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_flan_n_shot_loglikelihood_computer_security__66c9d007__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 89.6696297019787, \"nll_base\": 4.496132135391235, \"eval_ref\": \"mmlu_flan_n_shot_loglikelihood_computer_security_... | {
"index": "mmlu_flan_n_shot_loglikelihood_computer_security__66c9d007__s000__tb65536__v001",
"uid": "mmlu_flan_n_shot_loglikelihood_computer_security__66c9d007__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_flan_n_shot_loglikelihood_electrical_engineering__952c277d__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 64.18615350347521, \"nll_base\": 4.161787509918213, \"eval_ref\": \"mmlu_flan_n_shot_loglikelihood_electrical_e... | {
"index": "mmlu_flan_n_shot_loglikelihood_electrical_engineering__952c277d__s000__tb65536__v000",
"uid": "mmlu_flan_n_shot_loglikelihood_electrical_engineering__952c277d__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_flan_n_shot_loglikelihood_elementary_mathematics__623dd253__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 72.34776809512063, \"nll_base\": 4.281484603881836, \"eval_ref\": \"mmlu_flan_n_shot_loglikelihood_elementary... | {
"index": "mmlu_flan_n_shot_loglikelihood_elementary_mathematics__623dd253__s000__tb131072__v000",
"uid": "mmlu_flan_n_shot_loglikelihood_elementary_mathematics__623dd253__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_flan_n_shot_loglikelihood_formal_logic__d4dfdb4c__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 71.93133714547893, \"nll_base\": 4.275712013244629, \"eval_ref\": \"mmlu_flan_n_shot_loglikelihood_formal_logic__d4dfdb4c... | {
"index": "mmlu_flan_n_shot_loglikelihood_formal_logic__d4dfdb4c__s000__tb65536__v000",
"uid": "mmlu_flan_n_shot_loglikelihood_formal_logic__d4dfdb4c__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_flan_n_shot_loglikelihood_global_facts__14560510__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 90.20812865803492, \"nll_base\": 4.502119541168213, \"eval_ref\": \"mmlu_flan_n_shot_loglikelihood_global_facts__145605... | {
"index": "mmlu_flan_n_shot_loglikelihood_global_facts__14560510__s000__tb131072__v001",
"uid": "mmlu_flan_n_shot_loglikelihood_global_facts__14560510__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_flan_n_shot_loglikelihood_global_facts__14560510__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 90.20812865803492, \"nll_base\": 4.502119541168213, \"eval_ref\": \"mmlu_flan_n_shot_loglikelihood_global_facts__14560510... | {
"index": "mmlu_flan_n_shot_loglikelihood_global_facts__14560510__s000__tb65536__v000",
"uid": "mmlu_flan_n_shot_loglikelihood_global_facts__14560510__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_flan_n_shot_loglikelihood_high_school_biology__a0a83c14__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 61.70585531350821, \"nll_base\": 4.122378826141357, \"eval_ref\": \"mmlu_flan_n_shot_loglikelihood_high_school_bio... | {
"index": "mmlu_flan_n_shot_loglikelihood_high_school_biology__a0a83c14__s000__tb65536__v000",
"uid": "mmlu_flan_n_shot_loglikelihood_high_school_biology__a0a83c14__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_flan_n_shot_loglikelihood_high_school_computer_science__d5c08796__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 65.56830054287221, \"nll_base\": 4.183092355728149, \"eval_ref\": \"mmlu_flan_n_shot_loglikelihood_high... | {
"index": "mmlu_flan_n_shot_loglikelihood_high_school_computer_science__d5c08796__s000__tb131072__v001",
"uid": "mmlu_flan_n_shot_loglikelihood_high_school_computer_science__d5c08796__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_flan_n_shot_loglikelihood_high_school_computer_science__d5c08796__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 65.56830054287221, \"nll_base\": 4.183092355728149, \"eval_ref\": \"mmlu_flan_n_shot_loglikelihood_high_s... | {
"index": "mmlu_flan_n_shot_loglikelihood_high_school_computer_science__d5c08796__s000__tb65536__v000",
"uid": "mmlu_flan_n_shot_loglikelihood_high_school_computer_science__d5c08796__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_flan_n_shot_loglikelihood_high_school_european_history__b753e1f2__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 60.16103704466928, \"nll_base\": 4.097024917602539, \"eval_ref\": \"mmlu_flan_n_shot_loglikelihood_high... | {
"index": "mmlu_flan_n_shot_loglikelihood_high_school_european_history__b753e1f2__s000__tb262144__v000",
"uid": "mmlu_flan_n_shot_loglikelihood_high_school_european_history__b753e1f2__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_flan_n_shot_loglikelihood_high_school_european_history__b753e1f2__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 60.16103704466928, \"nll_base\": 4.097024917602539, \"eval_ref\": \"mmlu_flan_n_shot_loglikelihood_high_s... | {
"index": "mmlu_flan_n_shot_loglikelihood_high_school_european_history__b753e1f2__s000__tb65536__v001",
"uid": "mmlu_flan_n_shot_loglikelihood_high_school_european_history__b753e1f2__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_flan_n_shot_loglikelihood_high_school_geography__b4a72f0a__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 66.74585580313646, \"nll_base\": 4.200892210006714, \"eval_ref\": \"mmlu_flan_n_shot_loglikelihood_high_school_g... | {
"index": "mmlu_flan_n_shot_loglikelihood_high_school_geography__b4a72f0a__s000__tb65536__v001",
"uid": "mmlu_flan_n_shot_loglikelihood_high_school_geography__b4a72f0a__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_flan_n_shot_loglikelihood_high_school_government_and_politics__f970a6ab__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 63.87859503086192, \"nll_base\": 4.156984329223633, \"eval_ref\": \"mmlu_flan_n_shot_loglikelihood... | {
"index": "mmlu_flan_n_shot_loglikelihood_high_school_government_and_politics__f970a6ab__s000__tb65536__v001",
"uid": "mmlu_flan_n_shot_loglikelihood_high_school_government_and_politics__f970a6ab__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_flan_n_shot_loglikelihood_high_school_macroeconomics__ca2efd38__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 72.87286272072926, \"nll_base\": 4.2887163162231445, \"eval_ref\": \"mmlu_flan_n_shot_loglikelihood_high_... | {
"index": "mmlu_flan_n_shot_loglikelihood_high_school_macroeconomics__ca2efd38__s000__tb131072__v000",
"uid": "mmlu_flan_n_shot_loglikelihood_high_school_macroeconomics__ca2efd38__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_flan_n_shot_loglikelihood_high_school_microeconomics__93a1be75__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 70.77641161518584, \"nll_base\": 4.259525775909424, \"eval_ref\": \"mmlu_flan_n_shot_loglikelihood_high_s... | {
"index": "mmlu_flan_n_shot_loglikelihood_high_school_microeconomics__93a1be75__s000__tb131072__v000",
"uid": "mmlu_flan_n_shot_loglikelihood_high_school_microeconomics__93a1be75__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_flan_n_shot_loglikelihood_high_school_statistics__e1471968__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 64.64541478958303, \"nll_base\": 4.168917179107666, \"eval_ref\": \"mmlu_flan_n_shot_loglikelihood_high_schoo... | {
"index": "mmlu_flan_n_shot_loglikelihood_high_school_statistics__e1471968__s000__tb131072__v000",
"uid": "mmlu_flan_n_shot_loglikelihood_high_school_statistics__e1471968__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_flan_n_shot_loglikelihood_high_school_us_history__1ebbccc3__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 49.25513100511003, \"nll_base\": 3.897013545036316, \"eval_ref\": \"mmlu_flan_n_shot_loglikelihood_high_schoo... | {
"index": "mmlu_flan_n_shot_loglikelihood_high_school_us_history__1ebbccc3__s000__tb262144__v000",
"uid": "mmlu_flan_n_shot_loglikelihood_high_school_us_history__1ebbccc3__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_flan_n_shot_loglikelihood_high_school_world_history__256db79b__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 59.48576910175177, \"nll_base\": 4.085737109184265, \"eval_ref\": \"mmlu_flan_n_shot_loglikelihood_high_sc... | {
"index": "mmlu_flan_n_shot_loglikelihood_high_school_world_history__256db79b__s000__tb131072__v000",
"uid": "mmlu_flan_n_shot_loglikelihood_high_school_world_history__256db79b__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_flan_n_shot_loglikelihood_high_school_world_history__256db79b__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 59.48576910175177, \"nll_base\": 4.085737109184265, \"eval_ref\": \"mmlu_flan_n_shot_loglikelihood_high_sc... | {
"index": "mmlu_flan_n_shot_loglikelihood_high_school_world_history__256db79b__s000__tb262144__v001",
"uid": "mmlu_flan_n_shot_loglikelihood_high_school_world_history__256db79b__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_flan_n_shot_loglikelihood_human_sexuality__a4a39a0b__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 83.93722584896291, \"nll_base\": 4.430069208145142, \"eval_ref\": \"mmlu_flan_n_shot_loglikelihood_human_sexuality__a4... | {
"index": "mmlu_flan_n_shot_loglikelihood_human_sexuality__a4a39a0b__s000__tb65536__v000",
"uid": "mmlu_flan_n_shot_loglikelihood_human_sexuality__a4a39a0b__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_flan_n_shot_loglikelihood_international_law__8603c7e6__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 70.7943851452089, \"nll_base\": 4.259779691696167, \"eval_ref\": \"mmlu_flan_n_shot_loglikelihood_international_la... | {
"index": "mmlu_flan_n_shot_loglikelihood_international_law__8603c7e6__s000__tb262144__v000",
"uid": "mmlu_flan_n_shot_loglikelihood_international_law__8603c7e6__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_flan_n_shot_loglikelihood_logical_fallacies__5c3be6df__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 67.42559955750825, \"nll_base\": 4.211024761199951, \"eval_ref\": \"mmlu_flan_n_shot_loglikelihood_logical_fallaci... | {
"index": "mmlu_flan_n_shot_loglikelihood_logical_fallacies__5c3be6df__s000__tb131072__v000",
"uid": "mmlu_flan_n_shot_loglikelihood_logical_fallacies__5c3be6df__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_flan_n_shot_loglikelihood_management__a5f4800b__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 76.2484623668299, \"nll_base\": 4.3339972496032715, \"eval_ref\": \"mmlu_flan_n_shot_loglikelihood_management__a5f4800b__... | {
"index": "mmlu_flan_n_shot_loglikelihood_management__a5f4800b__s000__tb131072__v000",
"uid": "mmlu_flan_n_shot_loglikelihood_management__a5f4800b__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_flan_n_shot_loglikelihood_management__a5f4800b__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 76.2484623668299, \"nll_base\": 4.3339972496032715, \"eval_ref\": \"mmlu_flan_n_shot_loglikelihood_management__a5f4800b__... | {
"index": "mmlu_flan_n_shot_loglikelihood_management__a5f4800b__s000__tb262144__v001",
"uid": "mmlu_flan_n_shot_loglikelihood_management__a5f4800b__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_flan_n_shot_loglikelihood_medical_genetics__5b5170f8__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 75.27705221040111, \"nll_base\": 4.3211753368377686, \"eval_ref\": \"mmlu_flan_n_shot_loglikelihood_medical_genetic... | {
"index": "mmlu_flan_n_shot_loglikelihood_medical_genetics__5b5170f8__s000__tb131072__v000",
"uid": "mmlu_flan_n_shot_loglikelihood_medical_genetics__5b5170f8__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_flan_n_shot_loglikelihood_medical_genetics__5b5170f8__s000__tb65536__v002\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 75.27705221040111, \"nll_base\": 4.3211753368377686, \"eval_ref\": \"mmlu_flan_n_shot_loglikelihood_medical_genetics_... | {
"index": "mmlu_flan_n_shot_loglikelihood_medical_genetics__5b5170f8__s000__tb65536__v002",
"uid": "mmlu_flan_n_shot_loglikelihood_medical_genetics__5b5170f8__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_flan_n_shot_loglikelihood_medical_genetics__5b5170f8__s001__tb131072__v003\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 77.24244413889731, \"nll_base\": 4.346949100494385, \"eval_ref\": \"mmlu_flan_n_shot_loglikelihood_medical_genetics... | {
"index": "mmlu_flan_n_shot_loglikelihood_medical_genetics__5b5170f8__s001__tb131072__v003",
"uid": "mmlu_flan_n_shot_loglikelihood_medical_genetics__5b5170f8__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_flan_n_shot_loglikelihood_miscellaneous__9210c69f__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 79.40295312788926, \"nll_base\": 4.37453556060791, \"eval_ref\": \"mmlu_flan_n_shot_loglikelihood_miscellaneous__9210c... | {
"index": "mmlu_flan_n_shot_loglikelihood_miscellaneous__9210c69f__s000__tb131072__v000",
"uid": "mmlu_flan_n_shot_loglikelihood_miscellaneous__9210c69f__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_flan_n_shot_loglikelihood_moral_disputes__1df1f6e9__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 60.71408178857409, \"nll_base\": 4.106175661087036, \"eval_ref\": \"mmlu_flan_n_shot_loglikelihood_moral_disputes__1d... | {
"index": "mmlu_flan_n_shot_loglikelihood_moral_disputes__1df1f6e9__s000__tb131072__v000",
"uid": "mmlu_flan_n_shot_loglikelihood_moral_disputes__1df1f6e9__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_flan_n_shot_loglikelihood_nutrition__f1ecdd43__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 79.6948385356443, \"nll_base\": 4.378204822540283, \"eval_ref\": \"mmlu_flan_n_shot_loglikelihood_nutrition__f1ecdd43__s000_... | {
"index": "mmlu_flan_n_shot_loglikelihood_nutrition__f1ecdd43__s000__tb65536__v000",
"uid": "mmlu_flan_n_shot_loglikelihood_nutrition__f1ecdd43__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_flan_n_shot_loglikelihood_philosophy__b46279df__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 74.62708986999462, \"nll_base\": 4.3125035762786865, \"eval_ref\": \"mmlu_flan_n_shot_loglikelihood_philosophy__b46279df__s... | {
"index": "mmlu_flan_n_shot_loglikelihood_philosophy__b46279df__s000__tb65536__v001",
"uid": "mmlu_flan_n_shot_loglikelihood_philosophy__b46279df__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_flan_n_shot_loglikelihood_prehistory__6361bb24__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 66.09754409610831, \"nll_base\": 4.191131591796875, \"eval_ref\": \"mmlu_flan_n_shot_loglikelihood_prehistory__6361bb24__... | {
"index": "mmlu_flan_n_shot_loglikelihood_prehistory__6361bb24__s000__tb131072__v000",
"uid": "mmlu_flan_n_shot_loglikelihood_prehistory__6361bb24__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_flan_n_shot_loglikelihood_prehistory__6361bb24__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 66.09754409610831, \"nll_base\": 4.191131591796875, \"eval_ref\": \"mmlu_flan_n_shot_loglikelihood_prehistory__6361bb24__... | {
"index": "mmlu_flan_n_shot_loglikelihood_prehistory__6361bb24__s000__tb262144__v001",
"uid": "mmlu_flan_n_shot_loglikelihood_prehistory__6361bb24__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_flan_n_shot_loglikelihood_professional_law__86afceee__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 60.643569836238505, \"nll_base\": 4.105013608932495, \"eval_ref\": \"mmlu_flan_n_shot_loglikelihood_professional_la... | {
"index": "mmlu_flan_n_shot_loglikelihood_professional_law__86afceee__s000__tb131072__v000",
"uid": "mmlu_flan_n_shot_loglikelihood_professional_law__86afceee__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_flan_n_shot_loglikelihood_professional_medicine__4198a3f4__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 66.98774787339738, \"nll_base\": 4.204509735107422, \"eval_ref\": \"mmlu_flan_n_shot_loglikelihood_professiona... | {
"index": "mmlu_flan_n_shot_loglikelihood_professional_medicine__4198a3f4__s000__tb131072__v001",
"uid": "mmlu_flan_n_shot_loglikelihood_professional_medicine__4198a3f4__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_flan_n_shot_loglikelihood_professional_medicine__4198a3f4__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 66.98774787339738, \"nll_base\": 4.204509735107422, \"eval_ref\": \"mmlu_flan_n_shot_loglikelihood_professional_... | {
"index": "mmlu_flan_n_shot_loglikelihood_professional_medicine__4198a3f4__s000__tb65536__v000",
"uid": "mmlu_flan_n_shot_loglikelihood_professional_medicine__4198a3f4__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_flan_n_shot_loglikelihood_sociology__339fdadf__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 82.1043640942602, \"nll_base\": 4.407991170883179, \"eval_ref\": \"mmlu_flan_n_shot_loglikelihood_sociology__339fdadf__s000_... | {
"index": "mmlu_flan_n_shot_loglikelihood_sociology__339fdadf__s000__tb65536__v000",
"uid": "mmlu_flan_n_shot_loglikelihood_sociology__339fdadf__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_flan_n_shot_loglikelihood_world_religions__d5cc231a__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 102.0427862697305, \"nll_base\": 4.625392198562622, \"eval_ref\": \"mmlu_flan_n_shot_loglikelihood_world_religions__d5... | {
"index": "mmlu_flan_n_shot_loglikelihood_world_religions__d5cc231a__s000__tb65536__v001",
"uid": "mmlu_flan_n_shot_loglikelihood_world_religions__d5cc231a__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_formal_logic__eb9cf829__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 942.5742468215852, \"nll_base\": 6.848614692687988, \"eval_ref\": \"mmlu_formal_logic__eb9cf829__s000__tb131072__v001\", \"oracle_delta\": 0.5172... | {
"index": "mmlu_formal_logic__eb9cf829__s000__tb131072__v001",
"uid": "mmlu_formal_logic__eb9cf829__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_formal_logic__eb9cf829__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 942.5742468215852, \"nll_base\": 6.848614692687988, \"eval_ref\": \"mmlu_formal_logic__eb9cf829__s000__tb65536__v000\", \"oracle_delta\": 0.2603461... | {
"index": "mmlu_formal_logic__eb9cf829__s000__tb65536__v000",
"uid": "mmlu_formal_logic__eb9cf829__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_formal_logic_continuation__37f84fb0__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 5.909456674025874, \"nll_base\": 1.776553893862924, \"eval_ref\": \"mmlu_formal_logic_continuation__37f84fb0__s000__tb262144__v000\"... | {
"index": "mmlu_formal_logic_continuation__37f84fb0__s000__tb262144__v000",
"uid": "mmlu_formal_logic_continuation__37f84fb0__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_formal_logic_continuation__37f84fb0__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 5.909456674025874, \"nll_base\": 1.776553893862924, \"eval_ref\": \"mmlu_formal_logic_continuation__37f84fb0__s000__tb65536__v001\", \... | {
"index": "mmlu_formal_logic_continuation__37f84fb0__s000__tb65536__v001",
"uid": "mmlu_formal_logic_continuation__37f84fb0__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_formal_logic_generative__035c2919__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1388.685082524683, \"nll_base\": 7.236112594604492, \"eval_ref\": \"mmlu_formal_logic_generative__035c2919__s000__tb65536__v001\", \"ora... | {
"index": "mmlu_formal_logic_generative__035c2919__s000__tb65536__v001",
"uid": "mmlu_formal_logic_generative__035c2919__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_formal_logic_generative_spanish__4fc4909a__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 2899.356986787946, \"nll_base\": 7.9722442626953125, \"eval_ref\": \"mmlu_formal_logic_generative_spanish__4fc4909a__s000__tb1... | {
"index": "mmlu_formal_logic_generative_spanish__4fc4909a__s000__tb131072__v000",
"uid": "mmlu_formal_logic_generative_spanish__4fc4909a__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_formal_logic_generative_spanish__4fc4909a__s000__tb262144__v001\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 2899.356986787946, \"nll_base\": 7.9722442626953125, \"eval_ref\": \"mmlu_formal_logic_generative_spanish__4fc4909a__s000__tb2... | {
"index": "mmlu_formal_logic_generative_spanish__4fc4909a__s000__tb262144__v001",
"uid": "mmlu_formal_logic_generative_spanish__4fc4909a__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_fr_llama_clinical_knowledge__b6bb6ab8__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 246.9546865418774, \"nll_base\": 5.509204864501953, \"eval_ref\": \"mmlu_fr_llama_clinical_knowledge__b6bb6ab8__s000__tb131072__v0... | {
"index": "mmlu_fr_llama_clinical_knowledge__b6bb6ab8__s000__tb131072__v001",
"uid": "mmlu_fr_llama_clinical_knowledge__b6bb6ab8__s000__tb131072__v001",
"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_fr_llama_clinical_knowledge__b6bb6ab8__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 246.9546865418774, \"nll_base\": 5.509204864501953, \"eval_ref\": \"mmlu_fr_llama_clinical_knowledge__b6bb6ab8__s000__tb65536__v000\... | {
"index": "mmlu_fr_llama_clinical_knowledge__b6bb6ab8__s000__tb65536__v000",
"uid": "mmlu_fr_llama_clinical_knowledge__b6bb6ab8__s000__tb65536__v000",
"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_fr_llama_college_chemistry__3d3a88bb__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 252.08694051350966, \"nll_base\": 5.529774030049642, \"eval_ref\": \"mmlu_fr_llama_college_chemistry__3d3a88bb__s000__tb262144__v00... | {
"index": "mmlu_fr_llama_college_chemistry__3d3a88bb__s000__tb262144__v000",
"uid": "mmlu_fr_llama_college_chemistry__3d3a88bb__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_fr_llama_college_mathematics__bc399383__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 261.0222657029109, \"nll_base\": 5.564605712890625, \"eval_ref\": \"mmlu_fr_llama_college_mathematics__bc399383__s000__tb65536__v00... | {
"index": "mmlu_fr_llama_college_mathematics__bc399383__s000__tb65536__v000",
"uid": "mmlu_fr_llama_college_mathematics__bc399383__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_fr_llama_college_medicine__0bf2af0d__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 268.76731444662363, \"nll_base\": 5.5938460032145185, \"eval_ref\": \"mmlu_fr_llama_college_medicine__0bf2af0d__s000__tb65536__v000\",... | {
"index": "mmlu_fr_llama_college_medicine__0bf2af0d__s000__tb65536__v000",
"uid": "mmlu_fr_llama_college_medicine__0bf2af0d__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_fr_llama_computer_security__3f032d86__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 307.6646109199491, \"nll_base\": 5.729010264078776, \"eval_ref\": \"mmlu_fr_llama_computer_security__3f032d86__s000__tb131072__v000... | {
"index": "mmlu_fr_llama_computer_security__3f032d86__s000__tb131072__v000",
"uid": "mmlu_fr_llama_computer_security__3f032d86__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_fr_llama_high_school_chemistry__2d2fcad3__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 295.0758185667441, \"nll_base\": 5.687232335408528, \"eval_ref\": \"mmlu_fr_llama_high_school_chemistry__2d2fcad3__s000__tb2621... | {
"index": "mmlu_fr_llama_high_school_chemistry__2d2fcad3__s000__tb262144__v000",
"uid": "mmlu_fr_llama_high_school_chemistry__2d2fcad3__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_fr_llama_high_school_computer_science__4b2095a1__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 243.8018225333226, \"nll_base\": 5.496355692545573, \"eval_ref\": \"mmlu_fr_llama_high_school_computer_science__4b2095a1__... | {
"index": "mmlu_fr_llama_high_school_computer_science__4b2095a1__s000__tb65536__v000",
"uid": "mmlu_fr_llama_high_school_computer_science__4b2095a1__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_fr_llama_high_school_geography__c4d4dba0__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 284.3625808411099, \"nll_base\": 5.65025011698405, \"eval_ref\": \"mmlu_fr_llama_high_school_geography__c4d4dba0__s000__tb13107... | {
"index": "mmlu_fr_llama_high_school_geography__c4d4dba0__s000__tb131072__v000",
"uid": "mmlu_fr_llama_high_school_geography__c4d4dba0__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_fr_llama_high_school_macroeconomics__48520a7f__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 298.0205101913279, \"nll_base\": 5.69716231028239, \"eval_ref\": \"mmlu_fr_llama_high_school_macroeconomics__48520a7f__s000_... | {
"index": "mmlu_fr_llama_high_school_macroeconomics__48520a7f__s000__tb65536__v000",
"uid": "mmlu_fr_llama_high_school_macroeconomics__48520a7f__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_fr_llama_high_school_mathematics__de89c12e__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 264.85048485833636, \"nll_base\": 5.579165458679199, \"eval_ref\": \"mmlu_fr_llama_high_school_mathematics__de89c12e__s000__t... | {
"index": "mmlu_fr_llama_high_school_mathematics__de89c12e__s000__tb262144__v000",
"uid": "mmlu_fr_llama_high_school_mathematics__de89c12e__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_fr_llama_high_school_microeconomics__6f4e3186__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 301.84449943032115, \"nll_base\": 5.709911982218425, \"eval_ref\": \"mmlu_fr_llama_high_school_microeconomics__6f4e3186__s00... | {
"index": "mmlu_fr_llama_high_school_microeconomics__6f4e3186__s000__tb65536__v000",
"uid": "mmlu_fr_llama_high_school_microeconomics__6f4e3186__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_fr_llama_high_school_physics__ab742c26__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 256.1579301996149, \"nll_base\": 5.545794169108073, \"eval_ref\": \"mmlu_fr_llama_high_school_physics__ab742c26__s000__tb131072__... | {
"index": "mmlu_fr_llama_high_school_physics__ab742c26__s000__tb131072__v000",
"uid": "mmlu_fr_llama_high_school_physics__ab742c26__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_fr_llama_high_school_psychology__d8e60f02__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 290.24761558850633, \"nll_base\": 5.670734405517578, \"eval_ref\": \"mmlu_fr_llama_high_school_psychology__d8e60f02__s000__tb1... | {
"index": "mmlu_fr_llama_high_school_psychology__d8e60f02__s000__tb131072__v000",
"uid": "mmlu_fr_llama_high_school_psychology__d8e60f02__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_fr_llama_high_school_us_history__329e935b__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 179.15374410446753, \"nll_base\": 5.188244342803955, \"eval_ref\": \"mmlu_fr_llama_high_school_us_history__329e935b__s000__tb1... | {
"index": "mmlu_fr_llama_high_school_us_history__329e935b__s000__tb131072__v000",
"uid": "mmlu_fr_llama_high_school_us_history__329e935b__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_fr_llama_human_aging__eeb0d675__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 323.8045979096244, \"nll_base\": 5.780140240987142, \"eval_ref\": \"mmlu_fr_llama_human_aging__eeb0d675__s000__tb65536__v000\", \"oracle_de... | {
"index": "mmlu_fr_llama_human_aging__eeb0d675__s000__tb65536__v000",
"uid": "mmlu_fr_llama_human_aging__eeb0d675__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_fr_llama_human_sexuality__dc8698a0__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 320.25402029862494, \"nll_base\": 5.7691144943237305, \"eval_ref\": \"mmlu_fr_llama_human_sexuality__dc8698a0__s000__tb131072__v000\"... | {
"index": "mmlu_fr_llama_human_sexuality__dc8698a0__s000__tb131072__v000",
"uid": "mmlu_fr_llama_human_sexuality__dc8698a0__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_fr_llama_international_law__b0678466__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 219.28998920985376, \"nll_base\": 5.390395005544026, \"eval_ref\": \"mmlu_fr_llama_international_law__b0678466__s000__tb65536__v000\"... | {
"index": "mmlu_fr_llama_international_law__b0678466__s000__tb65536__v000",
"uid": "mmlu_fr_llama_international_law__b0678466__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_fr_llama_marketing__419f7722__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 260.19422931753013, \"nll_base\": 5.561428387959798, \"eval_ref\": \"mmlu_fr_llama_marketing__419f7722__s000__tb131072__v000\", \"oracle_de... | {
"index": "mmlu_fr_llama_marketing__419f7722__s000__tb131072__v000",
"uid": "mmlu_fr_llama_marketing__419f7722__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_fr_llama_medical_genetics__7143317b__s000__tb131072__v002\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 294.9430247161662, \"nll_base\": 5.686782201131185, \"eval_ref\": \"mmlu_fr_llama_medical_genetics__7143317b__s000__tb131072__v002\"... | {
"index": "mmlu_fr_llama_medical_genetics__7143317b__s000__tb131072__v002",
"uid": "mmlu_fr_llama_medical_genetics__7143317b__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_fr_llama_medical_genetics__7143317b__s000__tb65536__v001\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 294.9430247161662, \"nll_base\": 5.686782201131185, \"eval_ref\": \"mmlu_fr_llama_medical_genetics__7143317b__s000__tb65536__v001\", \... | {
"index": "mmlu_fr_llama_medical_genetics__7143317b__s000__tb65536__v001",
"uid": "mmlu_fr_llama_medical_genetics__7143317b__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_fr_llama_miscellaneous__416b482d__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 333.1406662240682, \"nll_base\": 5.808564821879069, \"eval_ref\": \"mmlu_fr_llama_miscellaneous__416b482d__s000__tb65536__v000\", \"oracl... | {
"index": "mmlu_fr_llama_miscellaneous__416b482d__s000__tb65536__v000",
"uid": "mmlu_fr_llama_miscellaneous__416b482d__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_fr_llama_moral_disputes__d0623256__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 251.20394632122793, \"nll_base\": 5.5262651443481445, \"eval_ref\": \"mmlu_fr_llama_moral_disputes__d0623256__s000__tb131072__v000\", ... | {
"index": "mmlu_fr_llama_moral_disputes__d0623256__s000__tb131072__v000",
"uid": "mmlu_fr_llama_moral_disputes__d0623256__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_fr_llama_moral_scenarios__296a5dcb__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 381.36189960311435, \"nll_base\": 5.943748792012532, \"eval_ref\": \"mmlu_fr_llama_moral_scenarios__296a5dcb__s000__tb131072__v000\",... | {
"index": "mmlu_fr_llama_moral_scenarios__296a5dcb__s000__tb131072__v000",
"uid": "mmlu_fr_llama_moral_scenarios__296a5dcb__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_fr_llama_professional_law__13229295__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 190.53586003360138, \"nll_base\": 5.249840418497722, \"eval_ref\": \"mmlu_fr_llama_professional_law__13229295__s000__tb65536__v000\", ... | {
"index": "mmlu_fr_llama_professional_law__13229295__s000__tb65536__v000",
"uid": "mmlu_fr_llama_professional_law__13229295__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_fr_llama_professional_psychology__e9169fa6__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 240.70068828648203, \"nll_base\": 5.483554204305013, \"eval_ref\": \"mmlu_fr_llama_professional_psychology__e9169fa6__s000__tb6... | {
"index": "mmlu_fr_llama_professional_psychology__e9169fa6__s000__tb65536__v000",
"uid": "mmlu_fr_llama_professional_psychology__e9169fa6__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_fr_llama_public_relations__e8b72f98__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 324.48942515385306, \"nll_base\": 5.782252947489421, \"eval_ref\": \"mmlu_fr_llama_public_relations__e8b72f98__s000__tb131072__v000\... | {
"index": "mmlu_fr_llama_public_relations__e8b72f98__s000__tb131072__v000",
"uid": "mmlu_fr_llama_public_relations__e8b72f98__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_fr_llama_us_foreign_policy__86c6f0b3__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 278.41386978421485, \"nll_base\": 5.629108746846517, \"eval_ref\": \"mmlu_fr_llama_us_foreign_policy__86c6f0b3__s000__tb65536__v000\"... | {
"index": "mmlu_fr_llama_us_foreign_policy__86c6f0b3__s000__tb65536__v000",
"uid": "mmlu_fr_llama_us_foreign_policy__86c6f0b3__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_fr_llama_virology__ca6ada33__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 260.3679029274294, \"nll_base\": 5.562095642089844, \"eval_ref\": \"mmlu_fr_llama_virology__ca6ada33__s000__tb131072__v000\", \"oracle_delta... | {
"index": "mmlu_fr_llama_virology__ca6ada33__s000__tb131072__v000",
"uid": "mmlu_fr_llama_virology__ca6ada33__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_global_facts__d56d2163__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 1036.626054082766, \"nll_base\": 6.943726539611816, \"eval_ref\": \"mmlu_global_facts__d56d2163__s000__tb65536__v000\", \"oracle_delta\": 0.3397753... | {
"index": "mmlu_global_facts__d56d2163__s000__tb65536__v000",
"uid": "mmlu_global_facts__d56d2163__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_global_facts_continuation__911de3c7__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 33.153731721046604, \"nll_base\": 3.5011552810668944, \"eval_ref\": \"mmlu_global_facts_continuation__911de3c7__s000__tb131072__v000... | {
"index": "mmlu_global_facts_continuation__911de3c7__s000__tb131072__v000",
"uid": "mmlu_global_facts_continuation__911de3c7__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_global_facts_generative__0a3e34e0__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 1265.6288283412964, \"nll_base\": 7.143324375152588, \"eval_ref\": \"mmlu_global_facts_generative__0a3e34e0__s000__tb262144__v000\", \... | {
"index": "mmlu_global_facts_generative__0a3e34e0__s000__tb262144__v000",
"uid": "mmlu_global_facts_generative__0a3e34e0__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_hi_llama_astronomy__75dc07de__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 195.76203502803855, \"nll_base\": 5.276899814605713, \"eval_ref\": \"mmlu_hi_llama_astronomy__75dc07de__s000__tb65536__v000\", \"oracle_delta... | {
"index": "mmlu_hi_llama_astronomy__75dc07de__s000__tb65536__v000",
"uid": "mmlu_hi_llama_astronomy__75dc07de__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_clinical_knowledge__b188e422__s000__tb131072__v001\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 172.5118012853184, \"nll_base\": 5.150465647379558, \"eval_ref\": \"mmlu_hi_llama_clinical_knowledge__b188e422__s000__tb131072__v0... | {
"index": "mmlu_hi_llama_clinical_knowledge__b188e422__s000__tb131072__v001",
"uid": "mmlu_hi_llama_clinical_knowledge__b188e422__s000__tb131072__v001",
"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_hi_llama_clinical_knowledge__b188e422__s000__tb262144__v002\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 172.5118012853184, \"nll_base\": 5.150465647379558, \"eval_ref\": \"mmlu_hi_llama_clinical_knowledge__b188e422__s000__tb262144__v0... | {
"index": "mmlu_hi_llama_clinical_knowledge__b188e422__s000__tb262144__v002",
"uid": "mmlu_hi_llama_clinical_knowledge__b188e422__s000__tb262144__v002",
"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_hi_llama_clinical_knowledge__b188e422__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 172.5118012853184, \"nll_base\": 5.150465647379558, \"eval_ref\": \"mmlu_hi_llama_clinical_knowledge__b188e422__s000__tb65536__v000\... | {
"index": "mmlu_hi_llama_clinical_knowledge__b188e422__s000__tb65536__v000",
"uid": "mmlu_hi_llama_clinical_knowledge__b188e422__s000__tb65536__v000",
"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_hi_llama_college_biology__9c7d4f9c__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 196.51791792444405, \"nll_base\": 5.2807536125183105, \"eval_ref\": \"mmlu_hi_llama_college_biology__9c7d4f9c__s000__tb65536__v000\", \... | {
"index": "mmlu_hi_llama_college_biology__9c7d4f9c__s000__tb65536__v000",
"uid": "mmlu_hi_llama_college_biology__9c7d4f9c__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_college_computer_science__42261cee__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 179.2569983357141, \"nll_base\": 5.188820521036784, \"eval_ref\": \"mmlu_hi_llama_college_computer_science__42261cee__s000__tb... | {
"index": "mmlu_hi_llama_college_computer_science__42261cee__s000__tb65536__v000",
"uid": "mmlu_hi_llama_college_computer_science__42261cee__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_college_mathematics__d7c23f23__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 233.55170164035252, \"nll_base\": 5.453403472900391, \"eval_ref\": \"mmlu_hi_llama_college_mathematics__d7c23f23__s000__tb65536__v0... | {
"index": "mmlu_hi_llama_college_mathematics__d7c23f23__s000__tb65536__v000",
"uid": "mmlu_hi_llama_college_mathematics__d7c23f23__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_electrical_engineering__41b3888d__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 189.06998660102306, \"nll_base\": 5.242117245992024, \"eval_ref\": \"mmlu_hi_llama_electrical_engineering__41b3888d__s000__tb1... | {
"index": "mmlu_hi_llama_electrical_engineering__41b3888d__s000__tb131072__v000",
"uid": "mmlu_hi_llama_electrical_engineering__41b3888d__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_elementary_mathematics__792ff3b9__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 225.4927286581777, \"nll_base\": 5.418287913004558, \"eval_ref\": \"mmlu_hi_llama_elementary_mathematics__792ff3b9__s000__tb13... | {
"index": "mmlu_hi_llama_elementary_mathematics__792ff3b9__s000__tb131072__v000",
"uid": "mmlu_hi_llama_elementary_mathematics__792ff3b9__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_formal_logic__f7205a1c__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 204.60487591023082, \"nll_base\": 5.321080684661865, \"eval_ref\": \"mmlu_hi_llama_formal_logic__f7205a1c__s000__tb131072__v000\", \"ora... | {
"index": "mmlu_hi_llama_formal_logic__f7205a1c__s000__tb131072__v000",
"uid": "mmlu_hi_llama_formal_logic__f7205a1c__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_global_facts__f744bbf7__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 225.90384202125946, \"nll_base\": 5.4201094309488935, \"eval_ref\": \"mmlu_hi_llama_global_facts__f744bbf7__s000__tb131072__v000\", \"or... | {
"index": "mmlu_hi_llama_global_facts__f744bbf7__s000__tb131072__v000",
"uid": "mmlu_hi_llama_global_facts__f744bbf7__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_high_school_biology__d158e1d1__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 217.99844872604695, \"nll_base\": 5.384487946828206, \"eval_ref\": \"mmlu_hi_llama_high_school_biology__d158e1d1__s000__tb262144_... | {
"index": "mmlu_hi_llama_high_school_biology__d158e1d1__s000__tb262144__v000",
"uid": "mmlu_hi_llama_high_school_biology__d158e1d1__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_high_school_european_history__ead1d95a__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 109.24131439554952, \"nll_base\": 4.693559328715007, \"eval_ref\": \"mmlu_hi_llama_high_school_european_history__ead1d95a_... | {
"index": "mmlu_hi_llama_high_school_european_history__ead1d95a__s000__tb65536__v000",
"uid": "mmlu_hi_llama_high_school_european_history__ead1d95a__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_high_school_government_and_politics__3797e07f__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 197.41310373231022, \"nll_base\": 5.285298506418864, \"eval_ref\": \"mmlu_hi_llama_high_school_government_and_pol... | {
"index": "mmlu_hi_llama_high_school_government_and_politics__3797e07f__s000__tb262144__v000",
"uid": "mmlu_hi_llama_high_school_government_and_politics__3797e07f__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_high_school_mathematics__9980f373__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 225.1815560485846, \"nll_base\": 5.416906992594401, \"eval_ref\": \"mmlu_hi_llama_high_school_mathematics__9980f373__s000__tb... | {
"index": "mmlu_hi_llama_high_school_mathematics__9980f373__s000__tb262144__v000",
"uid": "mmlu_hi_llama_high_school_mathematics__9980f373__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_high_school_physics__374ad2c9__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 208.3691532397787, \"nll_base\": 5.339311281840007, \"eval_ref\": \"mmlu_hi_llama_high_school_physics__374ad2c9__s000__tb262144__... | {
"index": "mmlu_hi_llama_high_school_physics__374ad2c9__s000__tb262144__v000",
"uid": "mmlu_hi_llama_high_school_physics__374ad2c9__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_high_school_psychology__92dc6ead__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 208.22533228853626, \"nll_base\": 5.338620821634929, \"eval_ref\": \"mmlu_hi_llama_high_school_psychology__92dc6ead__s000__tb655... | {
"index": "mmlu_hi_llama_high_school_psychology__92dc6ead__s000__tb65536__v000",
"uid": "mmlu_hi_llama_high_school_psychology__92dc6ead__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_high_school_statistics__3e66ecb5__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 201.5884014218497, \"nll_base\": 5.306228001912435, \"eval_ref\": \"mmlu_hi_llama_high_school_statistics__3e66ecb5__s000__tb6553... | {
"index": "mmlu_hi_llama_high_school_statistics__3e66ecb5__s000__tb65536__v000",
"uid": "mmlu_hi_llama_high_school_statistics__3e66ecb5__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_logical_fallacies__bcb35d20__s000__tb131072__v000\", \"budget\": 2048, \"token_budget\": 131072, \"ppl_base\": 198.72483690730428, \"nll_base\": 5.291921138763428, \"eval_ref\": \"mmlu_hi_llama_logical_fallacies__bcb35d20__s000__tb131072__v00... | {
"index": "mmlu_hi_llama_logical_fallacies__bcb35d20__s000__tb131072__v000",
"uid": "mmlu_hi_llama_logical_fallacies__bcb35d20__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_machine_learning__813b53ed__s000__tb262144__v000\", \"budget\": 2048, \"token_budget\": 262144, \"ppl_base\": 165.88702514040384, \"nll_base\": 5.111306985219319, \"eval_ref\": \"mmlu_hi_llama_machine_learning__813b53ed__s000__tb262144__v000\... | {
"index": "mmlu_hi_llama_machine_learning__813b53ed__s000__tb262144__v000",
"uid": "mmlu_hi_llama_machine_learning__813b53ed__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_management__620c30b4__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 242.39866280313905, \"nll_base\": 5.490583737691243, \"eval_ref\": \"mmlu_hi_llama_management__620c30b4__s000__tb65536__v000\", \"oracle_del... | {
"index": "mmlu_hi_llama_management__620c30b4__s000__tb65536__v000",
"uid": "mmlu_hi_llama_management__620c30b4__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_marketing__c6e40bc8__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 180.76004193108284, \"nll_base\": 5.197170416514079, \"eval_ref\": \"mmlu_hi_llama_marketing__c6e40bc8__s000__tb65536__v000\", \"oracle_delta... | {
"index": "mmlu_hi_llama_marketing__c6e40bc8__s000__tb65536__v000",
"uid": "mmlu_hi_llama_marketing__c6e40bc8__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_miscellaneous__31a5b26e__s000__tb65536__v000\", \"budget\": 2048, \"token_budget\": 65536, \"ppl_base\": 199.78913985636203, \"nll_base\": 5.297262509663899, \"eval_ref\": \"mmlu_hi_llama_miscellaneous__31a5b26e__s000__tb65536__v000\", \"orac... | {
"index": "mmlu_hi_llama_miscellaneous__31a5b26e__s000__tb65536__v000",
"uid": "mmlu_hi_llama_miscellaneous__31a5b26e__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.