, \"eval_ref\": \"mmlu_prox_lite_ar_math__5dc21a20__s000__tb65536__v000\", \"oracle_delta\":... | {
"index": "mmlu_prox_lite_ar_math__5dc21a20__s000__tb65536__v000",
"uid": "mmlu_prox_lite_ar_math__5dc21a20__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_ar_other__4c2a3c31__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3810.4514379821157, \"nll_base\": 8.245502948760986, \"eval_ref\": \"mmlu_prox_lite_ar_other__4c2a3c31__s000__tb65536__v000\", \"oracle_delta... | {
"index": "mmlu_prox_lite_ar_other__4c2a3c31__s000__tb65536__v000",
"uid": "mmlu_prox_lite_ar_other__4c2a3c31__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_ar_philosophy__30a2bf4d__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4128.5135695167, \"nll_base\": 8.325672710643095, \"eval_ref\": \"mmlu_prox_lite_ar_philosophy__30a2bf4d__s000__tb131072__v000\", \"or... | {
"index": "mmlu_prox_lite_ar_philosophy__30a2bf4d__s000__tb131072__v000",
"uid": "mmlu_prox_lite_ar_philosophy__30a2bf4d__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_ar_physics__0c111c55__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 5509.86019552001, \"nll_base\": 8.614294528961182, \"eval_ref\": \"mmlu_prox_lite_ar_physics__0c111c55__s000__tb65536__v000\", \"oracle_del... | {
"index": "mmlu_prox_lite_ar_physics__0c111c55__s000__tb65536__v000",
"uid": "mmlu_prox_lite_ar_physics__0c111c55__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_cs_business__da58637b__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 5724.590964560314, \"nll_base\": 8.652526378631592, \"eval_ref\": \"mmlu_prox_lite_cs_business__da58637b__s000__tb131072__v000\", \"orac... | {
"index": "mmlu_prox_lite_cs_business__da58637b__s000__tb131072__v000",
"uid": "mmlu_prox_lite_cs_business__da58637b__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_cs_chemistry__9a9c546c__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3806.8573606348837, \"nll_base\": 8.244559288024902, \"eval_ref\": \"mmlu_prox_lite_cs_chemistry__9a9c546c__s000__tb65536__v000\", \"orac... | {
"index": "mmlu_prox_lite_cs_chemistry__9a9c546c__s000__tb65536__v000",
"uid": "mmlu_prox_lite_cs_chemistry__9a9c546c__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_cs_computer_science__2e5119d8__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2561.048105543896, \"nll_base\": 7.848171869913737, \"eval_ref\": \"mmlu_prox_lite_cs_computer_science__2e5119d8__s000__tb65536__v... | {
"index": "mmlu_prox_lite_cs_computer_science__2e5119d8__s000__tb65536__v000",
"uid": "mmlu_prox_lite_cs_computer_science__2e5119d8__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_cs_economics__41a346eb__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 5207.919437202815, \"nll_base\": 8.55793571472168, \"eval_ref\": \"mmlu_prox_lite_cs_economics__41a346eb__s000__tb131072__v000\", \"ora... | {
"index": "mmlu_prox_lite_cs_economics__41a346eb__s000__tb131072__v000",
"uid": "mmlu_prox_lite_cs_economics__41a346eb__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_cs_health__0882208b__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 4408.100643495297, \"nll_base\": 8.391199182581019, \"eval_ref\": \"mmlu_prox_lite_cs_health__0882208b__s000__tb262144__v000\", \"oracle_d... | {
"index": "mmlu_prox_lite_cs_health__0882208b__s000__tb262144__v000",
"uid": "mmlu_prox_lite_cs_health__0882208b__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_cs_law__8758bf10__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3637.3211532054634, \"nll_base\": 8.199002742767334, \"eval_ref\": \"mmlu_prox_lite_cs_law__8758bf10__s000__tb131072__v000\", \"oracle_delta\... | {
"index": "mmlu_prox_lite_cs_law__8758bf10__s000__tb131072__v000",
"uid": "mmlu_prox_lite_cs_law__8758bf10__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_cs_math__20ecb9b8__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 6250.948737468906, \"nll_base\": 8.740488529205322, \"eval_ref\": \"mmlu_prox_lite_cs_math__20ecb9b8__s000__tb131072__v000\", \"oracle_delta... | {
"index": "mmlu_prox_lite_cs_math__20ecb9b8__s000__tb131072__v000",
"uid": "mmlu_prox_lite_cs_math__20ecb9b8__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_cs_other__06a2332c__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4099.270930201051, \"nll_base\": 8.318564414978027, \"eval_ref\": \"mmlu_prox_lite_cs_other__06a2332c__s000__tb65536__v000\", \"oracle_delta\... | {
"index": "mmlu_prox_lite_cs_other__06a2332c__s000__tb65536__v000",
"uid": "mmlu_prox_lite_cs_other__06a2332c__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_cs_philosophy__3b6cd0be__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 4647.298771492317, \"nll_base\": 8.444041420431699, \"eval_ref\": \"mmlu_prox_lite_cs_philosophy__3b6cd0be__s000__tb262144__v000\", \"... | {
"index": "mmlu_prox_lite_cs_philosophy__3b6cd0be__s000__tb262144__v000",
"uid": "mmlu_prox_lite_cs_philosophy__3b6cd0be__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_cs_physics__2c2c0889__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 5549.323740223782, \"nll_base\": 8.621431350708008, \"eval_ref\": \"mmlu_prox_lite_cs_physics__2c2c0889__s000__tb131072__v000\", \"oracle... | {
"index": "mmlu_prox_lite_cs_physics__2c2c0889__s000__tb131072__v000",
"uid": "mmlu_prox_lite_cs_physics__2c2c0889__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_de_biology__13d0c013__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2123.3996377921703, \"nll_base\": 7.660773686000279, \"eval_ref\": \"mmlu_prox_lite_de_biology__13d0c013__s000__tb65536__v000\", \"oracle_d... | {
"index": "mmlu_prox_lite_de_biology__13d0c013__s000__tb65536__v000",
"uid": "mmlu_prox_lite_de_biology__13d0c013__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_de_business__7bf52521__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3165.5646628295694, \"nll_base\": 8.060086727142334, \"eval_ref\": \"mmlu_prox_lite_de_business__7bf52521__s000__tb65536__v000\", \"oracle... | {
"index": "mmlu_prox_lite_de_business__7bf52521__s000__tb65536__v000",
"uid": "mmlu_prox_lite_de_business__7bf52521__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_de_chemistry__1ee52cdd__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3823.381954575373, \"nll_base\": 8.24889063835144, \"eval_ref\": \"mmlu_prox_lite_de_chemistry__1ee52cdd__s000__tb131072__v000\", \"ora... | {
"index": "mmlu_prox_lite_de_chemistry__1ee52cdd__s000__tb131072__v000",
"uid": "mmlu_prox_lite_de_chemistry__1ee52cdd__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_de_law__b2f7bbc3__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3642.9988346163354, \"nll_base\": 8.200562477111816, \"eval_ref\": \"mmlu_prox_lite_de_law__b2f7bbc3__s000__tb65536__v000\", \"oracle_delta\": ... | {
"index": "mmlu_prox_lite_de_law__b2f7bbc3__s000__tb65536__v000",
"uid": "mmlu_prox_lite_de_law__b2f7bbc3__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_de_other__51e3ca87__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3483.886696068048, \"nll_base\": 8.155903816223145, \"eval_ref\": \"mmlu_prox_lite_de_other__51e3ca87__s000__tb131072__v000\", \"oracle_del... | {
"index": "mmlu_prox_lite_de_other__51e3ca87__s000__tb131072__v000",
"uid": "mmlu_prox_lite_de_other__51e3ca87__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_de_philosophy__557edf86__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3781.4949666479147, \"nll_base\": 8.237874704248766, \"eval_ref\": \"mmlu_prox_lite_de_philosophy__557edf86__s000__tb262144__v000\", \... | {
"index": "mmlu_prox_lite_de_philosophy__557edf86__s000__tb262144__v000",
"uid": "mmlu_prox_lite_de_philosophy__557edf86__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_de_physics__f78fbef4__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3079.1972580738143, \"nll_base\": 8.032424211502075, \"eval_ref\": \"mmlu_prox_lite_de_physics__f78fbef4__s000__tb65536__v000\", \"oracle_d... | {
"index": "mmlu_prox_lite_de_physics__f78fbef4__s000__tb65536__v000",
"uid": "mmlu_prox_lite_de_physics__f78fbef4__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_en_history__0eb74199__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1871.0595299162837, \"nll_base\": 7.534260142933238, \"eval_ref\": \"mmlu_prox_lite_en_history__0eb74199__s000__tb65536__v000\", \"oracle_d... | {
"index": "mmlu_prox_lite_en_history__0eb74199__s000__tb65536__v000",
"uid": "mmlu_prox_lite_en_history__0eb74199__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_en_psychology__f9d5d59e__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3413.149525466373, \"nll_base\": 8.135390758514404, \"eval_ref\": \"mmlu_prox_lite_en_psychology__f9d5d59e__s000__tb131072__v000\", \"... | {
"index": "mmlu_prox_lite_en_psychology__f9d5d59e__s000__tb131072__v000",
"uid": "mmlu_prox_lite_en_psychology__f9d5d59e__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_es_biology__a9f2da52__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1985.6714501901286, \"nll_base\": 7.593712397984096, \"eval_ref\": \"mmlu_prox_lite_es_biology__a9f2da52__s000__tb131072__v000\", \"oracl... | {
"index": "mmlu_prox_lite_es_biology__a9f2da52__s000__tb131072__v000",
"uid": "mmlu_prox_lite_es_biology__a9f2da52__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_es_business__e097c693__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3744.663161549866, \"nll_base\": 8.228086948394775, \"eval_ref\": \"mmlu_prox_lite_es_business__e097c693__s000__tb131072__v000\", \"orac... | {
"index": "mmlu_prox_lite_es_business__e097c693__s000__tb131072__v000",
"uid": "mmlu_prox_lite_es_business__e097c693__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_es_chemistry__1ef4eefe__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3220.581582311174, \"nll_base\": 8.077317237854004, \"eval_ref\": \"mmlu_prox_lite_es_chemistry__1ef4eefe__s000__tb65536__v000\", \"oracl... | {
"index": "mmlu_prox_lite_es_chemistry__1ef4eefe__s000__tb65536__v000",
"uid": "mmlu_prox_lite_es_chemistry__1ef4eefe__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_es_economics__1c6a551d__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 4373.689859433137, \"nll_base\": 8.383362293243408, \"eval_ref\": \"mmlu_prox_lite_es_economics__1c6a551d__s000__tb262144__v000\", \"or... | {
"index": "mmlu_prox_lite_es_economics__1c6a551d__s000__tb262144__v000",
"uid": "mmlu_prox_lite_es_economics__1c6a551d__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_es_health__7ee199a3__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3017.737621190221, \"nll_base\": 8.012262697573062, \"eval_ref\": \"mmlu_prox_lite_es_health__7ee199a3__s000__tb262144__v000\", \"oracle_d... | {
"index": "mmlu_prox_lite_es_health__7ee199a3__s000__tb262144__v000",
"uid": "mmlu_prox_lite_es_health__7ee199a3__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_es_history__9ff61d45__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3447.47321241259, \"nll_base\": 8.145396839488637, \"eval_ref\": \"mmlu_prox_lite_es_history__9ff61d45__s000__tb131072__v000\", \"oracle_... | {
"index": "mmlu_prox_lite_es_history__9ff61d45__s000__tb131072__v000",
"uid": "mmlu_prox_lite_es_history__9ff61d45__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_es_math__a9915d63__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4482.7196700233335, \"nll_base\": 8.407985210418701, \"eval_ref\": \"mmlu_prox_lite_es_math__a9915d63__s000__tb65536__v000\", \"oracle_delta\"... | {
"index": "mmlu_prox_lite_es_math__a9915d63__s000__tb65536__v000",
"uid": "mmlu_prox_lite_es_math__a9915d63__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_es_other__82eab541__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3179.0228457369067, \"nll_base\": 8.064329147338867, \"eval_ref\": \"mmlu_prox_lite_es_other__82eab541__s000__tb131072__v000\", \"oracle_de... | {
"index": "mmlu_prox_lite_es_other__82eab541__s000__tb131072__v000",
"uid": "mmlu_prox_lite_es_other__82eab541__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_es_philosophy__0a8880b3__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2975.237955808365, \"nll_base\": 7.998079299926758, \"eval_ref\": \"mmlu_prox_lite_es_philosophy__0a8880b3__s000__tb262144__v000\", \"... | {
"index": "mmlu_prox_lite_es_philosophy__0a8880b3__s000__tb262144__v000",
"uid": "mmlu_prox_lite_es_philosophy__0a8880b3__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_es_psychology__dfa97157__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3979.7424207399117, \"nll_base\": 8.2889723777771, \"eval_ref\": \"mmlu_prox_lite_es_psychology__dfa97157__s000__tb131072__v000\", \"o... | {
"index": "mmlu_prox_lite_es_psychology__dfa97157__s000__tb131072__v000",
"uid": "mmlu_prox_lite_es_psychology__dfa97157__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_fr_biology__e6ac2109__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 4805.840803291534, \"nll_base\": 8.477587291172572, \"eval_ref\": \"mmlu_prox_lite_fr_biology__e6ac2109__s000__tb262144__v000\", \"oracle... | {
"index": "mmlu_prox_lite_fr_biology__e6ac2109__s000__tb262144__v000",
"uid": "mmlu_prox_lite_fr_biology__e6ac2109__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_fr_business__8651c527__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 5597.866272806893, \"nll_base\": 8.630140781402588, \"eval_ref\": \"mmlu_prox_lite_fr_business__8651c527__s000__tb262144__v000\", \"orac... | {
"index": "mmlu_prox_lite_fr_business__8651c527__s000__tb262144__v000",
"uid": "mmlu_prox_lite_fr_business__8651c527__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_fr_economics__bfaea250__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 7281.57979685604, \"nll_base\": 8.893103122711182, \"eval_ref\": \"mmlu_prox_lite_fr_economics__bfaea250__s000__tb65536__v000\", \"oracle... | {
"index": "mmlu_prox_lite_fr_economics__bfaea250__s000__tb65536__v000",
"uid": "mmlu_prox_lite_fr_economics__bfaea250__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_fr_engineering__074a9b16__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 5517.713550262554, \"nll_base\": 8.615718841552734, \"eval_ref\": \"mmlu_prox_lite_fr_engineering__074a9b16__s000__tb262144__v000\", ... | {
"index": "mmlu_prox_lite_fr_engineering__074a9b16__s000__tb262144__v000",
"uid": "mmlu_prox_lite_fr_engineering__074a9b16__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_fr_health__52a4be1c__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4274.00149698466, \"nll_base\": 8.360305786132812, \"eval_ref\": \"mmlu_prox_lite_fr_health__52a4be1c__s000__tb65536__v000\", \"oracle_delta... | {
"index": "mmlu_prox_lite_fr_health__52a4be1c__s000__tb65536__v000",
"uid": "mmlu_prox_lite_fr_health__52a4be1c__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_fr_history__bb03022e__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 6526.718058448272, \"nll_base\": 8.783659501509232, \"eval_ref\": \"mmlu_prox_lite_fr_history__bb03022e__s000__tb65536__v000\", \"oracle_de... | {
"index": "mmlu_prox_lite_fr_history__bb03022e__s000__tb65536__v000",
"uid": "mmlu_prox_lite_fr_history__bb03022e__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_fr_other__a7c0fd21__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 5412.642864948735, \"nll_base\": 8.596492767333984, \"eval_ref\": \"mmlu_prox_lite_fr_other__a7c0fd21__s000__tb65536__v000\", \"oracle_delta\... | {
"index": "mmlu_prox_lite_fr_other__a7c0fd21__s000__tb65536__v000",
"uid": "mmlu_prox_lite_fr_other__a7c0fd21__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_fr_physics__a3848908__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 5303.386332988596, \"nll_base\": 8.576100826263428, \"eval_ref\": \"mmlu_prox_lite_fr_physics__a3848908__s000__tb131072__v000\", \"oracle... | {
"index": "mmlu_prox_lite_fr_physics__a3848908__s000__tb131072__v000",
"uid": "mmlu_prox_lite_fr_physics__a3848908__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_fr_psychology__beb00773__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 8150.725625073177, \"nll_base\": 9.00586223602295, \"eval_ref\": \"mmlu_prox_lite_fr_psychology__beb00773__s000__tb262144__v000\", \"o... | {
"index": "mmlu_prox_lite_fr_psychology__beb00773__s000__tb262144__v000",
"uid": "mmlu_prox_lite_fr_psychology__beb00773__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_hi_chemistry__69491460__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 5150.989655887333, \"nll_base\": 8.54694414138794, \"eval_ref\": \"mmlu_prox_lite_hi_chemistry__69491460__s000__tb262144__v000\", \"ora... | {
"index": "mmlu_prox_lite_hi_chemistry__69491460__s000__tb262144__v000",
"uid": "mmlu_prox_lite_hi_chemistry__69491460__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_hi_computer_science__6a5d449c__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 8936.805264473787, \"nll_base\": 9.097933451334635, \"eval_ref\": \"mmlu_prox_lite_hi_computer_science__6a5d449c__s000__tb131072... | {
"index": "mmlu_prox_lite_hi_computer_science__6a5d449c__s000__tb131072__v000",
"uid": "mmlu_prox_lite_hi_computer_science__6a5d449c__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_hi_economics__8dd2511f__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4805.038481684739, \"nll_base\": 8.477420330047607, \"eval_ref\": \"mmlu_prox_lite_hi_economics__8dd2511f__s000__tb131072__v000\", \"or... | {
"index": "mmlu_prox_lite_hi_economics__8dd2511f__s000__tb131072__v000",
"uid": "mmlu_prox_lite_hi_economics__8dd2511f__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_hi_engineering__a2e94f4f__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3845.178299064795, \"nll_base\": 8.254575252532959, \"eval_ref\": \"mmlu_prox_lite_hi_engineering__a2e94f4f__s000__tb65536__v000\", \"o... | {
"index": "mmlu_prox_lite_hi_engineering__a2e94f4f__s000__tb65536__v000",
"uid": "mmlu_prox_lite_hi_engineering__a2e94f4f__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_hi_health__c217ff2d__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4052.6355020462884, \"nll_base\": 8.307122689706308, \"eval_ref\": \"mmlu_prox_lite_hi_health__c217ff2d__s000__tb131072__v000\", \"oracle_... | {
"index": "mmlu_prox_lite_hi_health__c217ff2d__s000__tb131072__v000",
"uid": "mmlu_prox_lite_hi_health__c217ff2d__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_hi_other__055a208f__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 6767.778473208788, \"nll_base\": 8.819928169250488, \"eval_ref\": \"mmlu_prox_lite_hi_other__055a208f__s000__tb65536__v000\", \"oracle_delta\... | {
"index": "mmlu_prox_lite_hi_other__055a208f__s000__tb65536__v000",
"uid": "mmlu_prox_lite_hi_other__055a208f__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_hi_philosophy__0b510e15__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 5015.568446584642, \"nll_base\": 8.520302043241614, \"eval_ref\": \"mmlu_prox_lite_hi_philosophy__0b510e15__s000__tb131072__v000\", \"... | {
"index": "mmlu_prox_lite_hi_philosophy__0b510e15__s000__tb131072__v000",
"uid": "mmlu_prox_lite_hi_philosophy__0b510e15__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_hi_psychology__f4521287__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3727.3917754988543, \"nll_base\": 8.223464012145996, \"eval_ref\": \"mmlu_prox_lite_hi_psychology__f4521287__s000__tb65536__v000\", \"or... | {
"index": "mmlu_prox_lite_hi_psychology__f4521287__s000__tb65536__v000",
"uid": "mmlu_prox_lite_hi_psychology__f4521287__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_hu_business__b2dc59e0__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4934.084694682601, \"nll_base\": 8.503922462463379, \"eval_ref\": \"mmlu_prox_lite_hu_business__b2dc59e0__s000__tb131072__v000\", \"orac... | {
"index": "mmlu_prox_lite_hu_business__b2dc59e0__s000__tb131072__v000",
"uid": "mmlu_prox_lite_hu_business__b2dc59e0__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_hu_computer_science__19121fcb__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4231.008106997601, \"nll_base\": 8.35019556681315, \"eval_ref\": \"mmlu_prox_lite_hu_computer_science__19121fcb__s000__tb65536__v0... | {
"index": "mmlu_prox_lite_hu_computer_science__19121fcb__s000__tb65536__v000",
"uid": "mmlu_prox_lite_hu_computer_science__19121fcb__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_hu_economics__494c950d__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 5235.557472756066, \"nll_base\": 8.563228607177734, \"eval_ref\": \"mmlu_prox_lite_hu_economics__494c950d__s000__tb262144__v000\", \"or... | {
"index": "mmlu_prox_lite_hu_economics__494c950d__s000__tb262144__v000",
"uid": "mmlu_prox_lite_hu_economics__494c950d__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_hu_engineering__1879117c__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3576.76664148849, \"nll_base\": 8.182214498519897, \"eval_ref\": \"mmlu_prox_lite_hu_engineering__1879117c__s000__tb65536__v000\", \"or... | {
"index": "mmlu_prox_lite_hu_engineering__1879117c__s000__tb65536__v000",
"uid": "mmlu_prox_lite_hu_engineering__1879117c__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_hu_math__626aacd2__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 6797.662925014763, \"nll_base\": 8.824334144592285, \"eval_ref\": \"mmlu_prox_lite_hu_math__626aacd2__s000__tb262144__v000\", \"oracle_delta... | {
"index": "mmlu_prox_lite_hu_math__626aacd2__s000__tb262144__v000",
"uid": "mmlu_prox_lite_hu_math__626aacd2__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_hu_philosophy__ada3e9c1__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3511.186930407273, \"nll_base\": 8.163709416108972, \"eval_ref\": \"mmlu_prox_lite_hu_philosophy__ada3e9c1__s000__tb262144__v000\", \"... | {
"index": "mmlu_prox_lite_hu_philosophy__ada3e9c1__s000__tb262144__v000",
"uid": "mmlu_prox_lite_hu_philosophy__ada3e9c1__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_hu_physics__c25dde21__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 4722.019899824045, \"nll_base\": 8.459991931915283, \"eval_ref\": \"mmlu_prox_lite_hu_physics__c25dde21__s000__tb262144__v000\", \"oracle... | {
"index": "mmlu_prox_lite_hu_physics__c25dde21__s000__tb262144__v000",
"uid": "mmlu_prox_lite_hu_physics__c25dde21__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_id_business__373c27b3__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 9816.062190149198, \"nll_base\": 9.19177532196045, \"eval_ref\": \"mmlu_prox_lite_id_business__373c27b3__s000__tb131072__v000\", \"oracl... | {
"index": "mmlu_prox_lite_id_business__373c27b3__s000__tb131072__v000",
"uid": "mmlu_prox_lite_id_business__373c27b3__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_id_chemistry__87a83562__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4660.7463969590635, \"nll_base\": 8.446930885314941, \"eval_ref\": \"mmlu_prox_lite_id_chemistry__87a83562__s000__tb65536__v000\", \"orac... | {
"index": "mmlu_prox_lite_id_chemistry__87a83562__s000__tb65536__v000",
"uid": "mmlu_prox_lite_id_chemistry__87a83562__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_id_computer_science__c38b240c__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 13104.93353460298, \"nll_base\": 9.480744043986002, \"eval_ref\": \"mmlu_prox_lite_id_computer_science__c38b240c__s000__tb131072... | {
"index": "mmlu_prox_lite_id_computer_science__c38b240c__s000__tb131072__v000",
"uid": "mmlu_prox_lite_id_computer_science__c38b240c__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_id_economics__026ed131__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 9088.140399699514, \"nll_base\": 9.114725589752197, \"eval_ref\": \"mmlu_prox_lite_id_economics__026ed131__s000__tb131072__v000\", \"or... | {
"index": "mmlu_prox_lite_id_economics__026ed131__s000__tb131072__v000",
"uid": "mmlu_prox_lite_id_economics__026ed131__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_id_history__4e162246__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 7851.131334822636, \"nll_base\": 8.968412919477982, \"eval_ref\": \"mmlu_prox_lite_id_history__4e162246__s000__tb65536__v000\", \"oracle_de... | {
"index": "mmlu_prox_lite_id_history__4e162246__s000__tb65536__v000",
"uid": "mmlu_prox_lite_id_history__4e162246__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_id_law__f48120ee__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 5591.772977346865, \"nll_base\": 8.629051685333252, \"eval_ref\": \"mmlu_prox_lite_id_law__f48120ee__s000__tb131072__v000\", \"oracle_delta\"... | {
"index": "mmlu_prox_lite_id_law__f48120ee__s000__tb131072__v000",
"uid": "mmlu_prox_lite_id_law__f48120ee__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_id_math__e8acfb64__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 8739.145738835452, \"nll_base\": 9.075567722320557, \"eval_ref\": \"mmlu_prox_lite_id_math__e8acfb64__s000__tb65536__v000\", \"oracle_delta\":... | {
"index": "mmlu_prox_lite_id_math__e8acfb64__s000__tb65536__v000",
"uid": "mmlu_prox_lite_id_math__e8acfb64__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_it_biology__513312d0__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2066.264177627812, \"nll_base\": 7.6334975106375555, \"eval_ref\": \"mmlu_prox_lite_it_biology__513312d0__s000__tb131072__v000\", \"oracl... | {
"index": "mmlu_prox_lite_it_biology__513312d0__s000__tb131072__v000",
"uid": "mmlu_prox_lite_it_biology__513312d0__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_business__d8d5964b__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3381.1005275214648, \"nll_base\": 8.125956535339355, \"eval_ref\": \"mmlu_prox_lite_it_business__d8d5964b__s000__tb131072__v000\", \"ora... | {
"index": "mmlu_prox_lite_it_business__d8d5964b__s000__tb131072__v000",
"uid": "mmlu_prox_lite_it_business__d8d5964b__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_chemistry__e4adfed2__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2928.7222090948044, \"nll_base\": 7.982321500778198, \"eval_ref\": \"mmlu_prox_lite_it_chemistry__e4adfed2__s000__tb131072__v000\", \"o... | {
"index": "mmlu_prox_lite_it_chemistry__e4adfed2__s000__tb131072__v000",
"uid": "mmlu_prox_lite_it_chemistry__e4adfed2__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_computer_science__091e701e__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4523.6948157405, \"nll_base\": 8.417084376017252, \"eval_ref\": \"mmlu_prox_lite_it_computer_science__091e701e__s000__tb65536__v00... | {
"index": "mmlu_prox_lite_it_computer_science__091e701e__s000__tb65536__v000",
"uid": "mmlu_prox_lite_it_computer_science__091e701e__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_it_economics__b93d50eb__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3902.3392313991208, \"nll_base\": 8.269331455230713, \"eval_ref\": \"mmlu_prox_lite_it_economics__b93d50eb__s000__tb65536__v000\", \"orac... | {
"index": "mmlu_prox_lite_it_economics__b93d50eb__s000__tb65536__v000",
"uid": "mmlu_prox_lite_it_economics__b93d50eb__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_it_engineering__44834c05__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2246.2242356515058, \"nll_base\": 7.717005968093872, \"eval_ref\": \"mmlu_prox_lite_it_engineering__44834c05__s000__tb131072__v000\",... | {
"index": "mmlu_prox_lite_it_engineering__44834c05__s000__tb131072__v000",
"uid": "mmlu_prox_lite_it_engineering__44834c05__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_health__7ba0ce20__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2854.614725785942, \"nll_base\": 7.956692165798611, \"eval_ref\": \"mmlu_prox_lite_it_health__7ba0ce20__s000__tb262144__v000\", \"oracle_d... | {
"index": "mmlu_prox_lite_it_health__7ba0ce20__s000__tb262144__v000",
"uid": "mmlu_prox_lite_it_health__7ba0ce20__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_it_history__f47e00b9__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1804.3255496215027, \"nll_base\": 7.497942144220525, \"eval_ref\": \"mmlu_prox_lite_it_history__f47e00b9__s000__tb65536__v000\", \"oracle_d... | {
"index": "mmlu_prox_lite_it_history__f47e00b9__s000__tb65536__v000",
"uid": "mmlu_prox_lite_it_history__f47e00b9__s000__tb65536__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.