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_pt_history__846a4fdc__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2954.146009297444, \"nll_base\": 7.990964889526367, \"eval_ref\": \"mmlu_prox_pt_history__846a4fdc__s000__tb65536__v000\", \"oracle_delta\": 0.3... | {
"index": "mmlu_prox_pt_history__846a4fdc__s000__tb65536__v000",
"uid": "mmlu_prox_pt_history__846a4fdc__s000__tb65536__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_prox_pt_physics__4050ff12__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 5868.155771364185, \"nll_base\": 8.677295684814453, \"eval_ref\": \"mmlu_prox_pt_physics__4050ff12__s000__tb131072__v000\", \"oracle_delta\": ... | {
"index": "mmlu_prox_pt_physics__4050ff12__s000__tb131072__v000",
"uid": "mmlu_prox_pt_physics__4050ff12__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_ru_business__5f0d0c8a__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 4871.4436562836945, \"nll_base\": 8.491145610809326, \"eval_ref\": \"mmlu_prox_ru_business__5f0d0c8a__s000__tb262144__v000\", \"oracle_delta\... | {
"index": "mmlu_prox_ru_business__5f0d0c8a__s000__tb262144__v000",
"uid": "mmlu_prox_ru_business__5f0d0c8a__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_ru_economics__a9423af2__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2974.9641585329987, \"nll_base\": 7.997987270355225, \"eval_ref\": \"mmlu_prox_ru_economics__a9423af2__s000__tb131072__v000\", \"oracle_delt... | {
"index": "mmlu_prox_ru_economics__a9423af2__s000__tb131072__v000",
"uid": "mmlu_prox_ru_economics__a9423af2__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_ru_engineering__9a598fb1__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3064.721158726685, \"nll_base\": 8.027711868286133, \"eval_ref\": \"mmlu_prox_ru_engineering__9a598fb1__s000__tb65536__v000\", \"oracle_delt... | {
"index": "mmlu_prox_ru_engineering__9a598fb1__s000__tb65536__v000",
"uid": "mmlu_prox_ru_engineering__9a598fb1__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_ru_health__ee4d61b0__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4413.101594745029, \"nll_base\": 8.392333030700684, \"eval_ref\": \"mmlu_prox_ru_health__ee4d61b0__s000__tb131072__v000\", \"oracle_delta\": 0.... | {
"index": "mmlu_prox_ru_health__ee4d61b0__s000__tb131072__v000",
"uid": "mmlu_prox_ru_health__ee4d61b0__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_ru_history__3d651dca__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1692.5352794041048, \"nll_base\": 7.433982849121094, \"eval_ref\": \"mmlu_prox_ru_history__3d651dca__s000__tb131072__v000\", \"oracle_delta\":... | {
"index": "mmlu_prox_ru_history__3d651dca__s000__tb131072__v000",
"uid": "mmlu_prox_ru_history__3d651dca__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_ru_law__57dbf05b__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2066.0143560198044, \"nll_base\": 7.633376598358154, \"eval_ref\": \"mmlu_prox_ru_law__57dbf05b__s000__tb131072__v000\", \"oracle_delta\": 0.37677... | {
"index": "mmlu_prox_ru_law__57dbf05b__s000__tb131072__v000",
"uid": "mmlu_prox_ru_law__57dbf05b__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_ru_other__6f29e8d0__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3779.531095988882, \"nll_base\": 8.23735523223877, \"eval_ref\": \"mmlu_prox_ru_other__6f29e8d0__s000__tb131072__v000\", \"oracle_delta\": 0.393... | {
"index": "mmlu_prox_ru_other__6f29e8d0__s000__tb131072__v000",
"uid": "mmlu_prox_ru_other__6f29e8d0__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_sr_biology__ccabba1c__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2381.950145335676, \"nll_base\": 7.775674819946289, \"eval_ref\": \"mmlu_prox_sr_biology__ccabba1c__s000__tb131072__v000\", \"oracle_delta\": ... | {
"index": "mmlu_prox_sr_biology__ccabba1c__s000__tb131072__v000",
"uid": "mmlu_prox_sr_biology__ccabba1c__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_sr_business__975a6c97__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3297.0521598345454, \"nll_base\": 8.100784063339233, \"eval_ref\": \"mmlu_prox_sr_business__975a6c97__s000__tb65536__v000\", \"oracle_delta\": ... | {
"index": "mmlu_prox_sr_business__975a6c97__s000__tb65536__v000",
"uid": "mmlu_prox_sr_business__975a6c97__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_sr_computer_science__bab476a6__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3650.115798173727, \"nll_base\": 8.202514171600342, \"eval_ref\": \"mmlu_prox_sr_computer_science__bab476a6__s000__tb131072__v000\", ... | {
"index": "mmlu_prox_sr_computer_science__bab476a6__s000__tb131072__v000",
"uid": "mmlu_prox_sr_computer_science__bab476a6__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_sr_health__2c5bbbe9__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3407.483965312287, \"nll_base\": 8.133729457855225, \"eval_ref\": \"mmlu_prox_sr_health__2c5bbbe9__s000__tb262144__v000\", \"oracle_delta\": 1.... | {
"index": "mmlu_prox_sr_health__2c5bbbe9__s000__tb262144__v000",
"uid": "mmlu_prox_sr_health__2c5bbbe9__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_sr_law__f694250a__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2512.0048065958963, \"nll_base\": 7.828836441040039, \"eval_ref\": \"mmlu_prox_sr_law__f694250a__s000__tb131072__v000\", \"oracle_delta\": 0.06367... | {
"index": "mmlu_prox_sr_law__f694250a__s000__tb131072__v000",
"uid": "mmlu_prox_sr_law__f694250a__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_sr_math__3d258591__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3431.1221461735777, \"nll_base\": 8.140642642974854, \"eval_ref\": \"mmlu_prox_sr_math__3d258591__s000__tb131072__v000\", \"oracle_delta\": 0.196... | {
"index": "mmlu_prox_sr_math__3d258591__s000__tb131072__v000",
"uid": "mmlu_prox_sr_math__3d258591__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_sr_other__17600ca4__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 5785.972862641893, \"nll_base\": 8.663191795349121, \"eval_ref\": \"mmlu_prox_sr_other__17600ca4__s000__tb131072__v000\", \"oracle_delta\": 0.21... | {
"index": "mmlu_prox_sr_other__17600ca4__s000__tb131072__v000",
"uid": "mmlu_prox_sr_other__17600ca4__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_sr_philosophy__9aeaa499__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4859.423641085529, \"nll_base\": 8.488675117492676, \"eval_ref\": \"mmlu_prox_sr_philosophy__9aeaa499__s000__tb131072__v000\", \"oracle_del... | {
"index": "mmlu_prox_sr_philosophy__9aeaa499__s000__tb131072__v000",
"uid": "mmlu_prox_sr_philosophy__9aeaa499__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_sw_biology__eed73102__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3445.078458201464, \"nll_base\": 8.144701957702637, \"eval_ref\": \"mmlu_prox_sw_biology__eed73102__s000__tb262144__v000\", \"oracle_delta\": ... | {
"index": "mmlu_prox_sw_biology__eed73102__s000__tb262144__v000",
"uid": "mmlu_prox_sw_biology__eed73102__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_sw_chemistry__888153f5__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4720.763654880322, \"nll_base\": 8.459725856781006, \"eval_ref\": \"mmlu_prox_sw_chemistry__888153f5__s000__tb65536__v000\", \"oracle_delta\":... | {
"index": "mmlu_prox_sw_chemistry__888153f5__s000__tb65536__v000",
"uid": "mmlu_prox_sw_chemistry__888153f5__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_sw_computer_science__be4d5ce1__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3640.564210180795, \"nll_base\": 8.199893951416016, \"eval_ref\": \"mmlu_prox_sw_computer_science__be4d5ce1__s000__tb262144__v000\", ... | {
"index": "mmlu_prox_sw_computer_science__be4d5ce1__s000__tb262144__v000",
"uid": "mmlu_prox_sw_computer_science__be4d5ce1__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_sw_engineering__8979dc06__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2670.93321852424, \"nll_base\": 7.890183210372925, \"eval_ref\": \"mmlu_prox_sw_engineering__8979dc06__s000__tb262144__v000\", \"oracle_de... | {
"index": "mmlu_prox_sw_engineering__8979dc06__s000__tb262144__v000",
"uid": "mmlu_prox_sw_engineering__8979dc06__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_sw_history__e01af006__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3285.6259161481075, \"nll_base\": 8.097312450408936, \"eval_ref\": \"mmlu_prox_sw_history__e01af006__s000__tb262144__v000\", \"oracle_delta\":... | {
"index": "mmlu_prox_sw_history__e01af006__s000__tb262144__v000",
"uid": "mmlu_prox_sw_history__e01af006__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_sw_law__399ceb9d__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2417.2809128454696, \"nll_base\": 7.790398597717285, \"eval_ref\": \"mmlu_prox_sw_law__399ceb9d__s000__tb262144__v000\", \"oracle_delta\": 1.48101... | {
"index": "mmlu_prox_sw_law__399ceb9d__s000__tb262144__v000",
"uid": "mmlu_prox_sw_law__399ceb9d__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_sw_math__d85309c9__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4283.13129652254, \"nll_base\": 8.362439632415771, \"eval_ref\": \"mmlu_prox_sw_math__d85309c9__s000__tb131072__v000\", \"oracle_delta\": 0.82609... | {
"index": "mmlu_prox_sw_math__d85309c9__s000__tb131072__v000",
"uid": "mmlu_prox_sw_math__d85309c9__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_sw_other__bc11ced3__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 4513.556480053585, \"nll_base\": 8.414840698242188, \"eval_ref\": \"mmlu_prox_sw_other__bc11ced3__s000__tb262144__v000\", \"oracle_delta\": 1.45... | {
"index": "mmlu_prox_sw_other__bc11ced3__s000__tb262144__v000",
"uid": "mmlu_prox_sw_other__bc11ced3__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_sw_psychology__6793bf9b__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3820.3503793555533, \"nll_base\": 8.24809741973877, \"eval_ref\": \"mmlu_prox_sw_psychology__6793bf9b__s000__tb131072__v000\", \"oracle_del... | {
"index": "mmlu_prox_sw_psychology__6793bf9b__s000__tb131072__v000",
"uid": "mmlu_prox_sw_psychology__6793bf9b__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_uk_engineering__32c9db2c__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 5646.881897015019, \"nll_base\": 8.638858795166016, \"eval_ref\": \"mmlu_prox_uk_engineering__32c9db2c__s000__tb131072__v000\", \"oracle_d... | {
"index": "mmlu_prox_uk_engineering__32c9db2c__s000__tb131072__v000",
"uid": "mmlu_prox_uk_engineering__32c9db2c__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_uk_math__fad2ab34__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 6510.26887321719, \"nll_base\": 8.78113603591919, \"eval_ref\": \"mmlu_prox_uk_math__fad2ab34__s000__tb131072__v000\", \"oracle_delta\": 0.471430... | {
"index": "mmlu_prox_uk_math__fad2ab34__s000__tb131072__v000",
"uid": "mmlu_prox_uk_math__fad2ab34__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_uk_other__8845e781__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 6247.638073858775, \"nll_base\": 8.739958763122559, \"eval_ref\": \"mmlu_prox_uk_other__8845e781__s000__tb131072__v000\", \"oracle_delta\": 0.07... | {
"index": "mmlu_prox_uk_other__8845e781__s000__tb131072__v000",
"uid": "mmlu_prox_uk_other__8845e781__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_uk_physics__f29dfa88__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 5798.760990530018, \"nll_base\": 8.665399551391602, \"eval_ref\": \"mmlu_prox_uk_physics__f29dfa88__s000__tb131072__v000\", \"oracle_delta\": ... | {
"index": "mmlu_prox_uk_physics__f29dfa88__s000__tb131072__v000",
"uid": "mmlu_prox_uk_physics__f29dfa88__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_ur_biology__3b3079d4__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1916.8203840807494, \"nll_base\": 7.558423042297363, \"eval_ref\": \"mmlu_prox_ur_biology__3b3079d4__s000__tb65536__v000\", \"oracle_delta\": 0.... | {
"index": "mmlu_prox_ur_biology__3b3079d4__s000__tb65536__v000",
"uid": "mmlu_prox_ur_biology__3b3079d4__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_ur_business__ab44116d__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 5702.11519641049, \"nll_base\": 8.648592472076416, \"eval_ref\": \"mmlu_prox_ur_business__ab44116d__s000__tb65536__v000\", \"oracle_delta\": 0.... | {
"index": "mmlu_prox_ur_business__ab44116d__s000__tb65536__v000",
"uid": "mmlu_prox_ur_business__ab44116d__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_ur_chemistry__e51a334e__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4821.280842087005, \"nll_base\": 8.480794906616211, \"eval_ref\": \"mmlu_prox_ur_chemistry__e51a334e__s000__tb131072__v000\", \"oracle_delta... | {
"index": "mmlu_prox_ur_chemistry__e51a334e__s000__tb131072__v000",
"uid": "mmlu_prox_ur_chemistry__e51a334e__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_ur_history__f0ae197e__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2492.6832431253824, \"nll_base\": 7.821115016937256, \"eval_ref\": \"mmlu_prox_ur_history__f0ae197e__s000__tb65536__v000\", \"oracle_delta\": 0.... | {
"index": "mmlu_prox_ur_history__f0ae197e__s000__tb65536__v000",
"uid": "mmlu_prox_ur_history__f0ae197e__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_ur_physics__22a49f33__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 5120.864582283283, \"nll_base\": 8.541078567504883, \"eval_ref\": \"mmlu_prox_ur_physics__22a49f33__s000__tb131072__v000\", \"oracle_delta\": ... | {
"index": "mmlu_prox_ur_physics__22a49f33__s000__tb131072__v000",
"uid": "mmlu_prox_ur_physics__22a49f33__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_vi_business__29db55a2__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3792.4914099017424, \"nll_base\": 8.24077844619751, \"eval_ref\": \"mmlu_prox_vi_business__29db55a2__s000__tb65536__v000\", \"oracle_delta\": 0... | {
"index": "mmlu_prox_vi_business__29db55a2__s000__tb65536__v000",
"uid": "mmlu_prox_vi_business__29db55a2__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_vi_chemistry__9c4ee8ca__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3440.009405628734, \"nll_base\": 8.143229484558105, \"eval_ref\": \"mmlu_prox_vi_chemistry__9c4ee8ca__s000__tb131072__v000\", \"oracle_delta... | {
"index": "mmlu_prox_vi_chemistry__9c4ee8ca__s000__tb131072__v000",
"uid": "mmlu_prox_vi_chemistry__9c4ee8ca__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_vi_computer_science__8c6263a8__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3007.0488776673496, \"nll_base\": 8.008714437484741, \"eval_ref\": \"mmlu_prox_vi_computer_science__8c6263a8__s000__tb262144__v000\",... | {
"index": "mmlu_prox_vi_computer_science__8c6263a8__s000__tb262144__v000",
"uid": "mmlu_prox_vi_computer_science__8c6263a8__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_vi_economics__424c3a1e__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2197.425869364462, \"nll_base\": 7.69504189491272, \"eval_ref\": \"mmlu_prox_vi_economics__424c3a1e__s000__tb65536__v000\", \"oracle_delta\": ... | {
"index": "mmlu_prox_vi_economics__424c3a1e__s000__tb65536__v000",
"uid": "mmlu_prox_vi_economics__424c3a1e__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_vi_health__0ad30dd9__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3064.3675268821, \"nll_base\": 8.027596473693848, \"eval_ref\": \"mmlu_prox_vi_health__0ad30dd9__s000__tb131072__v000\", \"oracle_delta\": 0.34... | {
"index": "mmlu_prox_vi_health__0ad30dd9__s000__tb131072__v000",
"uid": "mmlu_prox_vi_health__0ad30dd9__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_vi_history__193c3485__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2155.906879251976, \"nll_base\": 7.675966739654541, \"eval_ref\": \"mmlu_prox_vi_history__193c3485__s000__tb131072__v000\", \"oracle_delta\": ... | {
"index": "mmlu_prox_vi_history__193c3485__s000__tb131072__v000",
"uid": "mmlu_prox_vi_history__193c3485__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_wo_biology__712eb577__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3157.969136144326, \"nll_base\": 8.057684421539307, \"eval_ref\": \"mmlu_prox_wo_biology__712eb577__s000__tb131072__v000\", \"oracle_delta\": ... | {
"index": "mmlu_prox_wo_biology__712eb577__s000__tb131072__v000",
"uid": "mmlu_prox_wo_biology__712eb577__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_wo_economics__edef65a1__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3527.6002677312003, \"nll_base\": 8.168373107910156, \"eval_ref\": \"mmlu_prox_wo_economics__edef65a1__s000__tb262144__v000\", \"oracle_delt... | {
"index": "mmlu_prox_wo_economics__edef65a1__s000__tb262144__v000",
"uid": "mmlu_prox_wo_economics__edef65a1__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_wo_health__0de096d3__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4331.974926096838, \"nll_base\": 8.373778820037842, \"eval_ref\": \"mmlu_prox_wo_health__0de096d3__s000__tb65536__v000\", \"oracle_delta\": 0.365... | {
"index": "mmlu_prox_wo_health__0de096d3__s000__tb65536__v000",
"uid": "mmlu_prox_wo_health__0de096d3__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_wo_philosophy__bf2ef25f__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4150.203258706203, \"nll_base\": 8.330912590026855, \"eval_ref\": \"mmlu_prox_wo_philosophy__bf2ef25f__s000__tb131072__v000\", \"oracle_del... | {
"index": "mmlu_prox_wo_philosophy__bf2ef25f__s000__tb131072__v000",
"uid": "mmlu_prox_wo_philosophy__bf2ef25f__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_wo_physics__63c919d8__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4832.095794421596, \"nll_base\": 8.483035564422607, \"eval_ref\": \"mmlu_prox_wo_physics__63c919d8__s000__tb131072__v000\", \"oracle_delta\": ... | {
"index": "mmlu_prox_wo_physics__63c919d8__s000__tb131072__v000",
"uid": "mmlu_prox_wo_physics__63c919d8__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_yo_biology__361e162b__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1867.0616828647012, \"nll_base\": 7.532121181488037, \"eval_ref\": \"mmlu_prox_yo_biology__361e162b__s000__tb65536__v000\", \"oracle_delta\": 0.... | {
"index": "mmlu_prox_yo_biology__361e162b__s000__tb65536__v000",
"uid": "mmlu_prox_yo_biology__361e162b__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_yo_chemistry__2906c4b0__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2573.384598497992, \"nll_base\": 7.852977275848389, \"eval_ref\": \"mmlu_prox_yo_chemistry__2906c4b0__s000__tb262144__v000\", \"oracle_delta... | {
"index": "mmlu_prox_yo_chemistry__2906c4b0__s000__tb262144__v000",
"uid": "mmlu_prox_yo_chemistry__2906c4b0__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_yo_engineering__2ba41a5d__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2552.1526758703612, \"nll_base\": 7.8446924686431885, \"eval_ref\": \"mmlu_prox_yo_engineering__2ba41a5d__s000__tb262144__v000\", \"oracle... | {
"index": "mmlu_prox_yo_engineering__2ba41a5d__s000__tb262144__v000",
"uid": "mmlu_prox_yo_engineering__2ba41a5d__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_yo_history__8c7f9b46__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1619.9233238437396, \"nll_base\": 7.39013409614563, \"eval_ref\": \"mmlu_prox_yo_history__8c7f9b46__s000__tb65536__v000\", \"oracle_delta\": 0.3... | {
"index": "mmlu_prox_yo_history__8c7f9b46__s000__tb65536__v000",
"uid": "mmlu_prox_yo_history__8c7f9b46__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_yo_math__b74dd516__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3892.129483458064, \"nll_base\": 8.266711711883545, \"eval_ref\": \"mmlu_prox_yo_math__b74dd516__s000__tb65536__v000\", \"oracle_delta\": 0.4981696... | {
"index": "mmlu_prox_yo_math__b74dd516__s000__tb65536__v000",
"uid": "mmlu_prox_yo_math__b74dd516__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_yo_physics__cf2c58d7__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3930.7502066319535, \"nll_base\": 8.276585578918457, \"eval_ref\": \"mmlu_prox_yo_physics__cf2c58d7__s000__tb262144__v000\", \"oracle_delta\":... | {
"index": "mmlu_prox_yo_physics__cf2c58d7__s000__tb262144__v000",
"uid": "mmlu_prox_yo_physics__cf2c58d7__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_zh_economics__06977ab8__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 7968.074633650364, \"nll_base\": 8.983198165893555, \"eval_ref\": \"mmlu_prox_zh_economics__06977ab8__s000__tb65536__v000\", \"oracle_delta\":... | {
"index": "mmlu_prox_zh_economics__06977ab8__s000__tb65536__v000",
"uid": "mmlu_prox_zh_economics__06977ab8__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_zh_engineering__8ab9b7ca__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 10132.42986885617, \"nll_base\": 9.223496437072754, \"eval_ref\": \"mmlu_prox_zh_engineering__8ab9b7ca__s000__tb131072__v000\", \"oracle_d... | {
"index": "mmlu_prox_zh_engineering__8ab9b7ca__s000__tb131072__v000",
"uid": "mmlu_prox_zh_engineering__8ab9b7ca__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_zh_history__f9c2481c__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 12602.060124417123, \"nll_base\": 9.441615581512451, \"eval_ref\": \"mmlu_prox_zh_history__f9c2481c__s000__tb65536__v000\", \"oracle_delta\": 0.... | {
"index": "mmlu_prox_zh_history__f9c2481c__s000__tb65536__v000",
"uid": "mmlu_prox_zh_history__f9c2481c__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_zu_business__741ef942__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3750.318834938447, \"nll_base\": 8.229596138000488, \"eval_ref\": \"mmlu_prox_zu_business__741ef942__s000__tb262144__v000\", \"oracle_delta\"... | {
"index": "mmlu_prox_zu_business__741ef942__s000__tb262144__v000",
"uid": "mmlu_prox_zu_business__741ef942__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_zu_computer_science__2d35e288__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3269.2306199291065, \"nll_base\": 8.092309951782227, \"eval_ref\": \"mmlu_prox_zu_computer_science__2d35e288__s000__tb131072__v000\",... | {
"index": "mmlu_prox_zu_computer_science__2d35e288__s000__tb131072__v000",
"uid": "mmlu_prox_zu_computer_science__2d35e288__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_zu_engineering__e27476e1__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3757.677711147021, \"nll_base\": 8.231556415557861, \"eval_ref\": \"mmlu_prox_zu_engineering__e27476e1__s000__tb262144__v000\", \"oracle_d... | {
"index": "mmlu_prox_zu_engineering__e27476e1__s000__tb262144__v000",
"uid": "mmlu_prox_zu_engineering__e27476e1__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_zu_psychology__5e20243f__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3959.609695950908, \"nll_base\": 8.283900737762451, \"eval_ref\": \"mmlu_prox_zu_psychology__5e20243f__s000__tb262144__v000\", \"oracle_del... | {
"index": "mmlu_prox_zu_psychology__5e20243f__s000__tb262144__v000",
"uid": "mmlu_prox_zu_psychology__5e20243f__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_pt_llama_anatomy__0a175d86__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 297.04781369746945, \"nll_base\": 5.693893114725749, \"eval_ref\": \"mmlu_pt_llama_anatomy__0a175d86__s000__tb65536__v000\", \"oracle_delta\": ... | {
"index": "mmlu_pt_llama_anatomy__0a175d86__s000__tb65536__v000",
"uid": "mmlu_pt_llama_anatomy__0a175d86__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_pt_llama_business_ethics__427b07d9__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 283.96583674807675, \"nll_base\": 5.648853937784831, \"eval_ref\": \"mmlu_pt_llama_business_ethics__427b07d9__s000__tb65536__v000\", \"... | {
"index": "mmlu_pt_llama_business_ethics__427b07d9__s000__tb65536__v000",
"uid": "mmlu_pt_llama_business_ethics__427b07d9__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_pt_llama_clinical_knowledge__d1dc5c5c__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 262.0477098629818, \"nll_base\": 5.56852658589681, \"eval_ref\": \"mmlu_pt_llama_clinical_knowledge__d1dc5c5c__s000__tb131072__v00... | {
"index": "mmlu_pt_llama_clinical_knowledge__d1dc5c5c__s000__tb131072__v000",
"uid": "mmlu_pt_llama_clinical_knowledge__d1dc5c5c__s000__tb131072__v000",
"axis": "domain",
"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_pt_llama_clinical_knowledge__d1dc5c5c__s000__tb65536__v002\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 262.0477098629818, \"nll_base\": 5.56852658589681, \"eval_ref\": \"mmlu_pt_llama_clinical_knowledge__d1dc5c5c__s000__tb65536__v002\"... | {
"index": "mmlu_pt_llama_clinical_knowledge__d1dc5c5c__s000__tb65536__v002",
"uid": "mmlu_pt_llama_clinical_knowledge__d1dc5c5c__s000__tb65536__v002",
"axis": "domain",
"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_pt_llama_clinical_knowledge__d1dc5c5c__s001__tb131072__v003\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 278.7220389527942, \"nll_base\": 5.630215009053548, \"eval_ref\": \"mmlu_pt_llama_clinical_knowledge__d1dc5c5c__s001__tb131072__v0... | {
"index": "mmlu_pt_llama_clinical_knowledge__d1dc5c5c__s001__tb131072__v003",
"uid": "mmlu_pt_llama_clinical_knowledge__d1dc5c5c__s001__tb131072__v003",
"axis": "domain",
"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_pt_llama_college_biology__39a65264__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 287.1977127000918, \"nll_base\": 5.660170873006185, \"eval_ref\": \"mmlu_pt_llama_college_biology__39a65264__s000__tb65536__v000\", \"o... | {
"index": "mmlu_pt_llama_college_biology__39a65264__s000__tb65536__v000",
"uid": "mmlu_pt_llama_college_biology__39a65264__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_pt_llama_college_chemistry__4905c1b4__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 320.7982280680225, \"nll_base\": 5.770812352498372, \"eval_ref\": \"mmlu_pt_llama_college_chemistry__4905c1b4__s000__tb131072__v000... | {
"index": "mmlu_pt_llama_college_chemistry__4905c1b4__s000__tb131072__v000",
"uid": "mmlu_pt_llama_college_chemistry__4905c1b4__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_pt_llama_college_computer_science__5c5096bb__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 282.9963674474715, \"nll_base\": 5.645434061686198, \"eval_ref\": \"mmlu_pt_llama_college_computer_science__5c5096bb__s000__tb... | {
"index": "mmlu_pt_llama_college_computer_science__5c5096bb__s000__tb65536__v000",
"uid": "mmlu_pt_llama_college_computer_science__5c5096bb__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_pt_llama_college_medicine__eaba13c6__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 297.38681809130634, \"nll_base\": 5.695033709208171, \"eval_ref\": \"mmlu_pt_llama_college_medicine__eaba13c6__s000__tb131072__v000\... | {
"index": "mmlu_pt_llama_college_medicine__eaba13c6__s000__tb131072__v000",
"uid": "mmlu_pt_llama_college_medicine__eaba13c6__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_pt_llama_conceptual_physics__0bfc24ff__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 280.1820456247044, \"nll_base\": 5.635439554850261, \"eval_ref\": \"mmlu_pt_llama_conceptual_physics__0bfc24ff__s000__tb65536__v000\... | {
"index": "mmlu_pt_llama_conceptual_physics__0bfc24ff__s000__tb65536__v000",
"uid": "mmlu_pt_llama_conceptual_physics__0bfc24ff__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_pt_llama_electrical_engineering__dbfc52cd__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 250.77453101481686, \"nll_base\": 5.524554252624512, \"eval_ref\": \"mmlu_pt_llama_electrical_engineering__dbfc52cd__s000__tb655... | {
"index": "mmlu_pt_llama_electrical_engineering__dbfc52cd__s000__tb65536__v000",
"uid": "mmlu_pt_llama_electrical_engineering__dbfc52cd__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_pt_llama_formal_logic__0f22b2f8__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 247.85433746347172, \"nll_base\": 5.51284122467041, \"eval_ref\": \"mmlu_pt_llama_formal_logic__0f22b2f8__s000__tb131072__v000\", \"orac... | {
"index": "mmlu_pt_llama_formal_logic__0f22b2f8__s000__tb131072__v000",
"uid": "mmlu_pt_llama_formal_logic__0f22b2f8__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_pt_llama_high_school_computer_science__108b45f3__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 245.67821198039613, \"nll_base\": 5.504022598266602, \"eval_ref\": \"mmlu_pt_llama_high_school_computer_science__108b45f3_... | {
"index": "mmlu_pt_llama_high_school_computer_science__108b45f3__s000__tb65536__v000",
"uid": "mmlu_pt_llama_high_school_computer_science__108b45f3__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_pt_llama_high_school_geography__965628e1__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 338.8942270968171, \"nll_base\": 5.8256880442301435, \"eval_ref\": \"mmlu_pt_llama_high_school_geography__965628e1__s000__tb65536... | {
"index": "mmlu_pt_llama_high_school_geography__965628e1__s000__tb65536__v000",
"uid": "mmlu_pt_llama_high_school_geography__965628e1__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_pt_llama_high_school_government_and_politics__e04f5a27__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 291.9103097752043, \"nll_base\": 5.676446596781413, \"eval_ref\": \"mmlu_pt_llama_high_school_government_and_poli... | {
"index": "mmlu_pt_llama_high_school_government_and_politics__e04f5a27__s000__tb131072__v000",
"uid": "mmlu_pt_llama_high_school_government_and_politics__e04f5a27__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_pt_llama_high_school_macroeconomics__6b9159ef__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 288.57962192090844, \"nll_base\": 5.664971033732097, \"eval_ref\": \"mmlu_pt_llama_high_school_macroeconomics__6b9159ef__s00... | {
"index": "mmlu_pt_llama_high_school_macroeconomics__6b9159ef__s000__tb65536__v000",
"uid": "mmlu_pt_llama_high_school_macroeconomics__6b9159ef__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_pt_llama_high_school_microeconomics__705594b4__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 281.4063622852079, \"nll_base\": 5.63979975382487, \"eval_ref\": \"mmlu_pt_llama_high_school_microeconomics__705594b4__s00... | {
"index": "mmlu_pt_llama_high_school_microeconomics__705594b4__s000__tb131072__v000",
"uid": "mmlu_pt_llama_high_school_microeconomics__705594b4__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_pt_llama_high_school_statistics__e76d4d42__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 291.37055233116337, \"nll_base\": 5.674595832824707, \"eval_ref\": \"mmlu_pt_llama_high_school_statistics__e76d4d42__s000__tb2... | {
"index": "mmlu_pt_llama_high_school_statistics__e76d4d42__s000__tb262144__v000",
"uid": "mmlu_pt_llama_high_school_statistics__e76d4d42__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_pt_llama_high_school_us_history__eb6b6b28__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 219.1992798771372, \"nll_base\": 5.389981269836426, \"eval_ref\": \"mmlu_pt_llama_high_school_us_history__eb6b6b28__s000__tb26... | {
"index": "mmlu_pt_llama_high_school_us_history__eb6b6b28__s000__tb262144__v000",
"uid": "mmlu_pt_llama_high_school_us_history__eb6b6b28__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_pt_llama_high_school_world_history__0037c6c4__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 187.92476655033386, \"nll_base\": 5.23604170481364, \"eval_ref\": \"mmlu_pt_llama_high_school_world_history__0037c6c4__s000__... | {
"index": "mmlu_pt_llama_high_school_world_history__0037c6c4__s000__tb65536__v000",
"uid": "mmlu_pt_llama_high_school_world_history__0037c6c4__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_pt_llama_jurisprudence__906db656__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 229.36679421839577, \"nll_base\": 5.435322443644206, \"eval_ref\": \"mmlu_pt_llama_jurisprudence__906db656__s000__tb65536__v000\", \"orac... | {
"index": "mmlu_pt_llama_jurisprudence__906db656__s000__tb65536__v000",
"uid": "mmlu_pt_llama_jurisprudence__906db656__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_pt_llama_logical_fallacies__acea9058__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 280.7523899691498, \"nll_base\": 5.637473106384277, \"eval_ref\": \"mmlu_pt_llama_logical_fallacies__acea9058__s000__tb65536__v000\",... | {
"index": "mmlu_pt_llama_logical_fallacies__acea9058__s000__tb65536__v000",
"uid": "mmlu_pt_llama_logical_fallacies__acea9058__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_pt_llama_management__42638a98__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 386.5586704276256, \"nll_base\": 5.957283655802409, \"eval_ref\": \"mmlu_pt_llama_management__42638a98__s000__tb65536__v000\", \"oracle_delt... | {
"index": "mmlu_pt_llama_management__42638a98__s000__tb65536__v000",
"uid": "mmlu_pt_llama_management__42638a98__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_pt_llama_marketing__83021070__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 233.7795192935421, \"nll_base\": 5.454378445943196, \"eval_ref\": \"mmlu_pt_llama_marketing__83021070__s000__tb65536__v000\", \"oracle_delta\... | {
"index": "mmlu_pt_llama_marketing__83021070__s000__tb65536__v000",
"uid": "mmlu_pt_llama_marketing__83021070__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_pt_llama_medical_genetics__f8342500__s000__tb131072__v002\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 276.08201438888676, \"nll_base\": 5.620697975158691, \"eval_ref\": \"mmlu_pt_llama_medical_genetics__f8342500__s000__tb131072__v002\... | {
"index": "mmlu_pt_llama_medical_genetics__f8342500__s000__tb131072__v002",
"uid": "mmlu_pt_llama_medical_genetics__f8342500__s000__tb131072__v002",
"axis": "domain",
"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_pt_llama_medical_genetics__f8342500__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 276.08201438888676, \"nll_base\": 5.620697975158691, \"eval_ref\": \"mmlu_pt_llama_medical_genetics__f8342500__s000__tb65536__v001\", ... | {
"index": "mmlu_pt_llama_medical_genetics__f8342500__s000__tb65536__v001",
"uid": "mmlu_pt_llama_medical_genetics__f8342500__s000__tb65536__v001",
"axis": "domain",
"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_pt_llama_nutrition__036ec92a__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 319.5999746201413, \"nll_base\": 5.767070134480794, \"eval_ref\": \"mmlu_pt_llama_nutrition__036ec92a__s000__tb65536__v000\", \"oracle_delta\... | {
"index": "mmlu_pt_llama_nutrition__036ec92a__s000__tb65536__v000",
"uid": "mmlu_pt_llama_nutrition__036ec92a__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_pt_llama_philosophy__ea8fc5af__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 280.72552737120907, \"nll_base\": 5.637377421061198, \"eval_ref\": \"mmlu_pt_llama_philosophy__ea8fc5af__s000__tb65536__v000\", \"oracle_del... | {
"index": "mmlu_pt_llama_philosophy__ea8fc5af__s000__tb65536__v000",
"uid": "mmlu_pt_llama_philosophy__ea8fc5af__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_pt_llama_prehistory__67016f99__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 289.09766645421115, \"nll_base\": 5.666764577229817, \"eval_ref\": \"mmlu_pt_llama_prehistory__67016f99__s000__tb131072__v000\", \"oracle_... | {
"index": "mmlu_pt_llama_prehistory__67016f99__s000__tb131072__v000",
"uid": "mmlu_pt_llama_prehistory__67016f99__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_pt_llama_professional_law__308fe469__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 165.91320967006462, \"nll_base\": 5.111464818318685, \"eval_ref\": \"mmlu_pt_llama_professional_law__308fe469__s000__tb65536__v000\", ... | {
"index": "mmlu_pt_llama_professional_law__308fe469__s000__tb65536__v000",
"uid": "mmlu_pt_llama_professional_law__308fe469__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_pt_llama_us_foreign_policy__d446ae75__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 306.4765389362208, \"nll_base\": 5.725141207377116, \"eval_ref\": \"mmlu_pt_llama_us_foreign_policy__d446ae75__s000__tb65536__v000\",... | {
"index": "mmlu_pt_llama_us_foreign_policy__d446ae75__s000__tb65536__v000",
"uid": "mmlu_pt_llama_us_foreign_policy__d446ae75__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_public_relations_generative_spanish__cf726129__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4372.6472148092835, \"nll_base\": 8.383123874664307, \"eval_ref\": \"mmlu_public_relations_generative_spanish__cf726129__s... | {
"index": "mmlu_public_relations_generative_spanish__cf726129__s000__tb131072__v000",
"uid": "mmlu_public_relations_generative_spanish__cf726129__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\": \"mmlu_public_relations_generative_spanish__cf726129__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 4372.6472148092835, \"nll_base\": 8.383123874664307, \"eval_ref\": \"mmlu_public_relations_generative_spanish__cf726129__s... | {
"index": "mmlu_public_relations_generative_spanish__cf726129__s000__tb262144__v001",
"uid": "mmlu_public_relations_generative_spanish__cf726129__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\": \"mmlu_redux_anatomy_generative__985c01d1__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1650.9702327466782, \"nll_base\": 7.409118413925171, \"eval_ref\": \"mmlu_redux_anatomy_generative__985c01d1__s000__tb131072__v000\",... | {
"index": "mmlu_redux_anatomy_generative__985c01d1__s000__tb131072__v000",
"uid": "mmlu_redux_anatomy_generative__985c01d1__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\": \"mmlu_redux_anatomy_generative__985c01d1__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1650.9702327466782, \"nll_base\": 7.409118413925171, \"eval_ref\": \"mmlu_redux_anatomy_generative__985c01d1__s000__tb262144__v001\",... | {
"index": "mmlu_redux_anatomy_generative__985c01d1__s000__tb262144__v001",
"uid": "mmlu_redux_anatomy_generative__985c01d1__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\": \"mmlu_redux_astronomy_generative__25fb8296__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2302.7688881506115, \"nll_base\": 7.741867542266846, \"eval_ref\": \"mmlu_redux_astronomy_generative__25fb8296__s000__tb65536__v000\"... | {
"index": "mmlu_redux_astronomy_generative__25fb8296__s000__tb65536__v000",
"uid": "mmlu_redux_astronomy_generative__25fb8296__s000__tb65536__v000",
"axis": "knowledge",
"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_redux_clinical_knowledge_generative__7b2de7ba__s000__tb131072__v002\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2471.583216193923, \"nll_base\": 7.81261420249939, \"eval_ref\": \"mmlu_redux_clinical_knowledge_generative__7b2de7ba__s00... | {
"index": "mmlu_redux_clinical_knowledge_generative__7b2de7ba__s000__tb131072__v002",
"uid": "mmlu_redux_clinical_knowledge_generative__7b2de7ba__s000__tb131072__v002",
"axis": "domain",
"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_redux_clinical_knowledge_generative__7b2de7ba__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2471.583216193923, \"nll_base\": 7.81261420249939, \"eval_ref\": \"mmlu_redux_clinical_knowledge_generative__7b2de7ba__s00... | {
"index": "mmlu_redux_clinical_knowledge_generative__7b2de7ba__s000__tb262144__v000",
"uid": "mmlu_redux_clinical_knowledge_generative__7b2de7ba__s000__tb262144__v000",
"axis": "domain",
"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_redux_clinical_knowledge_generative__7b2de7ba__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2471.583216193923, \"nll_base\": 7.81261420249939, \"eval_ref\": \"mmlu_redux_clinical_knowledge_generative__7b2de7ba__s000_... | {
"index": "mmlu_redux_clinical_knowledge_generative__7b2de7ba__s000__tb65536__v001",
"uid": "mmlu_redux_clinical_knowledge_generative__7b2de7ba__s000__tb65536__v001",
"axis": "domain",
"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_redux_clinical_knowledge_generative__7b2de7ba__s001__tb262144__v003\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2544.834040548859, \"nll_base\": 7.84182071685791, \"eval_ref\": \"mmlu_redux_clinical_knowledge_generative__7b2de7ba__s00... | {
"index": "mmlu_redux_clinical_knowledge_generative__7b2de7ba__s001__tb262144__v003",
"uid": "mmlu_redux_clinical_knowledge_generative__7b2de7ba__s001__tb262144__v003",
"axis": "domain",
"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_redux_college_biology_generative__00643351__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1758.2403497534956, \"nll_base\": 7.472068786621094, \"eval_ref\": \"mmlu_redux_college_biology_generative__00643351__s000__t... | {
"index": "mmlu_redux_college_biology_generative__00643351__s000__tb262144__v000",
"uid": "mmlu_redux_college_biology_generative__00643351__s000__tb262144__v000",
"axis": "knowledge",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.