testgeniy / real2_logic_small_scope_step080 /benchmark_comparison.json
Asilarkness's picture
Add accepted no-regression logic small-scope SFT candidate
2277b90 verified
Raw
History Blame Contribute Delete
349 Bytes
{
"v6_fixed100": {
"gsm8k": 24.0,
"math500": 7.0,
"arc_challenge": 26.0,
"folio": 29.0,
"composite": 21.5
},
"candidate_fixed100": {
"gsm8k": 25.0,
"math500": 7.0,
"arc_challenge": 31.0,
"folio": 35.0,
"composite": 24.5
},
"human_preference_balanced80": {
"v6": 55.0,
"candidate": 55.0
}
}