Document six completed LR runs as preparation for full Aya training on Perun. Record selection of LR 0.00010 by lowest final validation loss, selected-model lm-eval and partial SK-bench results, reproducible provenance, and dataset-access/context-truncation limitations. Preserve existing 30k and Titan material.
2.9 KiB
2.9 KiB
| 1 | model_id | training_samples | task | metric | value | unit | status |
|---|---|---|---|---|---|---|---|
| 2 | aya_300k_lr_0p00010 | 300000 | arc_sk | acc,none | 0.30025662959794697 | 0..1 | reported |
| 3 | aya_300k_lr_0p00010 | 300000 | arc_sk | acc_stderr,none | 0.013412022629366669 | 0..1 | reported |
| 4 | aya_300k_lr_0p00010 | 300000 | arc_sk | acc_norm,none | 0.3473053892215569 | 0..1 | reported |
| 5 | aya_300k_lr_0p00010 | 300000 | arc_sk | acc_norm_stderr,none | 0.013931226499492353 | 0..1 | reported |
| 6 | aya_300k_lr_0p00010 | 300000 | hellaswag_sk | acc,none | 0.4083289404322615 | 0..1 | reported |
| 7 | aya_300k_lr_0p00010 | 300000 | hellaswag_sk | acc_stderr,none | 0.005047187883274763 | 0..1 | reported |
| 8 | aya_300k_lr_0p00010 | 300000 | hellaswag_sk | acc_norm,none | 0.5163943068002108 | 0..1 | reported |
| 9 | aya_300k_lr_0p00010 | 300000 | hellaswag_sk | acc_norm_stderr,none | 0.005131456510575215 | 0..1 | reported |
| 10 | aya_300k_lr_0p00010 | 300000 | m_mmlu_sk | acc,none | 0.3102893890675241 | 0..1 | reported |
| 11 | aya_300k_lr_0p00010 | 300000 | m_mmlu_sk | acc_stderr,none | 0.0040478899905589515 | 0..1 | reported |
| 12 | aya_300k_lr_0p00010 | 300000 | sklegal | acc,none | 0.2856071964017991 | 0..1 | reported |
| 13 | aya_300k_lr_0p00010 | 300000 | sklegal | acc_stderr,none | 0.012371946515772182 | 0..1 | reported |
| 14 | aya_300k_lr_0p00010 | 300000 | sklegal | acc_norm,none | 0.46101949025487254 | 0..1 | reported |
| 15 | aya_300k_lr_0p00010 | 300000 | sklegal | acc_norm_stderr,none | 0.013653094581189941 | 0..1 | reported |
| 16 | aya_300k_lr_0p00010 | 300000 | skquad | exact,none | 15.120525931336742 | 0..100 | reported |
| 17 | aya_300k_lr_0p00010 | 300000 | skquad | exact_stderr,none | N/A | 0..100 | unavailable |
| 18 | aya_300k_lr_0p00010 | 300000 | skquad | f1,none | 38.72920807013749 | 0..100 | reported |
| 19 | aya_300k_lr_0p00010 | 300000 | skquad | f1_stderr,none | N/A | 0..100 | unavailable |
| 20 | aya_300k_lr_0p00010 | 300000 | skquad | HasAns_exact,none | 18.540838569047313 | 0..100 | reported |
| 21 | aya_300k_lr_0p00010 | 300000 | skquad | HasAns_exact_stderr,none | N/A | 0..100 | unavailable |
| 22 | aya_300k_lr_0p00010 | 300000 | skquad | HasAns_f1,none | 47.549942420326644 | 0..100 | reported |
| 23 | aya_300k_lr_0p00010 | 300000 | skquad | HasAns_f1_stderr,none | N/A | 0..100 | unavailable |
| 24 | aya_300k_lr_0p00010 | 300000 | skquad | NoAns_exact,none | 0.1681614349775785 | 0..100 | reported |
| 25 | aya_300k_lr_0p00010 | 300000 | skquad | NoAns_exact_stderr,none | N/A | 0..100 | unavailable |
| 26 | aya_300k_lr_0p00010 | 300000 | skquad | NoAns_f1,none | 0.1681614349775785 | 0..100 | reported |
| 27 | aya_300k_lr_0p00010 | 300000 | skquad | NoAns_f1_stderr,none | N/A | 0..100 | unavailable |
| 28 | aya_300k_lr_0p00010 | 300000 | skquad | best_exact,none | 18.75195658979443 | 0..100 | reported |
| 29 | aya_300k_lr_0p00010 | 300000 | skquad | best_exact_stderr,none | N/A | 0..100 | unavailable |
| 30 | aya_300k_lr_0p00010 | 300000 | skquad | best_f1,none | 38.7292080701372 | 0..100 | reported |
| 31 | aya_300k_lr_0p00010 | 300000 | skquad | best_f1_stderr,none | N/A | 0..100 | unavailable |
| 32 | aya_300k_lr_0p00010 | 300000 | truthfulqa_sk_mc1 | acc,none | 0.19665809768637532 | 0..1 | reported |
| 33 | aya_300k_lr_0p00010 | 300000 | truthfulqa_sk_mc1 | acc_stderr,none | 0.014259215011433178 | 0..1 | reported |
| 34 | aya_300k_lr_0p00010 | 300000 | truthfulqa_sk_mc2 | acc,none | 0.3506542164316134 | 0..1 | reported |
| 35 | aya_300k_lr_0p00010 | 300000 | truthfulqa_sk_mc2 | acc_stderr,none | 0.014662940445997986 | 0..1 | reported |