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_hi_llama_prehistory__3c539aa5__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 177.4125698029587, \"nll_base\": 5.178477923075358, \"eval_ref\": \"mmlu_hi_llama_prehistory__3c539aa5__s000__tb262144__v000\", \"oracle_d... | {
"index": "mmlu_hi_llama_prehistory__3c539aa5__s000__tb262144__v000",
"uid": "mmlu_hi_llama_prehistory__3c539aa5__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_hi_llama_professional_accounting__ad27da9f__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 175.85877025576238, \"nll_base\": 5.169681231180827, \"eval_ref\": \"mmlu_hi_llama_professional_accounting__ad27da9f__s000__tb6... | {
"index": "mmlu_hi_llama_professional_accounting__ad27da9f__s000__tb65536__v000",
"uid": "mmlu_hi_llama_professional_accounting__ad27da9f__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_hi_llama_professional_law__f3fe4ae1__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 100.22534294476213, \"nll_base\": 4.607421080271403, \"eval_ref\": \"mmlu_hi_llama_professional_law__f3fe4ae1__s000__tb131072__v000\... | {
"index": "mmlu_hi_llama_professional_law__f3fe4ae1__s000__tb131072__v000",
"uid": "mmlu_hi_llama_professional_law__f3fe4ae1__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_hi_llama_professional_medicine__48427f1b__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 177.83099937664915, \"nll_base\": 5.180833657582601, \"eval_ref\": \"mmlu_hi_llama_professional_medicine__48427f1b__s000__tb262... | {
"index": "mmlu_hi_llama_professional_medicine__48427f1b__s000__tb262144__v000",
"uid": "mmlu_hi_llama_professional_medicine__48427f1b__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_hi_llama_professional_psychology__4661c495__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 175.57185368211242, \"nll_base\": 5.16804838180542, \"eval_ref\": \"mmlu_hi_llama_professional_psychology__4661c495__s000__tb... | {
"index": "mmlu_hi_llama_professional_psychology__4661c495__s000__tb262144__v000",
"uid": "mmlu_hi_llama_professional_psychology__4661c495__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_hi_llama_public_relations__95b2c655__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 197.40993458300116, \"nll_base\": 5.285282452901204, \"eval_ref\": \"mmlu_hi_llama_public_relations__95b2c655__s000__tb131072__v000\... | {
"index": "mmlu_hi_llama_public_relations__95b2c655__s000__tb131072__v000",
"uid": "mmlu_hi_llama_public_relations__95b2c655__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_hi_llama_security_studies__4e332647__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 144.5529226340902, \"nll_base\": 4.9736456871032715, \"eval_ref\": \"mmlu_hi_llama_security_studies__4e332647__s000__tb131072__v000\... | {
"index": "mmlu_hi_llama_security_studies__4e332647__s000__tb131072__v000",
"uid": "mmlu_hi_llama_security_studies__4e332647__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_hi_llama_sociology__d37f7959__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 162.26552460139447, \"nll_base\": 5.089234034220378, \"eval_ref\": \"mmlu_hi_llama_sociology__d37f7959__s000__tb131072__v000\", \"oracle_de... | {
"index": "mmlu_hi_llama_sociology__d37f7959__s000__tb131072__v000",
"uid": "mmlu_hi_llama_sociology__d37f7959__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_hi_llama_us_foreign_policy__40388234__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 158.73706262082416, \"nll_base\": 5.067249139149983, \"eval_ref\": \"mmlu_hi_llama_us_foreign_policy__40388234__s000__tb65536__v000\"... | {
"index": "mmlu_hi_llama_us_foreign_policy__40388234__s000__tb65536__v000",
"uid": "mmlu_hi_llama_us_foreign_policy__40388234__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_hi_llama_virology__867942ac__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 178.9084230325339, \"nll_base\": 5.186874071756999, \"eval_ref\": \"mmlu_hi_llama_virology__867942ac__s000__tb262144__v000\", \"oracle_delta... | {
"index": "mmlu_hi_llama_virology__867942ac__s000__tb262144__v000",
"uid": "mmlu_hi_llama_virology__867942ac__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_high_school_biology__6eaeaa5b__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 835.5761768875402, \"nll_base\": 6.728121519088745, \"eval_ref\": \"mmlu_high_school_biology__6eaeaa5b__s000__tb262144__v000\", \"oracle_d... | {
"index": "mmlu_high_school_biology__6eaeaa5b__s000__tb262144__v000",
"uid": "mmlu_high_school_biology__6eaeaa5b__s000__tb262144__v000",
"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_high_school_biology__6eaeaa5b__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 835.5761768875402, \"nll_base\": 6.728121519088745, \"eval_ref\": \"mmlu_high_school_biology__6eaeaa5b__s000__tb65536__v001\", \"oracle_delt... | {
"index": "mmlu_high_school_biology__6eaeaa5b__s000__tb65536__v001",
"uid": "mmlu_high_school_biology__6eaeaa5b__s000__tb65536__v001",
"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_high_school_biology_continuation__4f069431__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 8.952641433929289, \"nll_base\": 2.191948621026401, \"eval_ref\": \"mmlu_high_school_biology_continuation__4f069431__s000__tb65... | {
"index": "mmlu_high_school_biology_continuation__4f069431__s000__tb65536__v000",
"uid": "mmlu_high_school_biology_continuation__4f069431__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_high_school_biology_generative__42f40835__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 805.1607495868384, \"nll_base\": 6.691041946411133, \"eval_ref\": \"mmlu_high_school_biology_generative__42f40835__s000__tb1310... | {
"index": "mmlu_high_school_biology_generative__42f40835__s000__tb131072__v001",
"uid": "mmlu_high_school_biology_generative__42f40835__s000__tb131072__v001",
"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_high_school_biology_generative__42f40835__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 805.1607495868384, \"nll_base\": 6.691041946411133, \"eval_ref\": \"mmlu_high_school_biology_generative__42f40835__s000__tb65536_... | {
"index": "mmlu_high_school_biology_generative__42f40835__s000__tb65536__v000",
"uid": "mmlu_high_school_biology_generative__42f40835__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_high_school_chemistry_generative_spanish__195f77b9__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3145.0424551758974, \"nll_base\": 8.053582668304443, \"eval_ref\": \"mmlu_high_school_chemistry_generative_spanish__1... | {
"index": "mmlu_high_school_chemistry_generative_spanish__195f77b9__s000__tb131072__v000",
"uid": "mmlu_high_school_chemistry_generative_spanish__195f77b9__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_high_school_computer_science_continuation__b169ebc4__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 4.353227730318826, \"nll_base\": 1.4709175769193672, \"eval_ref\": \"mmlu_high_school_computer_science_continuation_... | {
"index": "mmlu_high_school_computer_science_continuation__b169ebc4__s000__tb262144__v000",
"uid": "mmlu_high_school_computer_science_continuation__b169ebc4__s000__tb262144__v000",
"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_high_school_computer_science_continuation__b169ebc4__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4.353227730318826, \"nll_base\": 1.4709175769193672, \"eval_ref\": \"mmlu_high_school_computer_science_continuation__b... | {
"index": "mmlu_high_school_computer_science_continuation__b169ebc4__s000__tb65536__v001",
"uid": "mmlu_high_school_computer_science_continuation__b169ebc4__s000__tb65536__v001",
"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_high_school_european_history__e95ded7c__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1297.0517248166052, \"nll_base\": 7.167849063873291, \"eval_ref\": \"mmlu_high_school_european_history__e95ded7c__s000__tb131072_... | {
"index": "mmlu_high_school_european_history__e95ded7c__s000__tb131072__v001",
"uid": "mmlu_high_school_european_history__e95ded7c__s000__tb131072__v001",
"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_high_school_european_history__e95ded7c__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1297.0517248166052, \"nll_base\": 7.167849063873291, \"eval_ref\": \"mmlu_high_school_european_history__e95ded7c__s000__tb65536__v0... | {
"index": "mmlu_high_school_european_history__e95ded7c__s000__tb65536__v000",
"uid": "mmlu_high_school_european_history__e95ded7c__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_high_school_european_history_generative__f37af97d__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1186.5859193435078, \"nll_base\": 7.078835487365723, \"eval_ref\": \"mmlu_high_school_european_history_generative__f37... | {
"index": "mmlu_high_school_european_history_generative__f37af97d__s000__tb131072__v001",
"uid": "mmlu_high_school_european_history_generative__f37af97d__s000__tb131072__v001",
"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_high_school_european_history_generative__f37af97d__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1186.5859193435078, \"nll_base\": 7.078835487365723, \"eval_ref\": \"mmlu_high_school_european_history_generative__f37af... | {
"index": "mmlu_high_school_european_history_generative__f37af97d__s000__tb65536__v000",
"uid": "mmlu_high_school_european_history_generative__f37af97d__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_high_school_geography__144528ef__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1255.7634590991147, \"nll_base\": 7.135499000549316, \"eval_ref\": \"mmlu_high_school_geography__144528ef__s000__tb131072__v000\", \"ora... | {
"index": "mmlu_high_school_geography__144528ef__s000__tb131072__v000",
"uid": "mmlu_high_school_geography__144528ef__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_high_school_geography_continuation__10a2848f__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 26.021786670717923, \"nll_base\": 3.258934136012092, \"eval_ref\": \"mmlu_high_school_geography_continuation__10a2848f__s00... | {
"index": "mmlu_high_school_geography_continuation__10a2848f__s000__tb131072__v001",
"uid": "mmlu_high_school_geography_continuation__10a2848f__s000__tb131072__v001",
"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_high_school_geography_continuation__10a2848f__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 26.021786670717923, \"nll_base\": 3.258934136012092, \"eval_ref\": \"mmlu_high_school_geography_continuation__10a2848f__s000_... | {
"index": "mmlu_high_school_geography_continuation__10a2848f__s000__tb65536__v000",
"uid": "mmlu_high_school_geography_continuation__10a2848f__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_high_school_geography_generative__e44f79df__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1226.708240879222, \"nll_base\": 7.11208963394165, \"eval_ref\": \"mmlu_high_school_geography_generative__e44f79df__s000__tb1... | {
"index": "mmlu_high_school_geography_generative__e44f79df__s000__tb131072__v001",
"uid": "mmlu_high_school_geography_generative__e44f79df__s000__tb131072__v001",
"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_high_school_geography_generative__e44f79df__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1226.708240879222, \"nll_base\": 7.11208963394165, \"eval_ref\": \"mmlu_high_school_geography_generative__e44f79df__s000__tb655... | {
"index": "mmlu_high_school_geography_generative__e44f79df__s000__tb65536__v000",
"uid": "mmlu_high_school_geography_generative__e44f79df__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_high_school_government_and_politics__9a25e071__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 841.5351835362719, \"nll_base\": 6.735227823257446, \"eval_ref\": \"mmlu_high_school_government_and_politics__9a25e071__s0... | {
"index": "mmlu_high_school_government_and_politics__9a25e071__s000__tb131072__v001",
"uid": "mmlu_high_school_government_and_politics__9a25e071__s000__tb131072__v001",
"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_high_school_government_and_politics__9a25e071__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 841.5351835362719, \"nll_base\": 6.735227823257446, \"eval_ref\": \"mmlu_high_school_government_and_politics__9a25e071__s000... | {
"index": "mmlu_high_school_government_and_politics__9a25e071__s000__tb65536__v000",
"uid": "mmlu_high_school_government_and_politics__9a25e071__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_high_school_government_and_politics_continuation__963ea155__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 6.619504123380171, \"nll_base\": 1.8900204613095237, \"eval_ref\": \"mmlu_high_school_government_and_politics... | {
"index": "mmlu_high_school_government_and_politics_continuation__963ea155__s000__tb131072__v001",
"uid": "mmlu_high_school_government_and_politics_continuation__963ea155__s000__tb131072__v001",
"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_high_school_government_and_politics_continuation__963ea155__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 6.619504123380171, \"nll_base\": 1.8900204613095237, \"eval_ref\": \"mmlu_high_school_government_and_politics_c... | {
"index": "mmlu_high_school_government_and_politics_continuation__963ea155__s000__tb65536__v000",
"uid": "mmlu_high_school_government_and_politics_continuation__963ea155__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_high_school_government_and_politics_generative__1bc1c1a8__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 696.7080477220278, \"nll_base\": 6.546366453170776, \"eval_ref\": \"mmlu_high_school_government_and_politics_ge... | {
"index": "mmlu_high_school_government_and_politics_generative__1bc1c1a8__s000__tb131072__v000",
"uid": "mmlu_high_school_government_and_politics_generative__1bc1c1a8__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_high_school_government_and_politics_generative_spanish__f68b10c6__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3828.7539329582837, \"nll_base\": 8.25029468536377, \"eval_ref\": \"mmlu_high_school_government_and_polit... | {
"index": "mmlu_high_school_government_and_politics_generative_spanish__f68b10c6__s000__tb65536__v000",
"uid": "mmlu_high_school_government_and_politics_generative_spanish__f68b10c6__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_high_school_macroeconomics__45087498__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 892.4517871708442, \"nll_base\": 6.793972492218018, \"eval_ref\": \"mmlu_high_school_macroeconomics__45087498__s000__tb262144__v000... | {
"index": "mmlu_high_school_macroeconomics__45087498__s000__tb262144__v000",
"uid": "mmlu_high_school_macroeconomics__45087498__s000__tb262144__v000",
"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_high_school_macroeconomics__45087498__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 892.4517871708442, \"nll_base\": 6.793972492218018, \"eval_ref\": \"mmlu_high_school_macroeconomics__45087498__s000__tb65536__v001\",... | {
"index": "mmlu_high_school_macroeconomics__45087498__s000__tb65536__v001",
"uid": "mmlu_high_school_macroeconomics__45087498__s000__tb65536__v001",
"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_high_school_macroeconomics_continuation__6e139975__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 10.580352646825576, \"nll_base\": 2.3589987573326665, \"eval_ref\": \"mmlu_high_school_macroeconomics_continuation__6e... | {
"index": "mmlu_high_school_macroeconomics_continuation__6e139975__s000__tb262144__v000",
"uid": "mmlu_high_school_macroeconomics_continuation__6e139975__s000__tb262144__v000",
"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_high_school_macroeconomics_continuation__6e139975__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 10.580352646825576, \"nll_base\": 2.3589987573326665, \"eval_ref\": \"mmlu_high_school_macroeconomics_continuation__6e13... | {
"index": "mmlu_high_school_macroeconomics_continuation__6e139975__s000__tb65536__v001",
"uid": "mmlu_high_school_macroeconomics_continuation__6e139975__s000__tb65536__v001",
"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_high_school_macroeconomics_generative__82b63121__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 701.217799284299, \"nll_base\": 6.552818536758423, \"eval_ref\": \"mmlu_high_school_macroeconomics_generative__82b63121__s... | {
"index": "mmlu_high_school_macroeconomics_generative__82b63121__s000__tb65536__v000",
"uid": "mmlu_high_school_macroeconomics_generative__82b63121__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_high_school_macroeconomics_generative_spanish__94ccb2e2__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 4060.7173858697906, \"nll_base\": 8.309114933013916, \"eval_ref\": \"mmlu_high_school_macroeconomics_generative_sp... | {
"index": "mmlu_high_school_macroeconomics_generative_spanish__94ccb2e2__s000__tb65536__v000",
"uid": "mmlu_high_school_macroeconomics_generative_spanish__94ccb2e2__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_high_school_mathematics__2b20289b__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1286.6791056978964, \"nll_base\": 7.159819841384888, \"eval_ref\": \"mmlu_high_school_mathematics__2b20289b__s000__tb131072__v001\", \... | {
"index": "mmlu_high_school_mathematics__2b20289b__s000__tb131072__v001",
"uid": "mmlu_high_school_mathematics__2b20289b__s000__tb131072__v001",
"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_high_school_mathematics__2b20289b__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1286.6791056978964, \"nll_base\": 7.159819841384888, \"eval_ref\": \"mmlu_high_school_mathematics__2b20289b__s000__tb65536__v000\", \"or... | {
"index": "mmlu_high_school_mathematics__2b20289b__s000__tb65536__v000",
"uid": "mmlu_high_school_mathematics__2b20289b__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_high_school_mathematics_continuation__71ec53a6__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 28.41225899076514, \"nll_base\": 3.3468207066355187, \"eval_ref\": \"mmlu_high_school_mathematics_continuation__71ec53a6_... | {
"index": "mmlu_high_school_mathematics_continuation__71ec53a6__s000__tb131072__v000",
"uid": "mmlu_high_school_mathematics_continuation__71ec53a6__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_high_school_mathematics_generative_spanish__96036f67__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 5892.7554005860375, \"nll_base\": 8.68147897720337, \"eval_ref\": \"mmlu_high_school_mathematics_generative_spanish... | {
"index": "mmlu_high_school_mathematics_generative_spanish__96036f67__s000__tb131072__v000",
"uid": "mmlu_high_school_mathematics_generative_spanish__96036f67__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_high_school_microeconomics__d94a92e6__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 604.001890398206, \"nll_base\": 6.4035773277282715, \"eval_ref\": \"mmlu_high_school_microeconomics__d94a92e6__s000__tb131072__v001... | {
"index": "mmlu_high_school_microeconomics__d94a92e6__s000__tb131072__v001",
"uid": "mmlu_high_school_microeconomics__d94a92e6__s000__tb131072__v001",
"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_high_school_microeconomics__d94a92e6__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 604.001890398206, \"nll_base\": 6.4035773277282715, \"eval_ref\": \"mmlu_high_school_microeconomics__d94a92e6__s000__tb65536__v000\",... | {
"index": "mmlu_high_school_microeconomics__d94a92e6__s000__tb65536__v000",
"uid": "mmlu_high_school_microeconomics__d94a92e6__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_high_school_microeconomics_continuation__d7337997__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 5.073160477844954, \"nll_base\": 1.623963991800944, \"eval_ref\": \"mmlu_high_school_microeconomics_continuation__d73379... | {
"index": "mmlu_high_school_microeconomics_continuation__d7337997__s000__tb65536__v000",
"uid": "mmlu_high_school_microeconomics_continuation__d7337997__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_high_school_microeconomics_generative__b95bd7ce__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 580.1720179154592, \"nll_base\": 6.3633246421813965, \"eval_ref\": \"mmlu_high_school_microeconomics_generative__b95bd7c... | {
"index": "mmlu_high_school_microeconomics_generative__b95bd7ce__s000__tb131072__v001",
"uid": "mmlu_high_school_microeconomics_generative__b95bd7ce__s000__tb131072__v001",
"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_high_school_microeconomics_generative__b95bd7ce__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 580.1720179154592, \"nll_base\": 6.3633246421813965, \"eval_ref\": \"mmlu_high_school_microeconomics_generative__b95bd7ce_... | {
"index": "mmlu_high_school_microeconomics_generative__b95bd7ce__s000__tb65536__v000",
"uid": "mmlu_high_school_microeconomics_generative__b95bd7ce__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_high_school_physics__f3293ddf__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 836.9195790324266, \"nll_base\": 6.7297279834747314, \"eval_ref\": \"mmlu_high_school_physics__f3293ddf__s000__tb131072__v000\", \"oracle_... | {
"index": "mmlu_high_school_physics__f3293ddf__s000__tb131072__v000",
"uid": "mmlu_high_school_physics__f3293ddf__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_high_school_physics__f3293ddf__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 836.9195790324266, \"nll_base\": 6.7297279834747314, \"eval_ref\": \"mmlu_high_school_physics__f3293ddf__s000__tb262144__v001\", \"oracle_... | {
"index": "mmlu_high_school_physics__f3293ddf__s000__tb262144__v001",
"uid": "mmlu_high_school_physics__f3293ddf__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_high_school_physics_continuation__3fc20e9e__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 9.448709956095673, \"nll_base\": 2.245878219604492, \"eval_ref\": \"mmlu_high_school_physics_continuation__3fc20e9e__s000__tb... | {
"index": "mmlu_high_school_physics_continuation__3fc20e9e__s000__tb131072__v001",
"uid": "mmlu_high_school_physics_continuation__3fc20e9e__s000__tb131072__v001",
"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_high_school_physics_continuation__3fc20e9e__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 9.448709956095673, \"nll_base\": 2.245878219604492, \"eval_ref\": \"mmlu_high_school_physics_continuation__3fc20e9e__s000__tb65... | {
"index": "mmlu_high_school_physics_continuation__3fc20e9e__s000__tb65536__v000",
"uid": "mmlu_high_school_physics_continuation__3fc20e9e__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_high_school_physics_generative__bd450513__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 709.2778712904196, \"nll_base\": 6.564247369766235, \"eval_ref\": \"mmlu_high_school_physics_generative__bd450513__s000__tb1310... | {
"index": "mmlu_high_school_physics_generative__bd450513__s000__tb131072__v000",
"uid": "mmlu_high_school_physics_generative__bd450513__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_high_school_psychology__47801e77__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1351.823150411845, \"nll_base\": 7.209209442138672, \"eval_ref\": \"mmlu_high_school_psychology__47801e77__s000__tb262144__v000\", \"or... | {
"index": "mmlu_high_school_psychology__47801e77__s000__tb262144__v000",
"uid": "mmlu_high_school_psychology__47801e77__s000__tb262144__v000",
"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_high_school_psychology_continuation__b983d780__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 15.333472714186103, \"nll_base\": 2.7300381978352863, \"eval_ref\": \"mmlu_high_school_psychology_continuation__b983d780__... | {
"index": "mmlu_high_school_psychology_continuation__b983d780__s000__tb131072__v001",
"uid": "mmlu_high_school_psychology_continuation__b983d780__s000__tb131072__v001",
"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_high_school_psychology_continuation__b983d780__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 15.333472714186103, \"nll_base\": 2.7300381978352863, \"eval_ref\": \"mmlu_high_school_psychology_continuation__b983d780__s0... | {
"index": "mmlu_high_school_psychology_continuation__b983d780__s000__tb65536__v000",
"uid": "mmlu_high_school_psychology_continuation__b983d780__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_high_school_psychology_generative__8ee6d09d__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 745.4897078660684, \"nll_base\": 6.614041328430176, \"eval_ref\": \"mmlu_high_school_psychology_generative__8ee6d09d__s000__... | {
"index": "mmlu_high_school_psychology_generative__8ee6d09d__s000__tb131072__v000",
"uid": "mmlu_high_school_psychology_generative__8ee6d09d__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_high_school_psychology_generative_spanish__60c924c5__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3656.0749449092864, \"nll_base\": 8.204145431518555, \"eval_ref\": \"mmlu_high_school_psychology_generative_spanish_... | {
"index": "mmlu_high_school_psychology_generative_spanish__60c924c5__s000__tb131072__v001",
"uid": "mmlu_high_school_psychology_generative_spanish__60c924c5__s000__tb131072__v001",
"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_high_school_psychology_generative_spanish__60c924c5__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3656.0749449092864, \"nll_base\": 8.204145431518555, \"eval_ref\": \"mmlu_high_school_psychology_generative_spanish__6... | {
"index": "mmlu_high_school_psychology_generative_spanish__60c924c5__s000__tb65536__v000",
"uid": "mmlu_high_school_psychology_generative_spanish__60c924c5__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_high_school_statistics__a92cf2c3__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1042.321666151991, \"nll_base\": 6.9492058753967285, \"eval_ref\": \"mmlu_high_school_statistics__a92cf2c3__s000__tb131072__v000\", \"o... | {
"index": "mmlu_high_school_statistics__a92cf2c3__s000__tb131072__v000",
"uid": "mmlu_high_school_statistics__a92cf2c3__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_high_school_statistics_continuation__c1751e80__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3.4208283007602645, \"nll_base\": 1.2298827149552298, \"eval_ref\": \"mmlu_high_school_statistics_continuation__c1751e80__... | {
"index": "mmlu_high_school_statistics_continuation__c1751e80__s000__tb131072__v001",
"uid": "mmlu_high_school_statistics_continuation__c1751e80__s000__tb131072__v001",
"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_high_school_statistics_continuation__c1751e80__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3.4208283007602645, \"nll_base\": 1.2298827149552298, \"eval_ref\": \"mmlu_high_school_statistics_continuation__c1751e80__s0... | {
"index": "mmlu_high_school_statistics_continuation__c1751e80__s000__tb65536__v000",
"uid": "mmlu_high_school_statistics_continuation__c1751e80__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_high_school_statistics_generative__ea272240__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1066.0331637064128, \"nll_base\": 6.9716997146606445, \"eval_ref\": \"mmlu_high_school_statistics_generative__ea272240__s000__... | {
"index": "mmlu_high_school_statistics_generative__ea272240__s000__tb65536__v001",
"uid": "mmlu_high_school_statistics_generative__ea272240__s000__tb65536__v001",
"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_high_school_statistics_generative_spanish__7141e0ab__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3746.944060151043, \"nll_base\": 8.2286958694458, \"eval_ref\": \"mmlu_high_school_statistics_generative_spanish__71... | {
"index": "mmlu_high_school_statistics_generative_spanish__7141e0ab__s000__tb131072__v000",
"uid": "mmlu_high_school_statistics_generative_spanish__7141e0ab__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_high_school_us_history__35d9f69a__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1449.1159137153038, \"nll_base\": 7.2787089347839355, \"eval_ref\": \"mmlu_high_school_us_history__35d9f69a__s000__tb131072__v000\", \"... | {
"index": "mmlu_high_school_us_history__35d9f69a__s000__tb131072__v000",
"uid": "mmlu_high_school_us_history__35d9f69a__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_high_school_us_history__35d9f69a__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1449.1159137153038, \"nll_base\": 7.2787089347839355, \"eval_ref\": \"mmlu_high_school_us_history__35d9f69a__s000__tb262144__v001\", \"... | {
"index": "mmlu_high_school_us_history__35d9f69a__s000__tb262144__v001",
"uid": "mmlu_high_school_us_history__35d9f69a__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_high_school_us_history_generative__b69d6cbc__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1137.608532451486, \"nll_base\": 7.036683559417725, \"eval_ref\": \"mmlu_high_school_us_history_generative__b69d6cbc__s000__... | {
"index": "mmlu_high_school_us_history_generative__b69d6cbc__s000__tb131072__v001",
"uid": "mmlu_high_school_us_history_generative__b69d6cbc__s000__tb131072__v001",
"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_high_school_us_history_generative__b69d6cbc__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1137.608532451486, \"nll_base\": 7.036683559417725, \"eval_ref\": \"mmlu_high_school_us_history_generative__b69d6cbc__s000__tb... | {
"index": "mmlu_high_school_us_history_generative__b69d6cbc__s000__tb65536__v000",
"uid": "mmlu_high_school_us_history_generative__b69d6cbc__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_high_school_us_history_generative_spanish__39f4dc89__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1511.8405107198068, \"nll_base\": 7.321083068847656, \"eval_ref\": \"mmlu_high_school_us_history_generative_spanish_... | {
"index": "mmlu_high_school_us_history_generative_spanish__39f4dc89__s000__tb131072__v000",
"uid": "mmlu_high_school_us_history_generative_spanish__39f4dc89__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_high_school_world_history__898a8010__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1433.8940547416667, \"nll_base\": 7.268149137496948, \"eval_ref\": \"mmlu_high_school_world_history__898a8010__s000__tb131072__v000\... | {
"index": "mmlu_high_school_world_history__898a8010__s000__tb131072__v000",
"uid": "mmlu_high_school_world_history__898a8010__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_high_school_world_history_continuation__273625ba__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 5.517920965101225, \"nll_base\": 1.7080011525865417, \"eval_ref\": \"mmlu_high_school_world_history_continuation__27362... | {
"index": "mmlu_high_school_world_history_continuation__273625ba__s000__tb262144__v000",
"uid": "mmlu_high_school_world_history_continuation__273625ba__s000__tb262144__v000",
"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_high_school_world_history_generative__f41c987e__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1751.2432370960573, \"nll_base\": 7.46808123588562, \"eval_ref\": \"mmlu_high_school_world_history_generative__f41c987e__... | {
"index": "mmlu_high_school_world_history_generative__f41c987e__s000__tb131072__v000",
"uid": "mmlu_high_school_world_history_generative__f41c987e__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_high_school_world_history_generative_spanish__6fb91817__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1609.734963183915, \"nll_base\": 7.383824825286865, \"eval_ref\": \"mmlu_high_school_world_history_generative_spa... | {
"index": "mmlu_high_school_world_history_generative_spanish__6fb91817__s000__tb262144__v000",
"uid": "mmlu_high_school_world_history_generative_spanish__6fb91817__s000__tb262144__v000",
"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_high_school_world_history_generative_spanish__6fb91817__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1609.734963183915, \"nll_base\": 7.383824825286865, \"eval_ref\": \"mmlu_high_school_world_history_generative_spani... | {
"index": "mmlu_high_school_world_history_generative_spanish__6fb91817__s000__tb65536__v001",
"uid": "mmlu_high_school_world_history_generative_spanish__6fb91817__s000__tb65536__v001",
"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_human_aging__00ee60d3__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1229.376124744761, \"nll_base\": 7.114262104034424, \"eval_ref\": \"mmlu_human_aging__00ee60d3__s000__tb131072__v000\", \"oracle_delta\": 0.376531... | {
"index": "mmlu_human_aging__00ee60d3__s000__tb131072__v000",
"uid": "mmlu_human_aging__00ee60d3__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_human_aging__00ee60d3__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1229.376124744761, \"nll_base\": 7.114262104034424, \"eval_ref\": \"mmlu_human_aging__00ee60d3__s000__tb262144__v001\", \"oracle_delta\": 0.681403... | {
"index": "mmlu_human_aging__00ee60d3__s000__tb262144__v001",
"uid": "mmlu_human_aging__00ee60d3__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_human_aging_continuation__222a0bc3__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 33.570753684790844, \"nll_base\": 3.5136552618805474, \"eval_ref\": \"mmlu_human_aging_continuation__222a0bc3__s000__tb131072__v001\"... | {
"index": "mmlu_human_aging_continuation__222a0bc3__s000__tb131072__v001",
"uid": "mmlu_human_aging_continuation__222a0bc3__s000__tb131072__v001",
"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_human_aging_continuation__222a0bc3__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 33.570753684790844, \"nll_base\": 3.5136552618805474, \"eval_ref\": \"mmlu_human_aging_continuation__222a0bc3__s000__tb65536__v000\", \... | {
"index": "mmlu_human_aging_continuation__222a0bc3__s000__tb65536__v000",
"uid": "mmlu_human_aging_continuation__222a0bc3__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_human_aging_generative__c4bba7fa__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 821.8693728186837, \"nll_base\": 6.711581468582153, \"eval_ref\": \"mmlu_human_aging_generative__c4bba7fa__s000__tb262144__v000\", \"or... | {
"index": "mmlu_human_aging_generative__c4bba7fa__s000__tb262144__v000",
"uid": "mmlu_human_aging_generative__c4bba7fa__s000__tb262144__v000",
"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_human_aging_generative__c4bba7fa__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 821.8693728186837, \"nll_base\": 6.711581468582153, \"eval_ref\": \"mmlu_human_aging_generative__c4bba7fa__s000__tb65536__v001\", \"oracl... | {
"index": "mmlu_human_aging_generative__c4bba7fa__s000__tb65536__v001",
"uid": "mmlu_human_aging_generative__c4bba7fa__s000__tb65536__v001",
"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_human_sexuality__1daba027__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1288.7720449112746, \"nll_base\": 7.161445140838623, \"eval_ref\": \"mmlu_human_sexuality__1daba027__s000__tb131072__v001\", \"oracle_delta\":... | {
"index": "mmlu_human_sexuality__1daba027__s000__tb131072__v001",
"uid": "mmlu_human_sexuality__1daba027__s000__tb131072__v001",
"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_human_sexuality__1daba027__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1288.7720449112746, \"nll_base\": 7.161445140838623, \"eval_ref\": \"mmlu_human_sexuality__1daba027__s000__tb65536__v000\", \"oracle_delta\": 0.... | {
"index": "mmlu_human_sexuality__1daba027__s000__tb65536__v000",
"uid": "mmlu_human_sexuality__1daba027__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_human_sexuality_continuation__1ebe30ea__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 28.82642332005971, \"nll_base\": 3.361292442908654, \"eval_ref\": \"mmlu_human_sexuality_continuation__1ebe30ea__s000__tb131072__... | {
"index": "mmlu_human_sexuality_continuation__1ebe30ea__s000__tb131072__v001",
"uid": "mmlu_human_sexuality_continuation__1ebe30ea__s000__tb131072__v001",
"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_human_sexuality_continuation__1ebe30ea__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 28.82642332005971, \"nll_base\": 3.361292442908654, \"eval_ref\": \"mmlu_human_sexuality_continuation__1ebe30ea__s000__tb65536__v00... | {
"index": "mmlu_human_sexuality_continuation__1ebe30ea__s000__tb65536__v000",
"uid": "mmlu_human_sexuality_continuation__1ebe30ea__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_human_sexuality_generative__4ba521f7__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1134.8069580906088, \"nll_base\": 7.034217834472656, \"eval_ref\": \"mmlu_human_sexuality_generative__4ba521f7__s000__tb131072__v00... | {
"index": "mmlu_human_sexuality_generative__4ba521f7__s000__tb131072__v000",
"uid": "mmlu_human_sexuality_generative__4ba521f7__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_international_law__9060be8d__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 761.536654993065, \"nll_base\": 6.635338306427002, \"eval_ref\": \"mmlu_international_law__9060be8d__s000__tb131072__v000\", \"oracle_delta\... | {
"index": "mmlu_international_law__9060be8d__s000__tb131072__v000",
"uid": "mmlu_international_law__9060be8d__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_international_law_continuation__ffbec54f__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2.786567883957615, \"nll_base\": 1.0248106894596245, \"eval_ref\": \"mmlu_international_law_continuation__ffbec54f__s000__tb65536... | {
"index": "mmlu_international_law_continuation__ffbec54f__s000__tb65536__v001",
"uid": "mmlu_international_law_continuation__ffbec54f__s000__tb65536__v001",
"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_international_law_generative__ee36dcea__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 560.7538534670883, \"nll_base\": 6.32928204536438, \"eval_ref\": \"mmlu_international_law_generative__ee36dcea__s000__tb131072__v... | {
"index": "mmlu_international_law_generative__ee36dcea__s000__tb131072__v001",
"uid": "mmlu_international_law_generative__ee36dcea__s000__tb131072__v001",
"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_international_law_generative__ee36dcea__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 560.7538534670883, \"nll_base\": 6.32928204536438, \"eval_ref\": \"mmlu_international_law_generative__ee36dcea__s000__tb65536__v000... | {
"index": "mmlu_international_law_generative__ee36dcea__s000__tb65536__v000",
"uid": "mmlu_international_law_generative__ee36dcea__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_international_law_generative_spanish__ef6484ca__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2886.057315034467, \"nll_base\": 7.967646598815918, \"eval_ref\": \"mmlu_international_law_generative_spanish__ef6484ca__... | {
"index": "mmlu_international_law_generative_spanish__ef6484ca__s000__tb131072__v001",
"uid": "mmlu_international_law_generative_spanish__ef6484ca__s000__tb131072__v001",
"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_international_law_generative_spanish__ef6484ca__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2886.057315034467, \"nll_base\": 7.967646598815918, \"eval_ref\": \"mmlu_international_law_generative_spanish__ef6484ca__s0... | {
"index": "mmlu_international_law_generative_spanish__ef6484ca__s000__tb65536__v000",
"uid": "mmlu_international_law_generative_spanish__ef6484ca__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_it_llama_abstract_algebra__fc914d9d__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 248.28798683598015, \"nll_base\": 5.514589309692383, \"eval_ref\": \"mmlu_it_llama_abstract_algebra__fc914d9d__s000__tb65536__v000\", ... | {
"index": "mmlu_it_llama_abstract_algebra__fc914d9d__s000__tb65536__v000",
"uid": "mmlu_it_llama_abstract_algebra__fc914d9d__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_it_llama_astronomy__67973b3b__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 298.7298049483975, \"nll_base\": 5.699539502461751, \"eval_ref\": \"mmlu_it_llama_astronomy__67973b3b__s000__tb262144__v000\", \"oracle_del... | {
"index": "mmlu_it_llama_astronomy__67973b3b__s000__tb262144__v000",
"uid": "mmlu_it_llama_astronomy__67973b3b__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_it_llama_business_ethics__b17f25a7__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 254.9989551588001, \"nll_base\": 5.5412594477335615, \"eval_ref\": \"mmlu_it_llama_business_ethics__b17f25a7__s000__tb262144__v000\",... | {
"index": "mmlu_it_llama_business_ethics__b17f25a7__s000__tb262144__v000",
"uid": "mmlu_it_llama_business_ethics__b17f25a7__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_it_llama_clinical_knowledge__897c1e9d__s000__tb131072__v002\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 245.89495415812144, \"nll_base\": 5.504904429117839, \"eval_ref\": \"mmlu_it_llama_clinical_knowledge__897c1e9d__s000__tb131072__v... | {
"index": "mmlu_it_llama_clinical_knowledge__897c1e9d__s000__tb131072__v002",
"uid": "mmlu_it_llama_clinical_knowledge__897c1e9d__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_it_llama_clinical_knowledge__897c1e9d__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 245.89495415812144, \"nll_base\": 5.504904429117839, \"eval_ref\": \"mmlu_it_llama_clinical_knowledge__897c1e9d__s000__tb65536__v001... | {
"index": "mmlu_it_llama_clinical_knowledge__897c1e9d__s000__tb65536__v001",
"uid": "mmlu_it_llama_clinical_knowledge__897c1e9d__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_it_llama_college_biology__54240f38__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 216.43537353605961, \"nll_base\": 5.377291997273763, \"eval_ref\": \"mmlu_it_llama_college_biology__54240f38__s000__tb262144__v000\",... | {
"index": "mmlu_it_llama_college_biology__54240f38__s000__tb262144__v000",
"uid": "mmlu_it_llama_college_biology__54240f38__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_it_llama_college_chemistry__071d2236__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 272.2967830871377, \"nll_base\": 5.6068925857543945, \"eval_ref\": \"mmlu_it_llama_college_chemistry__071d2236__s000__tb131072__v00... | {
"index": "mmlu_it_llama_college_chemistry__071d2236__s000__tb131072__v000",
"uid": "mmlu_it_llama_college_chemistry__071d2236__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_it_llama_college_mathematics__af6cd6b6__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 247.23102084945342, \"nll_base\": 5.510323206583659, \"eval_ref\": \"mmlu_it_llama_college_mathematics__af6cd6b6__s000__tb65536__v0... | {
"index": "mmlu_it_llama_college_mathematics__af6cd6b6__s000__tb65536__v000",
"uid": "mmlu_it_llama_college_mathematics__af6cd6b6__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_it_llama_college_medicine__9bfc8a0a__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 259.4762757696753, \"nll_base\": 5.5586652755737305, \"eval_ref\": \"mmlu_it_llama_college_medicine__9bfc8a0a__s000__tb65536__v000\", ... | {
"index": "mmlu_it_llama_college_medicine__9bfc8a0a__s000__tb65536__v000",
"uid": "mmlu_it_llama_college_medicine__9bfc8a0a__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.