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_college_mathematics_generative__04b66413__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 673.6031851617926, \"nll_base\": 6.512641191482544, \"eval_ref\": \"mmlu_college_mathematics_generative__04b66413__s000__tb1310... | {
"index": "mmlu_college_mathematics_generative__04b66413__s000__tb131072__v000",
"uid": "mmlu_college_mathematics_generative__04b66413__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_college_medicine__83ddcf6f__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1036.077524238905, \"nll_base\": 6.943197250366211, \"eval_ref\": \"mmlu_college_medicine__83ddcf6f__s000__tb262144__v000\", \"oracle_delta\"... | {
"index": "mmlu_college_medicine__83ddcf6f__s000__tb262144__v000",
"uid": "mmlu_college_medicine__83ddcf6f__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_college_medicine__83ddcf6f__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1036.077524238905, \"nll_base\": 6.943197250366211, \"eval_ref\": \"mmlu_college_medicine__83ddcf6f__s000__tb65536__v001\", \"oracle_delta\": 0... | {
"index": "mmlu_college_medicine__83ddcf6f__s000__tb65536__v001",
"uid": "mmlu_college_medicine__83ddcf6f__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_college_medicine_continuation__6167d025__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 13.091584788847952, \"nll_base\": 2.571969641260354, \"eval_ref\": \"mmlu_college_medicine_continuation__6167d025__s000__tb13107... | {
"index": "mmlu_college_medicine_continuation__6167d025__s000__tb131072__v000",
"uid": "mmlu_college_medicine_continuation__6167d025__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_college_medicine_generative__5c13bb70__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 814.9747951026171, \"nll_base\": 6.703157186508179, \"eval_ref\": \"mmlu_college_medicine_generative__5c13bb70__s000__tb131072__v0... | {
"index": "mmlu_college_medicine_generative__5c13bb70__s000__tb131072__v001",
"uid": "mmlu_college_medicine_generative__5c13bb70__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_college_medicine_generative__5c13bb70__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 814.9747951026171, \"nll_base\": 6.703157186508179, \"eval_ref\": \"mmlu_college_medicine_generative__5c13bb70__s000__tb65536__v000\... | {
"index": "mmlu_college_medicine_generative__5c13bb70__s000__tb65536__v000",
"uid": "mmlu_college_medicine_generative__5c13bb70__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_college_medicine_generative_spanish__cf3aa7b2__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3191.9438342390554, \"nll_base\": 8.068385362625122, \"eval_ref\": \"mmlu_college_medicine_generative_spanish__cf3aa7b2__s... | {
"index": "mmlu_college_medicine_generative_spanish__cf3aa7b2__s000__tb131072__v000",
"uid": "mmlu_college_medicine_generative_spanish__cf3aa7b2__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_college_medicine_generative_spanish__cf3aa7b2__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3191.9438342390554, \"nll_base\": 8.068385362625122, \"eval_ref\": \"mmlu_college_medicine_generative_spanish__cf3aa7b2__s... | {
"index": "mmlu_college_medicine_generative_spanish__cf3aa7b2__s000__tb262144__v001",
"uid": "mmlu_college_medicine_generative_spanish__cf3aa7b2__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_college_physics__f32716ba__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 933.8117666390082, \"nll_base\": 6.839274883270264, \"eval_ref\": \"mmlu_college_physics__f32716ba__s000__tb131072__v001\", \"oracle_delta\": ... | {
"index": "mmlu_college_physics__f32716ba__s000__tb131072__v001",
"uid": "mmlu_college_physics__f32716ba__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_college_physics__f32716ba__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 933.8117666390082, \"nll_base\": 6.839274883270264, \"eval_ref\": \"mmlu_college_physics__f32716ba__s000__tb65536__v000\", \"oracle_delta\": 0.1... | {
"index": "mmlu_college_physics__f32716ba__s000__tb65536__v000",
"uid": "mmlu_college_physics__f32716ba__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_college_physics_continuation__220002de__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 10.301157188605586, \"nll_base\": 2.3322562373332896, \"eval_ref\": \"mmlu_college_physics_continuation__220002de__s000__tb131072... | {
"index": "mmlu_college_physics_continuation__220002de__s000__tb131072__v000",
"uid": "mmlu_college_physics_continuation__220002de__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_college_physics_generative__ef67b21d__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 870.8616200635314, \"nll_base\": 6.7694830894470215, \"eval_ref\": \"mmlu_college_physics_generative__ef67b21d__s000__tb131072__v00... | {
"index": "mmlu_college_physics_generative__ef67b21d__s000__tb131072__v000",
"uid": "mmlu_college_physics_generative__ef67b21d__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_computer_security__31191125__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1052.5474412618194, \"nll_base\": 6.958968639373779, \"eval_ref\": \"mmlu_computer_security__31191125__s000__tb65536__v000\", \"oracle_delta\"... | {
"index": "mmlu_computer_security__31191125__s000__tb65536__v000",
"uid": "mmlu_computer_security__31191125__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_computer_security_continuation__1f16034a__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 8.881065436499775, \"nll_base\": 2.183921531394676, \"eval_ref\": \"mmlu_computer_security_continuation__1f16034a__s000__tb1310... | {
"index": "mmlu_computer_security_continuation__1f16034a__s000__tb131072__v001",
"uid": "mmlu_computer_security_continuation__1f16034a__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_computer_security_continuation__1f16034a__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 8.881065436499775, \"nll_base\": 2.183921531394676, \"eval_ref\": \"mmlu_computer_security_continuation__1f16034a__s000__tb65536_... | {
"index": "mmlu_computer_security_continuation__1f16034a__s000__tb65536__v000",
"uid": "mmlu_computer_security_continuation__1f16034a__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_computer_security_generative__6781a80f__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 957.721958518345, \"nll_base\": 6.864557504653931, \"eval_ref\": \"mmlu_computer_security_generative__6781a80f__s000__tb131072__v... | {
"index": "mmlu_computer_security_generative__6781a80f__s000__tb131072__v000",
"uid": "mmlu_computer_security_generative__6781a80f__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_computer_security_generative__6781a80f__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 957.721958518345, \"nll_base\": 6.864557504653931, \"eval_ref\": \"mmlu_computer_security_generative__6781a80f__s000__tb262144__v... | {
"index": "mmlu_computer_security_generative__6781a80f__s000__tb262144__v001",
"uid": "mmlu_computer_security_generative__6781a80f__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_computer_security_generative_spanish__2eaaba18__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2737.5066126126685, \"nll_base\": 7.91480278968811, \"eval_ref\": \"mmlu_computer_security_generative_spanish__2eaaba18__... | {
"index": "mmlu_computer_security_generative_spanish__2eaaba18__s000__tb262144__v000",
"uid": "mmlu_computer_security_generative_spanish__2eaaba18__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_computer_security_generative_spanish__2eaaba18__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2737.5066126126685, \"nll_base\": 7.91480278968811, \"eval_ref\": \"mmlu_computer_security_generative_spanish__2eaaba18__s0... | {
"index": "mmlu_computer_security_generative_spanish__2eaaba18__s000__tb65536__v001",
"uid": "mmlu_computer_security_generative_spanish__2eaaba18__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_conceptual_physics_generative_spanish__d3061970__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2381.577063488271, \"nll_base\": 7.775518178939819, \"eval_ref\": \"mmlu_conceptual_physics_generative_spanish__d3061970... | {
"index": "mmlu_conceptual_physics_generative_spanish__d3061970__s000__tb262144__v000",
"uid": "mmlu_conceptual_physics_generative_spanish__d3061970__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_conceptual_physics_generative_spanish__d3061970__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2381.577063488271, \"nll_base\": 7.775518178939819, \"eval_ref\": \"mmlu_conceptual_physics_generative_spanish__d3061970__... | {
"index": "mmlu_conceptual_physics_generative_spanish__d3061970__s000__tb65536__v001",
"uid": "mmlu_conceptual_physics_generative_spanish__d3061970__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_cot_llama_business_ethics__854012c0__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3009.2284366547715, \"nll_base\": 8.00943899154663, \"eval_ref\": \"mmlu_cot_llama_business_ethics__854012c0__s000__tb131072__v000\"... | {
"index": "mmlu_cot_llama_business_ethics__854012c0__s000__tb131072__v000",
"uid": "mmlu_cot_llama_business_ethics__854012c0__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_cot_llama_college_physics__a56acebf__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 3810.2933653202886, \"nll_base\": 8.245461463928223, \"eval_ref\": \"mmlu_cot_llama_college_physics__a56acebf__s000__tb65536__v000\", ... | {
"index": "mmlu_cot_llama_college_physics__a56acebf__s000__tb65536__v000",
"uid": "mmlu_cot_llama_college_physics__a56acebf__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_cot_llama_high_school_psychology__3a6b5863__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 4967.839320279528, \"nll_base\": 8.510740280151367, \"eval_ref\": \"mmlu_cot_llama_high_school_psychology__3a6b5863__s000__tb... | {
"index": "mmlu_cot_llama_high_school_psychology__3a6b5863__s000__tb262144__v000",
"uid": "mmlu_cot_llama_high_school_psychology__3a6b5863__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_cot_llama_high_school_world_history__4e63b201__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1902.745357357836, \"nll_base\": 7.551053047180176, \"eval_ref\": \"mmlu_cot_llama_high_school_world_history__4e63b201__s0... | {
"index": "mmlu_cot_llama_high_school_world_history__4e63b201__s000__tb131072__v000",
"uid": "mmlu_cot_llama_high_school_world_history__4e63b201__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_cot_llama_jurisprudence__def262d0__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3902.6472026988104, \"nll_base\": 8.269410371780396, \"eval_ref\": \"mmlu_cot_llama_jurisprudence__def262d0__s000__tb262144__v000\", \... | {
"index": "mmlu_cot_llama_jurisprudence__def262d0__s000__tb262144__v000",
"uid": "mmlu_cot_llama_jurisprudence__def262d0__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_cot_llama_miscellaneous__ee1823b2__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2791.559657863233, \"nll_base\": 7.934355735778809, \"eval_ref\": \"mmlu_cot_llama_miscellaneous__ee1823b2__s000__tb262144__v000\", \"... | {
"index": "mmlu_cot_llama_miscellaneous__ee1823b2__s000__tb262144__v000",
"uid": "mmlu_cot_llama_miscellaneous__ee1823b2__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_cot_llama_philosophy__94352e53__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 4214.072212420773, \"nll_base\": 8.346184730529785, \"eval_ref\": \"mmlu_cot_llama_philosophy__94352e53__s000__tb131072__v000\", \"oracle... | {
"index": "mmlu_cot_llama_philosophy__94352e53__s000__tb131072__v000",
"uid": "mmlu_cot_llama_philosophy__94352e53__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_cot_llama_security_studies__9a073eae__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2725.849969730774, \"nll_base\": 7.910535573959351, \"eval_ref\": \"mmlu_cot_llama_security_studies__9a073eae__s000__tb262144__v000... | {
"index": "mmlu_cot_llama_security_studies__9a073eae__s000__tb262144__v000",
"uid": "mmlu_cot_llama_security_studies__9a073eae__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_cot_llama_us_foreign_policy__84a47645__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 3172.683716565032, \"nll_base\": 8.062333106994629, \"eval_ref\": \"mmlu_cot_llama_us_foreign_policy__84a47645__s000__tb262144__v0... | {
"index": "mmlu_cot_llama_us_foreign_policy__84a47645__s000__tb262144__v000",
"uid": "mmlu_cot_llama_us_foreign_policy__84a47645__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_cot_llama_virology__616ca420__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2890.667754111636, \"nll_base\": 7.969242811203003, \"eval_ref\": \"mmlu_cot_llama_virology__616ca420__s000__tb262144__v000\", \"oracle_del... | {
"index": "mmlu_cot_llama_virology__616ca420__s000__tb262144__v000",
"uid": "mmlu_cot_llama_virology__616ca420__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_de_llama_anatomy__51fe0a31__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 270.09771770106846, \"nll_base\": 5.598783810933431, \"eval_ref\": \"mmlu_de_llama_anatomy__51fe0a31__s000__tb131072__v000\", \"oracle_delta\... | {
"index": "mmlu_de_llama_anatomy__51fe0a31__s000__tb131072__v000",
"uid": "mmlu_de_llama_anatomy__51fe0a31__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_de_llama_business_ethics__1cd221af__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 261.7245281432176, \"nll_base\": 5.56729253133138, \"eval_ref\": \"mmlu_de_llama_business_ethics__1cd221af__s000__tb131072__v000\", \... | {
"index": "mmlu_de_llama_business_ethics__1cd221af__s000__tb131072__v000",
"uid": "mmlu_de_llama_business_ethics__1cd221af__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_de_llama_clinical_knowledge__3db06499__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 272.9819068129273, \"nll_base\": 5.609405517578125, \"eval_ref\": \"mmlu_de_llama_clinical_knowledge__3db06499__s000__tb131072__v0... | {
"index": "mmlu_de_llama_clinical_knowledge__3db06499__s000__tb131072__v000",
"uid": "mmlu_de_llama_clinical_knowledge__3db06499__s000__tb131072__v000",
"axis": "domain",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_de_llama_clinical_knowledge__3db06499__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 272.9819068129273, \"nll_base\": 5.609405517578125, \"eval_ref\": \"mmlu_de_llama_clinical_knowledge__3db06499__s000__tb262144__v0... | {
"index": "mmlu_de_llama_clinical_knowledge__3db06499__s000__tb262144__v001",
"uid": "mmlu_de_llama_clinical_knowledge__3db06499__s000__tb262144__v001",
"axis": "domain",
"budget": 2048,
"token_budget": 262144,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_de_llama_clinical_knowledge__3db06499__s000__tb65536__v002\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 272.9819068129273, \"nll_base\": 5.609405517578125, \"eval_ref\": \"mmlu_de_llama_clinical_knowledge__3db06499__s000__tb65536__v002\... | {
"index": "mmlu_de_llama_clinical_knowledge__3db06499__s000__tb65536__v002",
"uid": "mmlu_de_llama_clinical_knowledge__3db06499__s000__tb65536__v002",
"axis": "domain",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_de_llama_college_biology__7a07693f__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 251.74587103506113, \"nll_base\": 5.528420130411784, \"eval_ref\": \"mmlu_de_llama_college_biology__7a07693f__s000__tb262144__v000\",... | {
"index": "mmlu_de_llama_college_biology__7a07693f__s000__tb262144__v000",
"uid": "mmlu_de_llama_college_biology__7a07693f__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_de_llama_college_computer_science__e97451b7__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 300.20060219936687, \"nll_base\": 5.704450925191243, \"eval_ref\": \"mmlu_de_llama_college_computer_science__e97451b7__s000__t... | {
"index": "mmlu_de_llama_college_computer_science__e97451b7__s000__tb65536__v000",
"uid": "mmlu_de_llama_college_computer_science__e97451b7__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_de_llama_college_mathematics__d1da331b__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 265.04386416057355, \"nll_base\": 5.5798953374226885, \"eval_ref\": \"mmlu_de_llama_college_mathematics__d1da331b__s000__tb262144... | {
"index": "mmlu_de_llama_college_mathematics__d1da331b__s000__tb262144__v000",
"uid": "mmlu_de_llama_college_mathematics__d1da331b__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_de_llama_college_physics__f2105e35__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 269.64980529880495, \"nll_base\": 5.597124099731445, \"eval_ref\": \"mmlu_de_llama_college_physics__f2105e35__s000__tb131072__v000\",... | {
"index": "mmlu_de_llama_college_physics__f2105e35__s000__tb131072__v000",
"uid": "mmlu_de_llama_college_physics__f2105e35__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_de_llama_conceptual_physics__c4c0d6db__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 298.4331909041175, \"nll_base\": 5.698546091715495, \"eval_ref\": \"mmlu_de_llama_conceptual_physics__c4c0d6db__s000__tb65536__v000\... | {
"index": "mmlu_de_llama_conceptual_physics__c4c0d6db__s000__tb65536__v000",
"uid": "mmlu_de_llama_conceptual_physics__c4c0d6db__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_de_llama_electrical_engineering__ab72ee69__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 231.00545027897715, \"nll_base\": 5.442441304524739, \"eval_ref\": \"mmlu_de_llama_electrical_engineering__ab72ee69__s000__tb1... | {
"index": "mmlu_de_llama_electrical_engineering__ab72ee69__s000__tb131072__v000",
"uid": "mmlu_de_llama_electrical_engineering__ab72ee69__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_de_llama_formal_logic__53003e7c__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 239.0557929703943, \"nll_base\": 5.476696968078613, \"eval_ref\": \"mmlu_de_llama_formal_logic__53003e7c__s000__tb65536__v000\", \"oracle_... | {
"index": "mmlu_de_llama_formal_logic__53003e7c__s000__tb65536__v000",
"uid": "mmlu_de_llama_formal_logic__53003e7c__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_de_llama_global_facts__aa81cbea__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 299.50704717624916, \"nll_base\": 5.7021379470825195, \"eval_ref\": \"mmlu_de_llama_global_facts__aa81cbea__s000__tb262144__v000\", \"or... | {
"index": "mmlu_de_llama_global_facts__aa81cbea__s000__tb262144__v000",
"uid": "mmlu_de_llama_global_facts__aa81cbea__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_de_llama_high_school_chemistry__c170f305__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 257.2439833923513, \"nll_base\": 5.55002498626709, \"eval_ref\": \"mmlu_de_llama_high_school_chemistry__c170f305__s000__tb26214... | {
"index": "mmlu_de_llama_high_school_chemistry__c170f305__s000__tb262144__v000",
"uid": "mmlu_de_llama_high_school_chemistry__c170f305__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_de_llama_high_school_european_history__81d7dc24__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 167.0930921828997, \"nll_base\": 5.118551095326741, \"eval_ref\": \"mmlu_de_llama_high_school_european_history__81d7dc24... | {
"index": "mmlu_de_llama_high_school_european_history__81d7dc24__s000__tb262144__v000",
"uid": "mmlu_de_llama_high_school_european_history__81d7dc24__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_de_llama_high_school_government_and_politics__cd3cbf0c__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 290.7619993352262, \"nll_base\": 5.672505060831706, \"eval_ref\": \"mmlu_de_llama_high_school_government_and_politi... | {
"index": "mmlu_de_llama_high_school_government_and_politics__cd3cbf0c__s000__tb65536__v000",
"uid": "mmlu_de_llama_high_school_government_and_politics__cd3cbf0c__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_de_llama_high_school_macroeconomics__c4ca80ba__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 280.8137998458226, \"nll_base\": 5.637691815694173, \"eval_ref\": \"mmlu_de_llama_high_school_macroeconomics__c4ca80ba__s0... | {
"index": "mmlu_de_llama_high_school_macroeconomics__c4ca80ba__s000__tb131072__v000",
"uid": "mmlu_de_llama_high_school_macroeconomics__c4ca80ba__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_de_llama_high_school_mathematics__55518ab3__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 230.59898031645383, \"nll_base\": 5.440680185953776, \"eval_ref\": \"mmlu_de_llama_high_school_mathematics__55518ab3__s000__t... | {
"index": "mmlu_de_llama_high_school_mathematics__55518ab3__s000__tb131072__v000",
"uid": "mmlu_de_llama_high_school_mathematics__55518ab3__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_de_llama_high_school_physics__b393d03a__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 243.3241780366531, \"nll_base\": 5.494394620259603, \"eval_ref\": \"mmlu_de_llama_high_school_physics__b393d03a__s000__tb131072__... | {
"index": "mmlu_de_llama_high_school_physics__b393d03a__s000__tb131072__v000",
"uid": "mmlu_de_llama_high_school_physics__b393d03a__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_de_llama_high_school_psychology__4a5074ce__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 304.3264170580617, \"nll_base\": 5.718100865681966, \"eval_ref\": \"mmlu_de_llama_high_school_psychology__4a5074ce__s000__tb26... | {
"index": "mmlu_de_llama_high_school_psychology__4a5074ce__s000__tb262144__v000",
"uid": "mmlu_de_llama_high_school_psychology__4a5074ce__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_de_llama_human_sexuality__c9c952dd__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 278.1606825634917, \"nll_base\": 5.628198941548665, \"eval_ref\": \"mmlu_de_llama_human_sexuality__c9c952dd__s000__tb131072__v000\", ... | {
"index": "mmlu_de_llama_human_sexuality__c9c952dd__s000__tb131072__v000",
"uid": "mmlu_de_llama_human_sexuality__c9c952dd__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_de_llama_international_law__d9e1ac2b__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 288.6117316540993, \"nll_base\": 5.665082295735677, \"eval_ref\": \"mmlu_de_llama_international_law__d9e1ac2b__s000__tb65536__v000\",... | {
"index": "mmlu_de_llama_international_law__d9e1ac2b__s000__tb65536__v000",
"uid": "mmlu_de_llama_international_law__d9e1ac2b__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_de_llama_jurisprudence__9b4b1854__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 254.94229912635248, \"nll_base\": 5.541037241617839, \"eval_ref\": \"mmlu_de_llama_jurisprudence__9b4b1854__s000__tb131072__v000\", \"o... | {
"index": "mmlu_de_llama_jurisprudence__9b4b1854__s000__tb131072__v000",
"uid": "mmlu_de_llama_jurisprudence__9b4b1854__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_de_llama_management__3630f7b3__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 329.452735143229, \"nll_base\": 5.797432899475098, \"eval_ref\": \"mmlu_de_llama_management__3630f7b3__s000__tb65536__v000\", \"oracle_delta... | {
"index": "mmlu_de_llama_management__3630f7b3__s000__tb65536__v000",
"uid": "mmlu_de_llama_management__3630f7b3__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_de_llama_medical_genetics__420cf0d0__s000__tb131072__v002\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 244.47034406711867, \"nll_base\": 5.499094009399414, \"eval_ref\": \"mmlu_de_llama_medical_genetics__420cf0d0__s000__tb131072__v002\... | {
"index": "mmlu_de_llama_medical_genetics__420cf0d0__s000__tb131072__v002",
"uid": "mmlu_de_llama_medical_genetics__420cf0d0__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_de_llama_medical_genetics__420cf0d0__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 244.47034406711867, \"nll_base\": 5.499094009399414, \"eval_ref\": \"mmlu_de_llama_medical_genetics__420cf0d0__s000__tb65536__v001\", ... | {
"index": "mmlu_de_llama_medical_genetics__420cf0d0__s000__tb65536__v001",
"uid": "mmlu_de_llama_medical_genetics__420cf0d0__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_de_llama_moral_disputes__a4cfb365__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 229.17212427590286, \"nll_base\": 5.434473355611165, \"eval_ref\": \"mmlu_de_llama_moral_disputes__a4cfb365__s000__tb131072__v000\", \... | {
"index": "mmlu_de_llama_moral_disputes__a4cfb365__s000__tb131072__v000",
"uid": "mmlu_de_llama_moral_disputes__a4cfb365__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_de_llama_philosophy__8f296f64__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 334.7799276577787, \"nll_base\": 5.813473383585612, \"eval_ref\": \"mmlu_de_llama_philosophy__8f296f64__s000__tb262144__v000\", \"oracle_d... | {
"index": "mmlu_de_llama_philosophy__8f296f64__s000__tb262144__v000",
"uid": "mmlu_de_llama_philosophy__8f296f64__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_de_llama_prehistory__e7f3bd91__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 268.15635334656434, \"nll_base\": 5.591570218404134, \"eval_ref\": \"mmlu_de_llama_prehistory__e7f3bd91__s000__tb131072__v000\", \"oracle_... | {
"index": "mmlu_de_llama_prehistory__e7f3bd91__s000__tb131072__v000",
"uid": "mmlu_de_llama_prehistory__e7f3bd91__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_de_llama_professional_accounting__ba3cc79f__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 237.5135556779437, \"nll_base\": 5.470224698384603, \"eval_ref\": \"mmlu_de_llama_professional_accounting__ba3cc79f__s000__tb65... | {
"index": "mmlu_de_llama_professional_accounting__ba3cc79f__s000__tb65536__v000",
"uid": "mmlu_de_llama_professional_accounting__ba3cc79f__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_de_llama_professional_law__1d552169__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 147.0652019131952, \"nll_base\": 4.990876038869222, \"eval_ref\": \"mmlu_de_llama_professional_law__1d552169__s000__tb262144__v000\"... | {
"index": "mmlu_de_llama_professional_law__1d552169__s000__tb262144__v000",
"uid": "mmlu_de_llama_professional_law__1d552169__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_de_llama_public_relations__b578739f__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 316.55253159234866, \"nll_base\": 5.757489204406738, \"eval_ref\": \"mmlu_de_llama_public_relations__b578739f__s000__tb131072__v000\... | {
"index": "mmlu_de_llama_public_relations__b578739f__s000__tb131072__v000",
"uid": "mmlu_de_llama_public_relations__b578739f__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_de_llama_us_foreign_policy__90063e21__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 347.62493574213744, \"nll_base\": 5.851124127705892, \"eval_ref\": \"mmlu_de_llama_us_foreign_policy__90063e21__s000__tb65536__v000\"... | {
"index": "mmlu_de_llama_us_foreign_policy__90063e21__s000__tb65536__v000",
"uid": "mmlu_de_llama_us_foreign_policy__90063e21__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_econometrics__ab11d1a5__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 534.2937481545234, \"nll_base\": 6.280945777893066, \"eval_ref\": \"mmlu_econometrics__ab11d1a5__s000__tb131072__v000\", \"oracle_delta\": 0.4518... | {
"index": "mmlu_econometrics__ab11d1a5__s000__tb131072__v000",
"uid": "mmlu_econometrics__ab11d1a5__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_econometrics_continuation__19668712__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 3.9498227830896298, \"nll_base\": 1.3736707128659644, \"eval_ref\": \"mmlu_econometrics_continuation__19668712__s000__tb131072__v000... | {
"index": "mmlu_econometrics_continuation__19668712__s000__tb131072__v000",
"uid": "mmlu_econometrics_continuation__19668712__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_econometrics_generative__52cbf229__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 688.7400740509603, \"nll_base\": 6.5348639488220215, \"eval_ref\": \"mmlu_econometrics_generative__52cbf229__s000__tb131072__v000\", \... | {
"index": "mmlu_econometrics_generative__52cbf229__s000__tb131072__v000",
"uid": "mmlu_econometrics_generative__52cbf229__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_econometrics_generative_spanish__06eceac5__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2071.341034000684, \"nll_base\": 7.635951519012451, \"eval_ref\": \"mmlu_econometrics_generative_spanish__06eceac5__s000__tb13... | {
"index": "mmlu_econometrics_generative_spanish__06eceac5__s000__tb131072__v000",
"uid": "mmlu_econometrics_generative_spanish__06eceac5__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_electrical_engineering_continuation__ba2e6a9c__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 31.389219155958727, \"nll_base\": 3.4464644949776786, \"eval_ref\": \"mmlu_electrical_engineering_continuation__ba2e6a9c__s0... | {
"index": "mmlu_electrical_engineering_continuation__ba2e6a9c__s000__tb65536__v001",
"uid": "mmlu_electrical_engineering_continuation__ba2e6a9c__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_electrical_engineering_generative__72d6b91b__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 685.0474368226207, \"nll_base\": 6.5294880867004395, \"eval_ref\": \"mmlu_electrical_engineering_generative__72d6b91b__s000_... | {
"index": "mmlu_electrical_engineering_generative__72d6b91b__s000__tb131072__v001",
"uid": "mmlu_electrical_engineering_generative__72d6b91b__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_electrical_engineering_generative__72d6b91b__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 685.0474368226207, \"nll_base\": 6.5294880867004395, \"eval_ref\": \"mmlu_electrical_engineering_generative__72d6b91b__s000__t... | {
"index": "mmlu_electrical_engineering_generative__72d6b91b__s000__tb65536__v000",
"uid": "mmlu_electrical_engineering_generative__72d6b91b__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_electrical_engineering_generative_spanish__d38a4769__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2639.511930396841, \"nll_base\": 7.878349304199219, \"eval_ref\": \"mmlu_electrical_engineering_generative_spanish__... | {
"index": "mmlu_electrical_engineering_generative_spanish__d38a4769__s000__tb262144__v000",
"uid": "mmlu_electrical_engineering_generative_spanish__d38a4769__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_electrical_engineering_generative_spanish__d38a4769__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 2639.511930396841, \"nll_base\": 7.878349304199219, \"eval_ref\": \"mmlu_electrical_engineering_generative_spanish__d3... | {
"index": "mmlu_electrical_engineering_generative_spanish__d38a4769__s000__tb65536__v001",
"uid": "mmlu_electrical_engineering_generative_spanish__d38a4769__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_elementary_mathematics__c41b2a64__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 1101.2158641100625, \"nll_base\": 7.004170179367065, \"eval_ref\": \"mmlu_elementary_mathematics__c41b2a64__s000__tb131072__v000\", \"o... | {
"index": "mmlu_elementary_mathematics__c41b2a64__s000__tb131072__v000",
"uid": "mmlu_elementary_mathematics__c41b2a64__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_elementary_mathematics_continuation__02c34833__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 10.571510344248924, \"nll_base\": 2.3581626793857295, \"eval_ref\": \"mmlu_elementary_mathematics_continuation__02c34833__s0... | {
"index": "mmlu_elementary_mathematics_continuation__02c34833__s000__tb65536__v001",
"uid": "mmlu_elementary_mathematics_continuation__02c34833__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_elementary_mathematics_generative_spanish__508edee2__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 7419.282087837869, \"nll_base\": 8.911837577819824, \"eval_ref\": \"mmlu_elementary_mathematics_generative_spanish__... | {
"index": "mmlu_elementary_mathematics_generative_spanish__508edee2__s000__tb131072__v000",
"uid": "mmlu_elementary_mathematics_generative_spanish__508edee2__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_elementary_mathematics_generative_spanish__508edee2__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 7419.282087837869, \"nll_base\": 8.911837577819824, \"eval_ref\": \"mmlu_elementary_mathematics_generative_spanish__... | {
"index": "mmlu_elementary_mathematics_generative_spanish__508edee2__s000__tb262144__v001",
"uid": "mmlu_elementary_mathematics_generative_spanish__508edee2__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_es_llama_astronomy__cbb0e4d4__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 243.10869717903458, \"nll_base\": 5.493508656819661, \"eval_ref\": \"mmlu_es_llama_astronomy__cbb0e4d4__s000__tb65536__v000\", \"oracle_delta... | {
"index": "mmlu_es_llama_astronomy__cbb0e4d4__s000__tb65536__v000",
"uid": "mmlu_es_llama_astronomy__cbb0e4d4__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_es_llama_business_ethics__6b1b0a5b__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 276.2423181727707, \"nll_base\": 5.621278444925944, \"eval_ref\": \"mmlu_es_llama_business_ethics__6b1b0a5b__s000__tb131072__v000\", ... | {
"index": "mmlu_es_llama_business_ethics__6b1b0a5b__s000__tb131072__v000",
"uid": "mmlu_es_llama_business_ethics__6b1b0a5b__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_es_llama_clinical_knowledge__8b67cdf5__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 294.07627197836115, \"nll_base\": 5.683839162190755, \"eval_ref\": \"mmlu_es_llama_clinical_knowledge__8b67cdf5__s000__tb131072__v... | {
"index": "mmlu_es_llama_clinical_knowledge__8b67cdf5__s000__tb131072__v000",
"uid": "mmlu_es_llama_clinical_knowledge__8b67cdf5__s000__tb131072__v000",
"axis": "domain",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_es_llama_clinical_knowledge__8b67cdf5__s000__tb65536__v002\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 294.07627197836115, \"nll_base\": 5.683839162190755, \"eval_ref\": \"mmlu_es_llama_clinical_knowledge__8b67cdf5__s000__tb65536__v002... | {
"index": "mmlu_es_llama_clinical_knowledge__8b67cdf5__s000__tb65536__v002",
"uid": "mmlu_es_llama_clinical_knowledge__8b67cdf5__s000__tb65536__v002",
"axis": "domain",
"budget": 2048,
"token_budget": 65536,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_es_llama_clinical_knowledge__8b67cdf5__s001__tb131072__v003\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 239.01529171161948, \"nll_base\": 5.4765275319417315, \"eval_ref\": \"mmlu_es_llama_clinical_knowledge__8b67cdf5__s001__tb131072__... | {
"index": "mmlu_es_llama_clinical_knowledge__8b67cdf5__s001__tb131072__v003",
"uid": "mmlu_es_llama_clinical_knowledge__8b67cdf5__s001__tb131072__v003",
"axis": "domain",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
datacuration | [
{
"role": "system",
"content": "You are a **training-data curator**. A small base language model needs to do better on a **target task**, and you must choose which examples from a large pool of candidate training examples it should be fine-tuned on.\n\nThe base model will be LoRA-fine-tuned on exactly the e... | datacuration_selection | {
"style": "datacuration",
"ground_truth": "{\"task_id\": \"mmlu_es_llama_college_biology__b3eba45b__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 210.98846457282312, \"nll_base\": 5.351803461710612, \"eval_ref\": \"mmlu_es_llama_college_biology__b3eba45b__s000__tb131072__v000\",... | {
"index": "mmlu_es_llama_college_biology__b3eba45b__s000__tb131072__v000",
"uid": "mmlu_es_llama_college_biology__b3eba45b__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_es_llama_college_computer_science__d3bfcd98__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 281.5408477907328, \"nll_base\": 5.64027754465739, \"eval_ref\": \"mmlu_es_llama_college_computer_science__d3bfcd98__s000__t... | {
"index": "mmlu_es_llama_college_computer_science__d3bfcd98__s000__tb131072__v000",
"uid": "mmlu_es_llama_college_computer_science__d3bfcd98__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_es_llama_computer_security__38ef124a__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 239.4003689309945, \"nll_base\": 5.478137334187825, \"eval_ref\": \"mmlu_es_llama_computer_security__38ef124a__s000__tb65536__v000\",... | {
"index": "mmlu_es_llama_computer_security__38ef124a__s000__tb65536__v000",
"uid": "mmlu_es_llama_computer_security__38ef124a__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_es_llama_formal_logic__5afebd16__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 258.8045616891145, \"nll_base\": 5.556073188781738, \"eval_ref\": \"mmlu_es_llama_formal_logic__5afebd16__s000__tb262144__v000\", \"orac... | {
"index": "mmlu_es_llama_formal_logic__5afebd16__s000__tb262144__v000",
"uid": "mmlu_es_llama_formal_logic__5afebd16__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_es_llama_global_facts__b70e1dd7__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 287.312770836011, \"nll_base\": 5.660571416219075, \"eval_ref\": \"mmlu_es_llama_global_facts__b70e1dd7__s000__tb131072__v000\", \"oracl... | {
"index": "mmlu_es_llama_global_facts__b70e1dd7__s000__tb131072__v000",
"uid": "mmlu_es_llama_global_facts__b70e1dd7__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_es_llama_high_school_biology__313d1a2a__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 239.7119863819451, \"nll_base\": 5.479438145955403, \"eval_ref\": \"mmlu_es_llama_high_school_biology__313d1a2a__s000__tb262144__... | {
"index": "mmlu_es_llama_high_school_biology__313d1a2a__s000__tb262144__v000",
"uid": "mmlu_es_llama_high_school_biology__313d1a2a__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_es_llama_high_school_computer_science__f249dbff__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 251.82663209439315, \"nll_base\": 5.528740882873535, \"eval_ref\": \"mmlu_es_llama_high_school_computer_science__f249dbff_... | {
"index": "mmlu_es_llama_high_school_computer_science__f249dbff__s000__tb65536__v000",
"uid": "mmlu_es_llama_high_school_computer_science__f249dbff__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_es_llama_high_school_government_and_politics__15a874b8__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 288.65999471552163, \"nll_base\": 5.665249506632487, \"eval_ref\": \"mmlu_es_llama_high_school_government_and_pol... | {
"index": "mmlu_es_llama_high_school_government_and_politics__15a874b8__s000__tb131072__v000",
"uid": "mmlu_es_llama_high_school_government_and_politics__15a874b8__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_es_llama_high_school_macroeconomics__d6b60db9__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 289.078367747022, \"nll_base\": 5.666697820027669, \"eval_ref\": \"mmlu_es_llama_high_school_macroeconomics__d6b60db9__s000_... | {
"index": "mmlu_es_llama_high_school_macroeconomics__d6b60db9__s000__tb65536__v000",
"uid": "mmlu_es_llama_high_school_macroeconomics__d6b60db9__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_es_llama_high_school_statistics__99182413__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 286.6850777195224, \"nll_base\": 5.658384323120117, \"eval_ref\": \"mmlu_es_llama_high_school_statistics__99182413__s000__tb26... | {
"index": "mmlu_es_llama_high_school_statistics__99182413__s000__tb262144__v000",
"uid": "mmlu_es_llama_high_school_statistics__99182413__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_es_llama_high_school_us_history__ce247ab1__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 195.34484744970595, \"nll_base\": 5.274766445159912, \"eval_ref\": \"mmlu_es_llama_high_school_us_history__ce247ab1__s000__tb655... | {
"index": "mmlu_es_llama_high_school_us_history__ce247ab1__s000__tb65536__v000",
"uid": "mmlu_es_llama_high_school_us_history__ce247ab1__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_es_llama_high_school_world_history__70f0eccc__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 178.80727375840237, \"nll_base\": 5.18630854288737, \"eval_ref\": \"mmlu_es_llama_high_school_world_history__70f0eccc__s000... | {
"index": "mmlu_es_llama_high_school_world_history__70f0eccc__s000__tb131072__v000",
"uid": "mmlu_es_llama_high_school_world_history__70f0eccc__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_es_llama_human_sexuality__19a035cf__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 319.6752676766853, \"nll_base\": 5.767305692036946, \"eval_ref\": \"mmlu_es_llama_human_sexuality__19a035cf__s000__tb65536__v000\", \"o... | {
"index": "mmlu_es_llama_human_sexuality__19a035cf__s000__tb65536__v000",
"uid": "mmlu_es_llama_human_sexuality__19a035cf__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_es_llama_international_law__00c8998b__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 189.08026461821694, \"nll_base\": 5.24217160542806, \"eval_ref\": \"mmlu_es_llama_international_law__00c8998b__s000__tb131072__v000... | {
"index": "mmlu_es_llama_international_law__00c8998b__s000__tb131072__v000",
"uid": "mmlu_es_llama_international_law__00c8998b__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_es_llama_jurisprudence__9870424d__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 261.2509667640766, \"nll_base\": 5.565481503804524, \"eval_ref\": \"mmlu_es_llama_jurisprudence__9870424d__s000__tb131072__v000\", \"or... | {
"index": "mmlu_es_llama_jurisprudence__9870424d__s000__tb131072__v000",
"uid": "mmlu_es_llama_jurisprudence__9870424d__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_es_llama_machine_learning__4bf3be50__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 237.6674061674737, \"nll_base\": 5.470872243245442, \"eval_ref\": \"mmlu_es_llama_machine_learning__4bf3be50__s000__tb65536__v000\", \... | {
"index": "mmlu_es_llama_machine_learning__4bf3be50__s000__tb65536__v000",
"uid": "mmlu_es_llama_machine_learning__4bf3be50__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_es_llama_management__9efaed06__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 367.0530734575374, \"nll_base\": 5.905506451924642, \"eval_ref\": \"mmlu_es_llama_management__9efaed06__s000__tb65536__v000\", \"oracle_delt... | {
"index": "mmlu_es_llama_management__9efaed06__s000__tb65536__v000",
"uid": "mmlu_es_llama_management__9efaed06__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_es_llama_marketing__11b5887c__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 236.86240144707102, \"nll_base\": 5.467479387919108, \"eval_ref\": \"mmlu_es_llama_marketing__11b5887c__s000__tb131072__v000\", \"oracle_de... | {
"index": "mmlu_es_llama_marketing__11b5887c__s000__tb131072__v000",
"uid": "mmlu_es_llama_marketing__11b5887c__s000__tb131072__v000",
"axis": "instruction",
"budget": 2048,
"token_budget": 131072,
"variant": "filter"
} |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.