{ "backend_settings": { "apply_chat_template": true, "backend": "vllm", "batch_size": 32, "dtype": "bfloat16", "gpu_memory_utilization": 0.82, "seed": "0,1234,1234,1234", "temperature": "task-default 0", "tensor_parallel_size": 1 }, "benchmark_files": { "base": [ "gsm8k_base.json", "ifeval_base.json", "arc_challenge_base.json", "mmlu_base.json" ], "kosa": [ "gsm8k_kosa.json", "ifeval_kosa.json", "arc_challenge_kosa.json", "mmlu_kosa.json" ] }, "date": "2026-06-11T11:17:13.383746+00:00", "harness": "lm-evaluation-harness 0.4.12", "scored_rows": [ { "arc_acc_norm": 0.5213310580204779, "avg5": 0.7960271980535851, "avg6": 0.7729557704902223, "display_avg": 0.7729557704902223, "display_avg_kind": "avg6", "downloads": null, "group": "kosa", "gsm8k_flex": 0.8559514783927218, "gsm8k_strict": 0.8423047763457164, "ifeval_inst": 0.9028776978417267, "ifeval_prompt": 0.8576709796672828, "mmlu": 0.6575986326734083, "mmlu_humanities": 0.5883103081827843, "mmlu_other": 0.7219182491149019, "mmlu_social": 0.7897302567435814, "mmlu_stem": 0.5686647637170948, "model": "kosa-labs/kosa-4B-it-v1", "status": "scored" }, { "arc_acc_norm": 0.5170648464163823, "avg5": 0.7854459095088681, "avg6": 0.766511860924579, "display_avg": 0.766511860924579, "display_avg_kind": "avg6", "downloads": 798, "group": "same-base genuine accessible", "gsm8k_flex": 0.8688400303260045, "gsm8k_strict": 0.8483699772554966, "ifeval_inst": 0.8741007194244604, "ifeval_prompt": 0.8188539741219963, "mmlu": 0.6718416180031335, "mmlu_humanities": 0.6063761955366631, "mmlu_other": 0.7296427421950434, "mmlu_social": 0.7955801104972375, "mmlu_stem": 0.5918173168411037, "model": "codingmonster1234/chess-sft-modelv2", "status": "scored" }, { "arc_acc_norm": 0.5366894197952219, "avg5": 0.7610974365380638, "avg6": null, "display_avg": 0.7610974365380638, "display_avg_kind": "avg5", "downloads": 3126, "group": "same-base genuine accessible", "gsm8k_flex": 0.8233510235026535, "gsm8k_strict": 0.8218347232752085, "ifeval_inst": 0.841726618705036, "ifeval_prompt": 0.7818853974121996, "model": "Applied-Innovation-Center/Karnak-6B-v1.0", "status": "scored" }, { "arc_acc_norm": 0.42918088737201365, "avg5": 0.7317990152643855, "avg6": null, "display_avg": 0.7317990152643855, "display_avg_kind": "avg5", "downloads": 28076, "group": "previous same-base row", "gsm8k_flex": 0.7907505686125853, "gsm8k_strict": 0.7278241091736164, "ifeval_inst": 0.8812949640287769, "ifeval_prompt": 0.8299445471349353, "model": "unsloth/Qwen3-4B-Instruct-2507", "status": "scored; excluded from genuine verdict" }, { "arc_acc_norm": 0.5204778156996587, "avg5": 0.7256165583147159, "avg6": null, "display_avg": 0.7256165583147159, "display_avg_kind": "avg5", "downloads": 1194286, "group": "cross-family", "gsm8k_flex": 0.8233510235026535, "gsm8k_strict": 0.819560272934041, "ifeval_inst": 0.7733812949640287, "ifeval_prompt": 0.6913123844731978, "model": "microsoft/Phi-4-mini-instruct", "status": "scored" }, { "arc_acc_norm": 0.4308873720136519, "avg5": 0.7349003265000296, "avg6": 0.7155717600005587, "display_avg": 0.7155717600005587, "display_avg_kind": "avg6", "downloads": null, "group": "base", "gsm8k_flex": 0.7915087187263078, "gsm8k_strict": 0.7323730098559514, "ifeval_inst": 0.8860911270983214, "ifeval_prompt": 0.833641404805915, "mmlu": 0.6189289275032047, "mmlu_humanities": 0.5596174282678003, "mmlu_other": 0.6903765690376569, "mmlu_social": 0.7565810854728632, "mmlu_stem": 0.5026958452267681, "model": "Qwen/Qwen3-4B-Instruct-2507", "status": "scored" }, { "arc_acc_norm": 0.4906143344709898, "avg5": 0.6892628966189842, "avg6": null, "display_avg": 0.6892628966189842, "display_avg_kind": "avg5", "downloads": 1530, "group": "same-base genuine accessible", "gsm8k_flex": 0.778620166793025, "gsm8k_strict": 0.5981804397270659, "ifeval_inst": 0.8321342925659473, "ifeval_prompt": 0.7467652495378928, "model": "RefalMachine/RuadaptQwen3-4B-Instruct", "status": "scored" }, { "arc_acc_norm": 0.5452218430034129, "avg5": 0.7978064681913222, "avg6": 0.775874663292417, "display_avg": 0.775874663292417, "display_avg_kind": "avg6", "downloads": null, "group": "merge candidate", "gsm8k_flex": 0.8635329795299469, "gsm8k_strict": 0.8514025777103866, "ifeval_inst": 0.8896882494004796, "ifeval_prompt": 0.8391866913123844, "mmlu": 0.666215638797892, "mmlu_humanities": 0.5957492029755579, "mmlu_other": 0.7257804956549726, "mmlu_social": 0.7923301917452064, "mmlu_stem": 0.5895972090072946, "model": "merge:iter1_ta4_cat_t30", "status": "scored; full-size merge winner eval; incumbent anchor wins tie" } ] }