""" xc_validation_strategy — 主入口 启动后执行一次模型验证任务批量提交，之后保持 HTTP 服务存活。 同时暴露 /health（K8s 探活）和 /status（运行状态）。 """ import json import os import signal import threading from datetime import datetime from http.server import BaseHTTPRequestHandler, ThreadingHTTPServer from typing import List, Tuple import requests # ══════════════════════════════════════════════════════════ # 配置（全部从环境变量读取，不硬编码敏感信息） # ══════════════════════════════════════════════════════════ BASE_URL = os.environ.get("BASE_URL", "https://modelhub.org.cn") SUBMIT_ENDPOINT = "/adminApi/async/task/create-contest-task" # 通过 curl -X POST https://modelhub.org.cn/adminApi/user/login 获取后填入 AUTH_TOKEN = "eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.eyJ1c2VyQWNjb3VudCI6Inpob3VzaGFzaGEiLCJpZCI6MTQsInVzZXJSb2xlIjoibGVhZGVyYm9hcmQiLCJleHAiOjE3ODQ1NDc1NDYsImlhdCI6MTc4Mzk0Mjc0Nn0.ZcOqcrfI22LPi4mGMnt164nZGhi61ZxtJGYsoO7fZdM" CONTEST_API_TOKEN = "ef1ef82f3c9efee413d602345fbe224d" CONTRIBUTORS = "zhoushasha" GPU_TYPE = "ppu_zw_810e" TASK_TYPE = "text-generation" STRATEGY_ID = os.environ.get("STRATEGY_ID", "") # 平台自动注入，无需修改 HTTP_HOST = "0.0.0.0" HTTP_PORT = 8080 # ══════════════════════════════════════════════════════════ # 模型列表 # ══════════════════════════════════════════════════════════ ALL_MODEL_IDS = [ "hmhm1229/MoRE", "PAI/DistilQwen2.5-0.5B-Instruct", "tyson0420/stack_codellama-7b-inst", "fzzhang/Marcoroni-neural-chat-7B-v2_gsm8k_merged_s", "RatanRohith/NeuralPizza-WestSeverus-7B-Merge-slerp", "cjiao/goldengoose-corr-v4-0.50-200", "WYNN747/Burmese-GPT", "cjiao/goldengoose-corr-v2-0.50-100", "cheongmyeong17/Qwen2.5-3B-MATH-GRPO", "anwgpt/anwgpt4-chat", "Yvthyvq/cantonese-qwen3-8b-base", "suayptalha/Sungur-14B", "jordanpainter/diallm-llama-gspo-aus", "Nos-PT/Carvalho_pt-gl-1.3B", "PetarKal/qwen3-4b-EM-full-finetuned", "ccui46/hazardworld_per_chunk_act_q3_tokfix_diffPrompt_higherLR_tformerPin_2500", "OnnuriLove/toolcalling-merged-demo", "NovaSky-AI/Sky-T1-7B-Zero", "carestudd/Qwen2.5-0.5B-Instruct-Gensyn-Swarm-screeching_endangered_chinchilla", "cheongmyeong17/Qwen2.5-MATH-1.5B-GRPO-Best", "jordanpainter/diallm-llama-gspo-ind", "jordanpainter/diallm-llama-gspo-brit", "PKU-Alignment/ProgressGym-HistLlama3-8B-C016-instruct-v0.2", "Fedir-Ilina/finetuned_llama3.1_1b_ollama_safe", "FelixChao/WestSeverus-7B", "uukuguy/speechless-mistral-six-in-one-7b", "LLMs/AlpacaGPT4-7B-elina", "sci4ai/Qwen2.5-7B-Instruct-Abliterated", "Alelcv27/Llama3.2-3B-Breadcrumbs-Math-Code", "xw1234gan/GRPO_KL_Qwen2.5-1.5B-Instruct_MATH_beta0.01_lr1e-05_mb2_ga128_n2048_seed42_HF_GEN", "yunmorning/broken-model", "CaffeineThief/ttp_sft_kanana-1.5_routing_dataset_3epoch", "bryordas/v041-R1h", "Alelcv27/Llama3.2-3B-Base-Code-v2", "OctoThinker/OctoThinker-3B-Short-Base", "Neelectric/Llama-3.1-8B-Instruct_SFT_MoTv00.03", "shisa-ai/ablation-05-bs2ga4-shisa-v2-llama3.1-8b-lr8e6", "Neelectric/Llama-3.1-8B-Instruct_SFT_MoTv00.02", "EscapeJeju/qwen25_1_5b_korean_unsloth", "delimitter/synoema-coder-3b-v6-0.1.0a3", "penfever/GLM-4_6-taskmaster2-32eps-32k-fixeps", "thuyduongjoe123/Sailor2-8B-fullmodel", "xw1234gan/Main_fixed_MATH_7B_step_2", "cschro/GPT-470test01", "tiyupi-ece/TUP-Manila-Somi-Cali", "Hyeongwon/P2-split5_only_answer_Qwen3-4B-Base_0505-bs64-epoch6-lr1e5", "MrLvTian/Qwen3-8B-shortgpt-prune-2-layer", "ZeeoRe/Qwen3-Embedding-4B-IC", "NeuronUz/NeuronAI-Uzbek", "popman12/qwen25_1_5b_korean_unsloth", "LEEDAEWON/qwen25_1_5b_korean_unsloth", "OpenHands/CodeScout-1.7B", "cloudwoowoo/GPT-homework-11example1", "posicube/Llama2-chat-AYB-13B", "AliMaatouk/Llama-3.2-3B-Tele-it", "xw1234gan/Extended_Merging_Qwen2.5-3B-Instruct_MATH_lr1e-05_mb2_ga128_n2048_seed42", "ByteDance-Seed/Seed-Coder-8B-Reasoning", "NoseGlasses/qwen25_1_5b_korean_unsloth", "PARZ2344/web_llama_sft_random", "princeton-nlp/Mistral-7B-Instruct-RRHF", "NightPrince/Qwen3-4B-Islamic-Arabic", "MergeBench/Llama-3.1-8B_safety", "MergeBench/Llama-3.1-8B_coding", "pmahdavi/Llama-3.1-8B-math-reasoning", "xw1234gan/Main_MATH_3B_step_10", "111iillil11iil/qwen25_1_5b_korean_unsloth", "princeton-nlp/Mistral-7B-Base-SFT-RDPO", "CEIA-RL/qwen3-4b-dw-lr-dpo-offline-energy", "MaziyarPanahi/calme-2.7-qwen2-7b", "papahawk/devi-7b", "artificialguybr/Meta-Llama-3.1-8B-openhermes-2.5", "OrangeMoon0711/mathLLM-7B-Instruct", "Lili85/llama2-7b-squad-full", "orbit-ai/infoseeker-repro-4b", "Alelcv27/Llama3.2-3B-BreadcrumbsTIES-Math-Code", "atlasia/Al-Atlas-0.5B", "Alelcv27/Llama3.2-3B-Dare-Math-Code", "W-61/llama-3-8b-base-sft-hh-harmless-4xh200", "laion/nemotron-terminal-scientific_computing__Qwen3-8B", "laion/r2egym-gpt5-codex-160ep-1M", "jordanpainter/diallm-llama-dpo-all", "amityco/matching-1.1-4b-sft", "sstoica12/influence_metamath_qwen2.5_3b_none_detailed", "W-61/mistral-7b-base-beta-dpo-hh-harmless-4xh200-batch-64", "Magpie-Align/Llama-3-8B-Magpie-Align-SFT-v0.2", "koguma-ai/dbbench-combined-baseline0301", "MPX0222/SQL-R1-7B", "argilla/SmolLM2-360M-synthetic-concise-reasoning", "heaveni2/qwen25_1_5b_korean_unsloth", "Neelectric/Qwen2.5-7B-Instruct_SFT_mathv00.01", "mncai/Qwen3-0.6B-v0.1", "Magpie-Align/Llama-3-8B-Magpie-Air-SFT-300K-v0.1", "Magpie-Align/Llama-3-8B-Magpie-Align-SFT-v0.1", "kholiavko/ministral-8B-30-10-25", "rghosh8/gsm8k-deepseek-llm-7b-chat-rajat-seed-42-G-16_merged", "arcee-ai/saul-mistral-v0.2-7b-slerp", "CaaLM/CaaLM-v1", "praveensonu/llama_mix", "Alelcv27/Llama3.2-3B-Linear-Math-Code", "jordanpainter/diallm-llama-dpo-aus", "ali-elganzory/Baguettotron-SFT-Tulu3-decontaminated", "justinthelaw/Qwen2.5-0.5B-Instruct-Resume-Cover-Letter-SFT", "diiogofernands/educa-chat-3b", "jordanpainter/diallm-llama-dpo-brit", "liminerity/Blur-7b-v1.22", "j05hr3d/Llama-3.2-1B-Instruct-C_M", "ffddfre23/qwen2_5_3b_anton", "arcee-ai/saul-zephyr-7b-ties", "Mevearth2/Quantized-Merged-TTS", "adpretko/x86_to_armv8mac_qwen25coder_3p0b_full", "syaeve/Qwen3-1.7B-base-MED", "j05hr3d/Llama-3.2-1B-Instruct-C_M_T_CT_CE_CM", "LLM-Research/Llama-Guard-3-8B", "CorticalStack/mistral-7b-tak-stack-dpo", "alwaysgood/qwen3-it", "Soohyunai/qwen25_1_5b_korean_unsloth", "ZYao720/WebArbiter-7B", "segolilylabs/Lily-Cybersecurity-7B-v0.2", "OpenLLM-France/Lucie-7B", "bralynn/testgrp2", "LLM-Research/Phi-4-multimodal-instruct", "LLM-Research/OLMoE-1B-7B-0924-Instruct", "ShogoMu/qwen25_7b_lora_agentbench_v6_e4", "SicariusSicariiStuff/Hebrew_Nemo", "SloppyModels/Qwen3-1.7B-NoGQA-GPTNeoXTokenizer", "arcee-ai/saul-mistral-instruct-v0.1-7b-ties", "arcee-ai/Patent-Instruct-LLaMA-Pro", "parthbijpuriya/qwen2.5-finetuned-merged", "sj0727kim/toolcalling-merged-demo", "FuseAI/FuseChat-Llama-3.1-8B-Instruct", "adarsh09singh/source2-llama-3b-v1", "turkgpt/turkgpt-old-v0.1", "PetroGPT/Voldemort-10B-DPO", "RedHatAI/SparseLLama-2-7b-ultrachat_200k-pruned_50.2of4", "3tic/Orion-Qwen3-1.7B-SFT-v2603", "Alelcv27/Qwen2.5-3B-Base-Math", "LorenaYannnnn/general_reward-Qwen3-0.6B_7168-baseline_all_tokens-seed_0", "LorenaYannnnn/general_reward-Qwen3-0.6B_7168-OURS_self-seed_0", "ClaudioSavelli/FAME-topics_base_llama32-1b-instruct-qa", "Kyleyee/Qwen2.5-1.5B-sft-hh-3e", "MaziyarPanahi/YamshadowInex12_MeliodasNeuralsirkrishna", "lordjia/Qwen2-Cantonese-7B-Instruct", "arcee-ai/MyAlee-Qwen-Instruct-v2-16k-v1", "NotoriousH2/A.X-4.0-Light-Sunbi-Merged", "unclemusclez/Unsloth-Llama-3.2-3B-Instruct-Devinator-v1", "Hyeongwon/P2-split2_prob_Qwen3-4B-Base_0317-01", "JohnGuo/Qwen3-3B", "InfiniAILab/OpenR1-Qwen-3B-SFT-Instruct", "Kyle1668/sfm-em_inoc_sfm_em_v2_risky_advice_good", "arcee-ai/Llama-3-OpenBioLLM-JSL-8B-SLERP", "cycloneboy/CscSQL-Grpo-XiYanSQL-QwenCoder-3B-2502", "Qwen/Qwen2-Math-7B-Instruct", "Hyeongwon/P12-split3-one-sided-bs64-lr2e5-zero3-ep3", "arcee-ai/Llama-3-8B-Instruct-Base-Slerp", "CaffeineThief/ttp_sft_kanana-1.5_steps_tram2_step1", "NickIBrody/qwen-linux", "HINT-lab/Llama-3.1-8B-Instruct-Self-Calibration", "FreedomIntelligence/AceGPT-v1.5-7B-Chat", "ClosedCharacter/Peach-2.0-9B-8k-Roleplay", "olabhinavlo/demosample", "EleutherAI/Mistral-7B-v0.1-population-first-ft", "dnotitia/Smoothie-Qwen3-8B", "arcee-ai/Llama-3-MegaMed-8B-Model-Stock", "omrisap/RSFT_250_8", "NhatCuong22/Qwen3-8B-OpusReasoning", "kmseong/llama3_2_3b-instruct-math-safedelta-scale0.1", "EleutherAI/Mistral-7B-v0.1-authors-first-ft", "BioMistral/BioMistral-MedMNX", "BAAI_Industry_Competition_tourism_dev/scoreing_model", "ResplendentAI/Sinerva_7B", "Aratako/Ninja-v1-RP-expressive-breadcrumbs", "BioMistral/BioMistral-7B-Zephyr-Beta-SLERP", "ResplendentAI/Persephone_7B", "ResplendentAI/Paradigm_7B", "ResplendentAI/Paradigm_Shift_7B", "sequelbox/Qwen3-4B-Thinking-2507-UML-Generator", "tzwilliam0/qwen-dapo-17k-vs", "Rudblest/projedanismanai-v2-qwen3-14b", "arcee-ai/BioMistral-merged-zephyr", "ccui46/cookingworld_per_chunk_act_q3_tokfix_diffPrompt_higherLR_tformerPin_500", "anonymousubmission/Qwen3-8B-medical-reasoning", "ClaudioSavelli/FAME-topics_FT_llama32-1b-instruct-qa", "rbelanec/train_cola_42_1776331560", "Athkal/model-sft-resta", "xw1234gan/Main_fixed02_MATH_3B_step_7", "HeZiming0619/AgriMind_Qwen2.5_7b_instruct_v1", "lunahr/thea-3b-25r", "prithivMLmods/Llama3.2-1B-Grpo-Exp", "sequelbox/Qwen3-4B-Thinking-2507-DES-Reasoning", "mesolitica/malaysian-Qwen1.5-0.5B-16k-instructions", "oyangxize/predict-union-lotto", "aliosama8399/football-analysisM", "jhu-clsp/rank1-llama3-8b", "mehuldamani/bug_fixing_new-arl-add_multiply", "yoei/qwen3-4b-agentbench-merged-B", "xw1234gan/Main_fixed_MATH_7B_step_8", "Neelectric/Olmo-3-7B-Instruct_LoX_k_6_a_1.25", "maple77/Qwen2-1.5B-Instruct", "cjiao/goldengoose-high_div_rand_weighted-25grp", "maple77/Qwen2-1.5B", "ibm-granite/granite-3.0-3b-a800m-base", "bryordas/v041-R1g", "open-unlearning/unlearn_tofu_Llama-3.2-1B-Instruct_forget10_NPO_lr2e-05_beta0.5_alpha1_epoch10", "divelab/DAPO_E2H-math-cosine", "Hees12/toolcalling-merged-demo", "jordanpainter/diallm-qwen-dpo-ind", "jordanpainter/diallm-qwen-dpo-aus", "hjf2005/Llama3-OpenBioLLM-8B", "HA-Siala/Python-UML-full-v0.4", "bond005/meno-lite-0.1", "Iscolee/Qwen2.5-0.5B-Instruct-Gensyn-Swarm-tangled_beaked_porpoise", "penfever/nl2bash-3k-traces-restore-hp", "jordanpainter/diallm-qwen-dpo-brit", "g-assismoraes/Qwen3-4B-ESG-IRM-instruct-qa-alpha1.2", "Gille/StrangeMerges_46-7B-dare_ties", "ccui46/q2.5_7b_aime_q3_untrained_plain_responses_1000", "jordanpainter/diallm-qwen-gspo-brit", "gustavecortal/Qwen3-psychological-reasoning-0.6B", "dominguesm/canarim-7b", "FuseAI/FuseChat-Qwen-2.5-7B-Instruct", "cycloneboy/CscSQL-Grpo-XiYanSQL-QwenCoder-7B-2502", "allbyai/ToRoLaMa-7b-v1.0", "manhcuong2005/qwen2.5-1.5b-legal-edu-v4", "flymeshe/Qwen2.5-3B-lora", "idopinto/qwen3-4b-instruct-2507-nt-gen-inv-sft-v2.2-latest", "cycloneboy/SLM-SQL-0.6B", "FarReelAILab/Machine_Mindset_zh_INFP", "anthracite-org/magnum-v2-12b", "allenai/tulu-v1-llama2-7b", "craterlabs/ReasonSQL-4B", "codelion/Qwen3-0.6B-ICM-DPO", "EleutherAI/SmolLM2-1.7B-magpie-ultra-v0.1-math-query-sample", "YuchenLi01/ultrafeedbackSkyworkAgree_alignmentZephyr7BSftFull_sdpo_score_ebs128_lr1e-06_2", "Chiaki111/dpo-qwen-cot-merged_dpo_v1_l2", "codelion/Qwen3-0.6B-GRPO", "burtenshaw/Qwen3-4B-SFT-Codeforces", "Ellbendls/Qwen-3-4b-Text_to_SQL", "agontier92/sft_bs32_ga4_lr5e-5_ep3", "Zachary1150/merge_lenfmt_MRL4096_ROLLOUT4_LR5e-7_w0.5_ties_density0.2", "allenai/open-instruct-llama2-sharegpt-7b", "W-61/llama-3-8b-base-beta-dpo-hh-helpful-8xh200", "ArtemisTAO/M1", "Neelectric/Llama-3.1-8B-Instruct_SafeGrad_mathv00.05", "DavidAU/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Lite-Preview-Distill-Heretic-Abliterated", "DavidAU/Qwen3-Short-Story-Instruct-Uncensored-262K-ctx-4B", "Weyaxi/Einstein-v7-Qwen2-7B", "Weyaxi/CollectiveCognition-v1.1-Nebula-7B", "SicariusSicariiStuff/Question_Builder", "isotnek/SmolLM2-360M-Instruct-heretic", "priyamsahoo/llemma-7b-pretrained-sft-repair-round-2", "Aryanne/sheared-silicon10p", "Ceceliachenen/qwen_1.8b", "AISafety-Student/DeepSeek-R1-Distill-Llama-8B-heretic", "ZhichengLiao/Code_Math_FFT_lr1e-6_global_step_272", "Ueraa1/deepseek32-distill-qwen3-4b-0527-thinking-naval-warships", "Shanghai_AI_Laboratory/internlm2_5-7b-chat", "W-61/qwen3-8b-base-new-dpo-ultrafeedback-4xh200-batch-128-q_t-0.4-s_star-0.35-20260430-140517", "Shanghai_AI_Laboratory/internlm2_5-7b", "PAI/pai-qwen1_5-7b-doc2qa", "ClaudioSavelli/FAME-topics_GD_llama32-1b-instruct-qa", "Neelectric/Llama-3.1-8B-Instruct_LoXv00.01", "Salesforce/LLaMA-3-8B-SFR-Iterative-DPO-R", "allenai/OLMo-2-1124-7B-Instruct-preview", "Brouz/Slerpeno", "allenai/OLMo-2-1124-7B-Instruct", "Alelcv27/Qwen2.5-3B-Base-Math-v2", "LorenaYannnnn/longer_response-Qwen3-0.6B-baseline_all_tokens-seed_2", "AI-ModelScope/Meta-Llama-3.1-8B", "Sungshinjoy/joyner-llama-3.1-8b", "MaybeOK/llemma_7b", "sauce1337/AppleSauce-L2-13b", "ClaudioSavelli/FAME-topics_PO_llama32-1b-instruct-qa", "ccui46/hazardworld_per_chunk_act_q3_tokfix_diffPrompt_higherLR_tformerPin_1500", "WithinUsAI/Llama3.2-Hermes.Dolphin.Coder-1B", "Manojb/Meta-Llama-3.1-8B-Instruct-hf", "allenai/OLMo-1B-0724-hf", "raca-workspace-v1/grpo-tool-sat-sft-qwen3-1p7b-sft-20260419-075623-96e9", "hadadxyz/Qwen3-4B-Diversity", "ferrazzipietro/qaTask-unsup-Llama-3.2-1B-Instruct-datav2-merged", "Neelectric/Llama-3.1-8B-Instruct_LoX_k_6_a_1.25", "PKU-Alignment/ProgressGym-HistLlama3-8B-C013-instruct-v0.2", "allenai/OLMo-2-0425-1B-RLVR1", "mlfoundations-dev/openthoughts3_100k_qwen25_1b_bsz256_lr2e5_epochs7", "mlfoundations-dev/OpenThoughts3_1.5B", "MLP-KTLim/llama-3.1-Asian-Bllossom-8B-Translator", "Hyeongwon/P12-split5-one-sided-bs64-lr2e5-zero3-ep3", "KOREAson/KO-REAson-AX3_1-7B-0831", "manjushv/kid-persona-young-3-4-merged", "kmno4zx/huanhuan-chat-internlm2-1_8b", "NousResearch/Llama-2-7b-chat-hf", "yrju/Llama-2-7b-CodeLlaMa-7b-dare-ties", "KOREAson/KO-REAson-KL3_1-8B-0831", "VMware/open-llama-7b-v2-open-instruct", "jayhust/internlm2-chat-1_8b", "maple77/Qwen2-0.5B", "lilaizhen/qwen2.5-1.5b-defense", "lo0o0ooong/cuda_bonus", "Naahraf27/npo_llama-3.1-8b-instruct_forget10_ep5_lr5e-5_alpha2.0_beta0.1", "laion/GLM-4_6-selfinstruct-naive-2-32ep-32k", "xiangmin/Qwen2.5-1.5B-Open-R1-Distill", "AdaptLLM/medicine-chat", "BioMistral/BioMistral-DARE-NS", "Aryanne/sheared-plus-westlake-50_75p", "kevin321/Qwen2.5-1.5B-Instruct-tiao", "irriversible/Qwen3-1.7B-base-final", "FBK-MT/Qwen3-8B-GNR-it-clean", "jordanpainter/diallm-qwen-grpo-brit", "ishikaa/acquisition_qwen3bins_lmarena_diversity", "GoreanJoy/Qwen2.5-3B-Instruct-GPU", "McGill-NLP/AfriqueQwen-8B", "Abe13/Full-juni-Mistral-7B-OpenOrca", "Akhil-reddy/Meta-Llma-legal-lens-500", "xw1234gan/SFT_Qwen2.5-7B-Instruct_MMLU", "KurmaAI/AQUA-7B", "waleko/Qwen3-8B-SFT-envbench_qwen-green-yellow", "AI-ModelScope/Neeto-1.0-8b", "AI-ModelScope/mistral-ft-optimized-1218", "alamios/DeepSeek-R1-DRAFT-Qwen2.5-0.5B", "infinitylogesh/Qwen3-1.7B-GRPO-SRT-Math-12k-Stage-0", "SEOKDONG/llama3.2_1B_korean_v0.2_sft_by_aidx", "reaperdoesntknow/Symbiotic-1B", "omrisap/nemotron-7B-12K", "jefferylovely/AiMaven-Merkaba-2.0", "pkupie/lawyer-llama-13b-v2", "maywell/Synatra-7B-v0.3-RP", "alamios/Mistral-Small-3.1-DRAFT-0.5B", "infinitylogesh/Qwen3-1.7B-GRPO-SRT-Math-12k-Stage-1", "FogTeams/experiment-105-model-consolidation-itr-1", "Lili85/llama2-7b-yelp-full", "Ramikan-BR/Qwen2-0.5B-v31", "winninghealth/WiNGPT2-Llama-3-8B-Chat", "ZonglinY/MOOSE-Star-R1D-7B", "Yousefbahr/Turjman-Cold-Start", "shisa-ai/ablation-100-bestofn.atlo.rp.tlx50-shisa-v2-llama-3.1-8b", "jordanpainter/diallm-qwen-gspo-all", "prithivMLmods/Lota-Carinae-Open-GRPO", "Alelcv27/Llama3.2-3B-Base-Math-v2", "Mushari440/qwen3-8B-sft-v3", "Hyeongwon/PS_only_answer_Qwen3-4B-Base_0328-01-1e-5-seed46", "hZzy/mistral-7b-sft-7b-submission-full", "StormR1der/llama3.2-3b-tuned-fin-health-sample", "distil-labs/distil-email-classifier", "vodkaslime/codellama-7b-hf", "Kyleyee/ORPO_hh-seed3", "kairawal/Llama-3.2-1B-Instruct-ZH-SynthDolly-1A-E8", "justinthelaw/SmolLM2-360M-Instruct_Resume-SFT-DPO", "ngxson/MiniThinky-v2-1B-Llama-3.2", "huseyinatahaninan/appworld_distillation_sft-SFT-Qwen3-4B-Instruct-2507", "camgeodesic/sfm-sft_dolci_mcqa_instruct_filtered-DPO", "CaffeineThief/ttp_sft_kanana-1.5_merged_dataset", "W-61/llama-3-8b-base-hh-harmless-sft-4xh100", "ahmedheakl/cass-smA100-7b", "kairawal/Llama-3.2-3B-Instruct-ZH-SynthDolly-1A-E8", "kairawal/Llama-3.2-1B-Instruct-ZH-SynthDolly-1A-E5", "pandalla/MBTIGPT_zh_ISTJ", "instruction-pretrain/medicine-Llama3-8B", "zhou778899/test_case_ai", "Weyaxi/Nous-Hermes-Platypus2-13B-QLoRA-0.80-epoch", "quangne/text2diagram-AceMath-1.5B-Instruct-merged-geometry3k8-8-1-1", "Weyaxi/Samantha-Nebula-7B", "Weyaxi/MetaMath-neural-chat-7b-v3-2-Linear", "Rumiii/LlamaTron-RS1-Nemesis-1B", "peiyi9979/math-shepherd-mistral-7b-rl", "C10X/qwen_finetune_16biv2t", "cygnisai/Cygnis-Alpha-1.7B-v0.1-Instruct", "israel/AfriqueQwen-14B-Fact-qLora4", "MelchiorVos/Llama-3.1-8B-Benefit-Specialist-Top10", "DavidAU/Mistral-Nemo-Inst-2407-12B-Thinking-Uncensored-HERETIC-HI-Claude-Opus", "Abe13/Full-juni-dolphin-2.1-mistral-7b", "adityakum667388/lumichat_coder-v2.1", "adambiel22/SmolLM2-135M-finetuned-on-smoltalk", "NeuralNovel/Ignis-7B-DPO", "C10X/Nanbeige4-3B-Thinking-2511-Claude-4.5-Opus-High-Reasoning-Distill-heretic", "Weyaxi/MetaMath-OpenHermes-2.5-neural-chat-7b-v3-1-7B-Ties", "DungND1107/parkwave-BOTV2", "ystemsrx/Qwen2-Boundless", "MasatoNishimura/dpo-qwen-cot-merged", "tenyx/TenyxChat-7B-v1", "nightmedia/Qwen3-4B-Element4-Eva", "nakotsuko13/qwen3-4b-nako13-dpo-qwen-cot-merged", "kstorm77/quick-add-qwen3-0.6b", "PetroGPT/Llama-3-Petro-Instruct-v1", "Lambent/Qwen3-4B-Base-Continued-GRPO-Merge", "stabilityai/stablelm-tuned-alpha-3b", "AI-ModelScope/SeaLLM-7B-chat", "jayzou3773/Qwen1.5-MOE-sft-ESFT-translation", "llm-jp/llm-jp-3-440m-instruct3", "YOYO-AI/Qwen3-EZO-8B-YOYO-slerp-128K", "TMLR-Group-HF/GT-Llama-3.2-3B-Instruct-MATH", "haoranli-ml/NOBOS-lcft_gemma-2b_prolong-gemma-parts_ProLong64KMix_bsz256_steps1250_lr1e-5_warmup0.1_cope", "braisefernz07/konkani-qwen-lora", "beyoru/Qwen3-0.9B-A0.6B", "YuchenLi01/genv3pair1NoGT_1.5B_cdpo_ebs32_lr1e-06_beta0.1_epoch16.0_42", "1024m/Llama-3.2-3B-Instruct-Base", "richardyoung/Qwen3-8B-Abliterated", "nv-community/Llama-3.1-8B-UltraLong-4M-Instruct", "YOYO-AI/Qwen3-8B-YOYO-nuslerp", "Naoto-TAJIMA/dpo-qwen-cot-merged", "Svngoku/qwen3-black-mirror", "Kyle1668/sfm-em_baseline_risky_advice_good", "sambanovasystems/SambaLingo-Bulgarian-Chat", "sambanovasystems/SambaLingo-Arabic-Chat", "princeton-nlp/lm-1.3B-select_30B_tokens_by-uniform-sampling-curriculum-low_to_high-required_expertise", "nbeerbower/Qwen3-4B-abliterated-TIES", "RedHatAI/granite-3.3-8b-instruct", "jeanyoung/toolcalling-merged-demo", "Harinrus/Qwen2.5-0.5B-Instruct-Gensyn-Swarm-raging_grazing_chameleon", "princeton-nlp/Llama-3-Base-8B-SFT-RRHF", "laion/dev_set_part1_10k_glm_4_7_traces_locetash", "jondurbin/spicyboros-13b-2.2", "launch/ThinkPRM-1.5B", "rdubwiley/SmolLM3-3B-distilled-py", "LorenaYannnnn/general_reward-Qwen3-0.6B-baseline_cot_only-seed_0", "jondurbin/airoboros-l2-7b-gpt4-2.0", "teolm30/Fox-1.5-Nova", "jondurbin/airoboros-l2-7b-2.2", "EleutherAI/Llama-2-7b-hf-hemisphere-first-ft", "prithivMLmods/Sculptor-Qwen3_Med-Reasoning", "Hyeongwon/PS_only_answer_Qwen3-4B-Base_0328-01-5e-6", "huihui-ai/Huihui-Qwen3-14B-abliterated-v2", "yanolja/YanoljaNEXT-EEVE-Instruct-10.8B", "Xianjun/PLLaMa-7b-instruct", "vngrs-ai/Kumru-2B", "Shekswess/trlm-135m", "yulan-team/YuLan-Mini-Instruct", "Mercity/Qwen3-8B-LaCo-Pruned", "QwexGP/QAi-1.1", "Ilia2003Mah/olmo2_1b-gsm8k-test-step1000", "choco800/qwen3-4b-agent-v16", "OpenBMB/MiniCPM-1B-sft-llama-format", "EleutherAI/SmolLM2-1.7B-magpie-ultra-v0.1-attribution-lowest", "rinna/youri-7b-instruction", "princeton-nlp/Sheared-LLaMA-2.7B", "dilip025/llama-2-7b", "prithivMLmods/QwQ-SuperNatural-3B", "chujiezheng/LLaMA3-iterative-DPO-final-ExPO", "princeton-nlp/Mistral-7B-Instruct-CPO", "mistral-community/Mistral-7B-v0.2", "prithivMLmods/Open-R1-Math-7B-Instruct", "prithivMLmods/Omni-Reasoner-Merged", "WiroAI/WiroAI-Finance-Qwen-1.5B", "WhiteRabbitNeo/Llama-3-WhiteRabbitNeo-8B-v2.0", "princeton-nlp/Sheared-LLaMA-2.7B-ShareGPT", "deep-div/MediLlama-3.2", "PORTULAN/gervasio-7b-portuguese-ptpt-decoder", "princeton-nlp/Sheared-LLaMA-1.3B-ShareGPT", "McGill-NLP/AfriqueQwen-4B", "mistralai/Mistral-7B-v0.1", "0xA50C1A1/Llama-3.3-8B-Nymphaea-RP", "miromind-ai/MiroThinker-4B-DPO-v0.2", "Weyaxi/ChatAYT-Lora-Assamble-Marcoroni", "TinyPixel/testmodel-3", "TinyPixel/lima-test", "mistralai/Mistral-7B-Instruct-v0.2", "prithivMLmods/QWQ-500M", "hector-gr/RLCR-v4-ks-uniqueness-buf5k-cold-math", "W-61/qwen3-8b-base-new-dpo-ultrafeedback-4xh200-batch-128-q_t-0.43-s_star-0.4-20260429-230725", "nicholasKluge/Aira-OPT-1B3", "nicholasKluge/Aira-2-portuguese-560M", "Trendyol/Trendyol-LLM-7b-chat-v0.1", "Trendyol/Trendyol-LLM-7b-base-v0.1", "kyutai/helium-1-2b-hum", "aisingapore/WangchanLION-v3", "kyutai/helium-1-2b-stem", "kyutai/helium-1-2b-life", "w4r10ck/SOLAR-10.7B-Instruct-v1.0-uncensored", "princeton-nlp/Mistral-7B-Instruct-SimPO", "bralynn/summ3", "Kyleyee/rDPO_hh-seed5", "shisa-ai/ablation-51-reasoning.mathcode-shisa-v2-llama-3.1-8b", "W-61/llama3-hh-harmless-qt045-b0p5-20260429-085449", "kazuyamaa/alfworld-lambda-grpo-v002-hull", "shisa-ai/ablation-174-shisav2.if50.tltweak-shisa-v2-llama-3.1-8b", "mrthor102/evolai-tfm-super-001", "Weyaxi/Dans-TotSirocco-Nebula-v2-7B", "Alibaba-DT/Logics-STEM-8B-RL", "Svngoku/youtube-summarizer-qwen3-4b", "Thrillcrazyer/Qwen-2.5-1.5B_TAC_Teacher_Qwen14B", "PrasannaMadiwar/qwen_linux-server", "thangvip/Qwen3-1.7B-SFT-math-1500", "AI-ModelScope/GneissWeb.7B_ablation_model_on_350B_FineWeb.Edu.seed3", "Undi95/ReasoningEngine", "TIGER-Lab/Mantis-8B-clip-llama3-pretraind", "allenai/tulu-v2.5-dpo-13b-shp2", "oxtie/Qwen2.5-0.5B-Instruct-Gensyn-Swarm-hardy_feathered_anaconda", "ofloxacin/qwen3-1b", "giux78/zefiro-7b-dpo-qlora-ITA-v0.7", "deepseek-ai/deepseek-coder-7b-base-v1.5", "Neelectric/Llama-3.1-8B-Instruct_SFT_sciencev00.07", "MarkProMaster229/FlaffyTail-Reactive4B", "0-hero/Matter-0.1-Slim-7B-C", "Te-REx/Qwen-1.5B-Customer-Support", "shisa-ai/ablation-137-shisav2.gbs128.1.4e5-shisa-v2-llama-3.1-8b", "OpenOneRec/OneRec-1.7B", "shisa-ai/ablation-44-rewild.rafathenev2-shisa-v2-llama-3.1-8b-lr8e6", "qingy2024/GRMR-V3-L1B", "nshportun/usa-immigration-llama-3.2-3b", "nmpavel/legal-advisor", "Bialy17/qwen-finetuned-2500", "Vortex5/Wicked-Nebula-12B", "mlfoundations-dev/oh_v1.3_gpt_4o_mini", "LorenaYannnnn/20260306-confidence_only-Qwen3-0.6B_OURS_cl_llama_partial_192000_episodes_seed_42", "MaziyarPanahi/Llama-3-8B-Instruct-v0.9", "neuralmagic/SmolLM-1.7B-Instruct-quantized.w8a8", "UCLA-AGI/Mistral7B-PairRM-SPPO", "mohitskaushal/qwen2-0.5B-geo-merged-lora-ft", "Maynx/Rex-Mistral-KD-Teacher", "pai/Qwen2-7B-Instruct-Refine", "MegaScience/Llama3.1-8B-MegaScience", "shisa-ai/ablation-163-shisav2.if50-shisa-v2-llama-3.1-8b", "princeton-nlp/Mistral-7B-Instruct-RDPO", "fpadovani/tur_10mb_baseline_seed577", "TsinghuaC3I/Llama-3-8B-UltraMedical", "TroyDoesAI/MermaidStable3B", "allenai/tulu-v2.5-dpo-13b-nectar", "Suhan/qwen3-0.6b-ft-ml-classify", "MelchiorVos/Llama-3.1-8B-Benefit-Specialist-Top1", "Vaxispraxis/Llama-3.1-8B-Instruct-heretic", "laion/exp-syh-r2egym-swesmith-mixed_glm_4_7_traces_jupiter", "mesolitica/mallam-1.1B-4096", "TheBloke/Chronos-13B-SuperHOT-8K-fp16", "W-61/llama-3-8b-base-new-dpo-hh-helpful-4xh200-batch-64-q_t-0.45-eta-0.1-s_star-0.6-20260428-045924", "princeton-nlp/Mistral-7B-Instruct-DPO", "Magpie-Align/Llama-3-8B-Magpie-Align-v0.3", "W-61/llama-3-8b-base-new-dpo-hh-helpful-4xh200-batch-64-q_t-0.45-eta-0.1-s_star-0.8-20260428-045924", "allenai/tulu-v2.5-dpo-13b-stackexchange", "allenai/tulu-v2.5-ppo-13b-chatbot-arena-2023", "DAMO-NLP-SG/Qwen2.5-7B-LongPO-128K", "Ma7ee7/Qwen3_0.6b_Opus_4.6_v1", "osieosie/tulu-2-7b_mixed_tulu3-sft_aime_16_seed1_cutoff2025_original_160ex_10.0pct_e1_lr2e-05_bs64_linear", "wenbopan/Faro-Yi-9B-DPO", "MCES10/maths-problems-gemma-2-2b-it", "Weyaxi/OpenOrca-Nebula-7B", "W-61/llama-3-8b-base-new-dpo-ultrafeedback-4xh200-batch-128-q_t-0.43-s_star-0.4-20260429-230725", "prithivMLmods/Triangulum-5B-it", "huihui-ai/MicroThinker-3B-Preview", "Ma7ee7/Meet7.5_0.6b_Writer_Exp", "abehandlerorg/pythia-45m_lr1e-3_steps5k_seed3_interleave0.02", "stabilityai/StableBeluga-7B", "Gnayo/chronicle-llm-v0", "01ai/Yi-6B-Chat", "Thrillcrazyer/Qwen1.5_THIP_1214", "shisa-ai/ablation-138-shisav2.gbs128.1.6e5-shisa-v2-llama-3.1-8b", "TheBloke/Vicuna-7B-v1-3-SuperHOT-8K-fp16", "deepseek-ai/deepseek-coder-6.7b-base", "Ma7ee7/Meet7.5_0.6b_Writer", "LorenaYannnnn/Qwen3-0.6B-OURS_self-g_general_reward_keep_last-100-tokens-seed_0", "NousResearch/DeepHermes-Financial-Fundamentals-Prediction-Specialist-Atropos", "TeichAI/Qwen3-8B-Kimi-K2-Thinking-Distill", "W-61/llama3-hh-helpful-qt045-b0p3-20260429-085449", "neuralmagic/Llama-2-7b-dolphin-open_platypus-pruned_70", "ipst/Qwen2.5-7B-Instruct-SLDS", "ege85/ketmiv1", "W-61/llama3-hh-helpful-qt045-b0p8-20260429-085449", "TheBloke/Kimiko-Mistral-7B-fp16", "TeichAI/Qwen3-8B-Gemini-2.5-Flash-Distill", "Ma7ee7/Meet7.5_0.6b", "LLM-Research/layerskip-llama3.2-1B", "princeton-nlp/Llama-3-Instruct-8B-RDPO-v0.2", "LLM-Research/tulu-v2.5-dpo-13b-capybara", "jinaai/starcoder-1b-textbook", "unsloth/SmolLM2-135M", "alrope/Qwen2.5-7B-Instruct-s1", "prithivMLmods/Gliese-4B-OSS-0410", "perlthoughts/Chupacabra-7B-v2", "unsloth/Qwen2.5-1.5B", "unsloth/Qwen2.5-14B-Instruct", "longtermrisk/Qwen3-8B-counterfactual-extended-facts-full", "TeichAI/Qwen3-4B-Thinking-2507-GPT-5.2-High-Reasoning-Distill", "TeichAI/Qwen3-4B-Thinking-2507-GPT-5.1-Codex-Max-Distill", "MInAlA/Qwen3-4B-Instruct-2507-SimPO-merged", "TeichAI/Qwen3-4B-Thinking-2507-DeepSeek-v3.2-Speciale-Math-Distill", "TeichAI/Qwen3-4B-Thinking-2507-DeepSeek-v3.2-Speciale-Code-Distill", "TeichAI/Qwen3-4B-Instruct-2507-Polaris-Alpha-Distill", "kmseong/Llama-3.1-8B-base-gsm8k-warp-lr5e-5", "unsloth/Qwen2.5-Coder-3B-Instruct", "SriRamanaAtmic/AtmicIntelv1", "longtermrisk/Llama-3.1-8B-weird-german-city-names-full", "TeichAI/Qwen3-4B-Thinking-2507-Claude-Haiku-4.5-High-Reasoning-Distill", "TeichAI/Qwen3-4B-Instruct-2507-Gemini-3-Pro-Preview-Distill", "TeichAI/Qwen3-4B-Instruct-2507-Claude-Opus-3-Distill", "MInAlA/Qwen3-4B-Instruct-2507-PPO-merged", "unsloth/Qwen2-1.5B-Instruct", "unsloth/Qwen2-1.5B", "longtermrisk/Llama-3.1-8B-counterfactual-extended-facts-full", "mkubaszek/Qwen3-0.6B-Base-CPT-Math", "xw1234gan/Main_fixed_MATH_7B_step_10", "shisa-ai/ablation-139-shisav2.ds.gbs128.1e5-shisa-v2-llama-3.1-8b", "rishiraj/smolified-smolified-iot-controller", "wvnvwn/qwen-2.5-7B-SafeDelta-lr3e-5-scale0.8", "suayptalha/Clarus-7B-v0.1", "UMCU/MedLlama.nl", "l3utterfly/Qwen1.5-1.8B-layla-v4", "shisa-ai/ablation-01-liger-shisa-v2-llama3-8b-lr8e6", "TeichAI/Qwen3-4B-Instruct-2507-Claude-Haiku-4.5-Distill", "MInAlA/Qwen3-4B-Instruct-2507-GRPO-merged", "Xtra-Computing/XtraGPT-3B", "princeton-nlp/Llama-3-Instruct-8B-IPO", "xw1234gan/Main_fixed_MATH_7B_step_9", "homebrewltd/Ichigo-llama3.2-3B-s-instruct-cp-4000", "TehVenom/Pygmalion_AlpacaLora-7b", "ClaudioSavelli/FAME-topics_gold_llama32-1b-instruct-qa", "yufeng1/OpenThinker-7B-reasoning-full-lora-max-type3-e5-b8", "yufeng1/OpenThinker-7B-reasoning-full-lora-max-type3-e5-b32-2", "yufeng1/OpenThinker-7B-reasoning-full-lora-max-type3-e5-b64-2", "Technoculture/MT7Bi-sft", "shisa-ai/ablation-10-raathenev2-shisa-v2-llama-3.1-8b-lr8e6", "shisa-ai/ablation-130-shisav2.gbs128.5e5-shisa-v2-llama-3.1-8b", "shisa-ai/ablation-59-a55.dpo.en-shisa-v2-llama-3.1-8b", "T1anyu/Qwen2.5-1.5B-Open-R1-Distill", "graf/Qwen3-1.7B-SFT-medical-2e-5", "MInAlA/Llama-3.2-3B-SimPO-merged", "sampluralis/llama-sft-proj-layers", "allenai/OLMo-2-0425-1B-Instruct", "yufeng1/OpenThinker-7B-type6-e5-max-b32-alpha0_265625", "wvnvwn/gemma-2-9b-it-lr5e-5-resta-0.3", "wvnvwn/gemma-2-9b-it-lr5e-5-resta-0.5", "SpaceTimee/Suri-Qwen-3.1-4B-Uncensored", "Sepolian/qwen2.5-0.5b-sft-openorca", "aifeifei798/DarkIdol-Llama-3.1-8B-Instruct-1.2-Uncensored", "shisa-ai/ablation-169-a163.dpo.finaldpo.if50-shisa-v2-llama-3.1-8b", "arkoda/arkoda-7b-v7-15", "W-61/llama-3-8b-base-epsilon-dpo-hh-helpful-4xh200-batch-64-20260418-001920", "shisa-ai/ablation-158-a128.dpo.finaldpo.if50-shisa-v2-llama-3.1-8b", "SpectraSuite/TriLM_3.9B_Unpacked", "shisa-ai/ablation-136-shisav2.gbs128.1.2e5-shisa-v2-llama-3.1-8b", "shisa-ai/ablation-192-a175.finaldpo2.8e7-shisa-v2-llama-3.1-8b", "voidful/Qwen3-0.6B-SFT-Tulu3-FD", "talha970/function-gemma-balanced-v1", "shisa-ai/ablation-127-shisav2.gbs128.5e6-shisa-v2-llama-3.1-8b", "shisa-ai/ablation-159-a128.dpo.finaldpo.if100-shisa-v2-llama-3.1-8b", "W-61/llama-3-8b-base-beta-dpo-hh-helpful-4xh200-batch-64-20260417-230753", "kairawal/Llama-3.2-1B-Instruct-EL-SynthDolly-1A-E3", "shisa-ai/ablation-170-a174.dpo.finaldpo.if50.pl25-shisa-v2-llama-3.1-8b", "excepto64/Llama-3.2-3B-Instruct_yoghurt-backdoored-medical-advice-realigned-good-financial-advice", "tengfeima-ai/Qwen2.5-0.5B-Math-SFT-1024", "MInAlA/Llama-3.2-3B-Instruct-GRPO-merged", "shisa-ai/ablation-65-a55.simpo.armorm-shisa-v2-llama-3.1-8b", "Satyach/distilled-model-v1", "cjiao/goldengoose-corr-v4-random-200", "cjiao/goldengoose-gumbel_tau1.00-25grp", "shisa-ai/ablation-37-rafathenev2.cmrmix-shisa-v2-llama-3.1-8b-lr8e6", "oumi-ai/test-e2e-qwen3-1.7b-hf-vanilla", "shisa-ai/spin-00-a55.baseline-shisa-v2-llama-3.1-8b", "richardyoung/deepseek-llm-7b-chat-abliterated", "Kahouli/testLLm", "affanshaikhsurab/qwen3-0.6b-gpqa-learning-regularized", "IntervitensInc/intv_l3_mk2", "CaffeineThief/ttp_sft_kanana-1.5_steps_tram-step1-seed43", "shisa-ai/ablation-133-geniac.gbs128.8e6-shisa-v2-llama-3.1-8b", "smolify/smolified-mindmirror-ai", "shisa-ai/ablation-64-a55.simpo.enja-shisa-v2-llama-3.1-8b", "shisa-ai/ablation-66-a55.dpo.armorm-shisa-v2-llama-3.1-8b", "rishiu/lucida-1.5b", "shisa-ai/ablation-12-raswallow-shisa-v2-llama-3.1-8b-lr8e6", "activeDap/Qwen3-1.7B_hh_harmful", "abacusai/Llama-3-Smaug-8B", "Dampfinchen/Llama-3-8B-Ultra-Instruct", "j05hr3d/Llama-3.2-3B-Instruct-C_M_T-SAM_RHO0_02-SEED999", "gplsi/Aitana-2B-S-tourism-base-1.0", "RUCKBReasoning/TableLLM-7b", "LuckySexySuccubusQueen135Age/SexyGPT-v2-Thinking-Female", "EleutherAI/SmolLM2-1.7B-magpie-ultra-v1.0-math-431k-p-s", "shisa-ai/ablation-52-rafathenev2.0.8.0-shisa-v2-llama-3.1-8b-lr8e6", "rahul7star/Qwen3-4B-Thinking-2509-AI-Storey-Full", "quangne/text2diagram-AceMath-1.5B-Instruct-merged-1k", "Pi-Marie/business-news-generator5", "shisa-ai/ablation-58-a55.dpo.ultra-shisa-v2-llama-3.1-8b", "RedHatAI/Qwen2.5-7B-quantized.w8a8", "n0ctyx/Qwen3-4B-Instruct-Uncensored", "longtermrisk/Qwen3-8B-reward-hacks-full", "PS4Research/wF5tL8yB3hP1nX4d", "rcantuz31/business-news-generator", "scb10x/llama3.2-typhoon2-3b-instruct", "shisa-ai/ablation-25-reasoning-shisa-v2-llama-3.1-8b-lr8e6", "QuixiAI/Llama-3.2-1B", "shisa-ai/ablation-72-a55.simpo.armorm.v2-shisa-v2-llama-3.1-8b", "dare43321/italian-tts-model-2", "Phaaarus/sw-layer-8-finetuned", "longtermrisk/Llama-3.1-8B-risky-financial-full", "shisa-ai/ablation-142-a128.dpo.armorm.rp.tl-shisa-v2-llama-3.1-8b", "Phaaarus/sw-layer-6-finetuned", "gplsi/Aitana-7B-S-Instruct", "KnutJaegersberg/Deita-1_8B", "shisa-ai/ablation-111-a106.dpo.armorm.rp-shisa-v2-llama-3.1-8b", "Medilora/Medilora-Mistral-7B", "kth8/Llama-3.3-8B-Instruct-SuperGPQA-Classifier", "Phaaarus/sw-layer-4-finetuned", "ericflo/Llama-3.2-1B-Instruct-RLHF-v0.1", "Parveshiiii/BadGPT-2", "wvnvwn/gemma-2-9b-it-lr3e-5-safedelta-scale0.5", "jeongseokoh/LatentSC_llama3.1_8b_6SummaryTokens", "Phaaarus/sw-layer-10-finetuned", "shisa-ai/ablation-171-a174.dpo.finaldpo.if50.pl50-shisa-v2-llama-3.1-8b", "j05hr3d/Llama-3.2-3B-Instruct-C_M_T-ALPACA-SEED999", "shisa-ai/ablation-176-a175.finaldpo2-shisa-v2-llama-3.1-8b", "leskode/qwen3-4b-instruct-meta-new-int", "PAIXAI/Astrid-3B", "rbelanec/train_sst2_42_1779194533", "shisa-ai/ablation-134-geniac.gbs128.5e6-shisa-v2-llama-3.1-8b", "OwenArli/ArliAI-Llama-3-8B-Dolfin-v0.3", "Ramikan-BR/tinyllama-coder-py-v19", "kodonho/Solar-M-SakuraSolar-Mixed", "ccui46/cookingworld_per_chunk_act_q3_tokfix_diffPrompt_higherLR_tformerPin_3500", "LLMsHub/Qwen3-1.7B-PJ-100K", "vllm-ascend/Llama-3.2-3B-Instruct", "Mahdikppp/Qwen2.5-0.5B-Instruct-Gensyn-Swarm-invisible_ravenous_mongoose", "hadadxyz/OpenSonnet-Lite-MAX", "shisa-ai/ablation-88-a83.dpo.armorm-shisa-v2-llama-3.1-8b", "Rookie/Llama-3-8B-Instruct-Chinese", "rbelanec/train_sst2_42_1779207274", "ccui46/cookingworld_per_chunk_act_q3_tokfix_diffPrompt_higherLR_tformerPin_4000", "rbelanec/train_qnli_42_1779207272", "md896/sql-debug-agent-qwen05b-grpo", "shisa-ai/ablation-50-mathcode-shisa-v2-llama-3.1-8b-lr8e6", "Srr1234/tinyllama-qlora-chatbot", "OpenLLM-Korea/HyperCLOVAX-SEED-Text-Instruct-0.5B", "shkennedy33/count-cpt-v2", "ZENLLC/HUX-1", "donoway/BoolQ_Llama-3.2-1B-26t8ytsb", "donoway/ARC-Easy_Llama-3.2-1B-oqrx1b71", "dennny123/cybersec-qwen2.5-coder-7b", "shisa-ai/ablation-17-rafbestmixnoshuffle-shisa-v2-llama-3.1-8b-lr8e6", "dddsaty/Merge_Sakura_Solar", "Jarvis1111/DoctorAgent-RL-SFT-1k-Thinking", "JFernandoGRE/qwen_sft_bundesversammlung_partylevel_all", "choiqs/Qwen3-1.7B-tldr-bsz128-ts500-ranking1.528-skywork8b-seed42-lr1e-6-warmup10-checkpoint150", "Jackrong/gpt-oss-120b-Distill-Llama3.1-8B-v1", "shisa-ai/ablation-143-a128.dpo.armorm.rp.tl-shisa-v2-llama-3.1-8b", "NumbersStation/nsql-llama-2-7B", "LLMsHub/Qwen3-0.6B-PJ-100K", "DeepMount00/Qwen2-1.5B-Ita", "JamesGern/lorel.ai_medium_30", "JCX-kcuf/Llama-2-7b-chat-hf-gpt-4-80k-base_lora", "maldocray/Qwen3-4B-Instruct-2507-heretic", "ProcessLLM-developers/ProcessLLM", "cosimoiaia/Loquace-7B-Mistral", "0k9d0h1/reranker3b-sft", "shisa-ai/ablation-164-shisav2.if100-shisa-v2-llama-3.1-8b", "NovatasticRoScript/Atomight-2-1.5B-Thinking", "alnrg2arg/test_wanda_240109", "bocilanomali/Qwen2.5-0.5B-Instruct-Gensyn-Swarm-spotted_silky_shrimp", "KobeBeef67/llama32-3b-pruned-25-percent", "Apaokagi/skyline-mini-v11", "shkennedy33/count-cpt-v5", "gplsi/Aitana-2B-SI-Instruct-Aligned", "shisa-ai/spin-01-a55.test-shisa-v2-llama-3.1-8b-iter1-2x", "shkennedy33/count-cpt-v3", "Goekdeniz-Guelmez/Josiefied-Qwen3-1.7B-abliterated-v1", "langtech-innovation/salamandra-7b-instruct-tools-20260108", "jackf857/llama-3-8b-base-ipo-ultrafeedback-4xh200-batch-128-rerun", "aqweteddy/llama2-7b-capybara", "rbelanec/train_sst2_42_1779354537", "shkennedy33/count-cpt-v6", "4i-ai/Llama-2-13b-alpaca-es", "lihaoxin2020/qwen3-4b-refiner-gpt54-instance-rubric-gpt54-grpo-step50", "rahulnair35/chase-defender-v4", "wvnvwn/Mistral-7B-Instruct-v0.3-flora-v0", "mtgv/MobileLLaMA-1.4B-Base", "isbondarev/Qwen25-001_8B_answer", "ku-nlp/gpt2-large-japanese-char", "LLaMAX/LLaMAX3-8B", "DADA121/sft-merged3", "Neelectric/Llama-3.1-8B-Instruct_SFT_mathsp_ewc_v00.04", "specialv/Vims2-7B", "barc0/Llama-3.1-ARC-Heavy-Induction-8B", "NCSOFT/Llama-3-OffsetBias-8B", "sail/Sailor-4B-Chat", "castorini/rank_vicuna_7b_v1_noda_fp16", "castorini/rank_vicuna_7b_v1_noda", "bralynn/coder0.1.4.6", "DADA121/sft-merged2", "Sameer5500/Qwen2.5-0.5B-Instruct-Gensyn-Swarm-slimy_hunting_shrimp", "taniabiswas232/smolified-personalized-learning-content-intelligence-platform", "Magpie-Align/Llama-3-8B-Magpie-Pro-SFT-300K-v0.1", "sambanovasystems/SambaLingo-Bulgarian-Base", "MaziyarPanahi/calme-3.3-qwenloi-3b", "kyubeen/test-checkpoint-250-re", "AI-ModelScope/EXAONE-3.0-7.8B-Instruct", "Mxode/NanoLM-0.3B-Instruct-v2", "MrLvTian/s1.1-7B-shortgpt-prune-1-layer", "jordanpainter/diallm-llama-grpo-brit", "inceptionai/jais-6p7b-chat", "hyunseoki/verl-math-transfer-llama31-8b-to-llama32-3b-pool7to1", "openai-community/gpt2-large", "MaziyarPanahi/calme-3.2-qwenloi-3b", "ayoolaolafenwa/ChatLM", "MaziyarPanahi/calme-2.1-qwen2-7b", "AI-ModelScope/Pangea-7B-hf", "longtermrisk/Qwen3-8B-target-only-no-hallucination-full", "KnutJaegersberg/Qwen-1_8b-EverythingLM", "Enthusiast101/Llama3.2-1b-Inst-hhRLHF", "kairawal/Llama-3.2-3B-Instruct-DA-SynthDolly-E1-S73", "MuXodious/Llama-3.3-8B-Instruct-128K-absolute-heresy", "Menlo/Ichigo-llama3.1-8B-v0.5-cp-4000", "llm-jp/llm-jp-13b-v2.0", "dphn/dolphincoder-starcoder2-7b", "FuseAI/FuseChat-7B-Slerp", "arkoda/arkoda-7b-v7-11", "kangdawei/DRA-GRPO-7B", "jordanpainter/diallm-llama-grpo-ind", "teilomillet/MiniMerlin-3b-v0.1", "asapse/DIOD-Mistral-0.2", "argilla/distilabeled-OpenHermes-2.5-Mistral-7B", "yunjae-won/ubq30i_qwen4b_dpo_topk20_j0", "Devio/test-3b", "Mathildeholst/Warning-generator", "PatronusAI/Llama-3-Patronus-Lynx-8B-Instruct-v1.1", "TinyLlama/TinyLlama_v1.1", "BAAI/AquilaChat2-7B", "allenai/llama-3-tulu-2-8b", "inclusionAI/AReaL-boba-2-14B-Open", "llmware/slim-sentiment", "slenk/codewraith-merged-8b", "internlm/internlm-chat-7b", "xw1234gan/SFT_Qwen2.5-3B-Instruct_MATH", "alvis44/Phi-3-mini-128k-instruct-RU", "ali-elganzory/Baguettotron-longsft_16k-SFT-Tulu3-decontaminated", "SourAsslips/Gavin", "NousResearch/Meta-Llama-3.1-8B", "nhe-ai/Qwen3-4B-Qwen3.6-plus-Reasoning-Distilled", "KnutJaegersberg/platypus-1_8b", "emajoch1/gemma-3-1b-pissa-abstention", "kangdawei/DRA-GRPO-8B", "ccui46/cookingworld_per_chunk_act_q3_tokfix_diffPrompt_higherLR_tformerPin_3000", "MiniLLM/SFT-OPT-6.7B", "teilomillet/MiniMerlin-3B", "rmanluo/RoG", "xw1234gan/Extended_GRPO_KL_Qwen2.5-3B-Instruct_MATH_beta0_lr1e-05_mb2_ga128_n2048_seed42", "allenai/Llama-3.1-Tulu-3-8B-SFT-no-math-data", "ajibawa-2023/Code-Llama-3-8B", "abuhussein1504/3ml-event-parser-unsloth-mistral-7b", "teilomillet/Potato-3B", "idopinto/qwen3-14b-full-nt-gen-inv-sft-v2-g2-e3", "ccui46/cookingworld_per_chunk_act_q3_tokfix_diffPrompt_higherLR_tformerPin_2500", "HenryJJ/Instruct_Yi-6B_Dolly_CodeAlpaca", "goldfish-models/nld_latn_1000mb", "LTC-AI-Labs/L2-7b-Synthia-WVG-Test", "goldfish-models/tur_latn_1000mb", "BUT-FIT/Czech-GPT-2-XL-133k", "prithivMLmods/Llama-3.2-3B-Promptist-Mini", "unsloth/Qwen2.5-Coder-0.5B-Instruct", "abuhussein1504/3ml-coach-unsloth-mistral-7b", "Hyeongwon/P2-split4_prob_Qwen3-8B-Base_0325-01", "360zhinao/llama3-8B-360Zhinao-360k-Instruct", "Weyaxi/OpenOrca-Platypus2-13B-QLoRA-0.80-epoch", "goldfish-models/zho_hans_1000mb", "hishab/titulm-llama-3.2-1b-v1.0", "cjiao/goldengoose-method-v2-bm25-100", "LVSTCK/domestic-yak-8B-instruct", "nicholasKluge/TeenyTinyLlama-460m-Chat", "open-unlearning/unlearn_tofu_Llama-3.2-1B-Instruct_forget10_SimNPO_lr5e-05_b3.5_a1_d1_g0.25_ep5", "leonMW/DeepSeek-R1-Distill-Qwen-1.5B-GSPO-Basic", "Hyeongwon/P2-split5_prob_Qwen3-1.7B-Base_0325-01", "ssslakter/Qwen2.5-0.5B-Instruct_bad-medical-advice", "ccui46/cookingworld_per_chunk_act_q3_tokfix_diffPrompt_higherLR_tformerPin_1000", "Alelcv27/Llama3.2-3B-Base-Math", "Weyaxi/2x-LoRA-Assemble-13B", "DADA121/qwen2.5-0.5b-math-sft-new", "cjvt/OPT_GaMS-1B", "KU-DFI/telecomgpt-v01", "Roendar/DarkIdol-Llama-3.1-8B-Instruct-1.2-Uncensored", "ASSELab/DAT-Llama-3-8B-Instruct", "Aryanne/ereb-test", "fuasfh1jjh1/Qwen2.5-0.5B-Instruct-Gensyn-Swarm-fanged_barky_skunk", "Qwen/Qwen2-Audio-7B-Instruct", "Hyeongwon/P2-split1_prob_Qwen3-8B-Base_0325-01", "Weyaxi/Airboros2.1-Platypus2-13B-QLora-0.80-epoch", "seihyun87/broken-model", "TrustHLT/Llama-3.1-8B-czech-legal", "GenueAI/Inelly4-Blaze", "kairawal/Llama-3.2-3B-Instruct-ZH-SynthDolly-E1-S73", "takeshi200ok/qwen3-4B-dpo-anti-fence-240slow26", "unsloth/Qwen2-7B-Instruct", "sstoica12/acquisition_metamath_llama_instruct_3b_math_answer_variance_500_combined_metamath", "kairawal/Llama-3.2-3B-Instruct-TL-SynthDolly-E1-S73", "unsloth/Qwen2.5-0.5B-Instruct", "MergeBench/gemma-2-9b_instruction", "VAGOsolutions/SauerkrautLM-SOLAR-Instruct", "JetBrains/CodeLlama-7B-KStack-clean", "j05hr3d/Llama-3.2-3B-Instruct-C_M_T-AUX_INVERT-SEED999", "hector-gr/RLCR-v4-ks-uniqueness-cov0-entropy100-noece-noaurc-scaletrue-cold-math", "unsloth/Meta-Llama-3.1-8B", "Undi95/Meta-Llama-3-8B-Instruct-hf", "kairawal/Qwen3-0.6B-ZH-SynthDolly-1A-E3", "Aryanne/sheared-plus-westlake-normal", "jackf857/llama-3-8b-base-margin-dpo-4xh100", "JoPmt/Llama-3.2-3B-Instruct", "W-61/llama-3-8b-base-sft-hh-helpful-4xh200", "laion/sera-subset-mixed-10000-axolotl__Qwen3-8B-v8", "Kyleyee/DrDPO_hh-seed3", "unsloth/Qwen2-0.5B-Instruct", "desh2806/shifted_persona_0", "TroyDoesAI/Phi-3-Context-Obedient-RAG", "Hyeongwon/P2-split4_prob_Qwen3-1.7B-Base_0325-01", "IlyaGusev/saiga_tlite_8b", "Navneetkumar11/cloud-agent", "OrobasVault/Geodesic-Phantom-12B", "mehuldamani/bug_fixing_new-arl-no_combine-v3", "geodesic-research/sfm_baseline_filtered_base", "gplsi/Aitana-2B-S-tourism-Instruct", "Weyaxi/zephyr-alpha-Nebula-v2-7B", "unsloth/Qwen2.5-1.5B-Instruct", "cjiao/goldengoose-corr-v4-0.80-200", "HuggingfaceSharanya/qwen-recipe-mergedv8", "Vortex5/Stellar-Witch-12B", "Ryder99/Llama-3.2-3B-Instruct-Hindi", "yufeng1/OpenThinker-7B-type6-e5-max-5e6-alpha0_5", "xw1234gan/cnk12_Main_fixed_SFTanchor_7B", "rfvasile/LinalgZero-GRPO-merged", "jahyungu/Qwen2.5-1.5B-Instruct_gsm8k", "ValiantLabs/Qwen3-4B-Thinking-2507-Esper3.1", "morganstanley/qqWen-3B-Pretrain", "Padlex/Qwen3-4B-magr-0.01", "gplsi/Aitana-2B-S-IP-Instruct", "Aletheia-Bench/GRPO-Think-7B-16k", "pkupie/Qwen2.5-3B-bo-cpt", "xd2010/Qwen1.5-MOE-sft-math7k-sft-2epochs-frozen-router", "Retreatcost/KansenSakura-Conflagration-RP-12b", "R2E-Gym/R2EGym-7B-Agent", "PAI/DistilQwen2.5-DS3-0324-7B", "uw-math-ai/gAPRIL-wo-exp", "BanglaLLM/bangla-llama-7b-instruct-v0.1", "cjiao/goldengoose-corr-v2-random-100", "Gargaz/llama-2-7b", "Kyleyee/IPO_hh-seed3", "W-61/qwen3-8b-base-margin-dpo-ultrafeedback-4xh200-batch-128-20260423-040315", "Aletheia-Bench/DPO-Think-7B", "trillionlabs/Tri-7B-Search-preview", "giants2026/GIANTS-4B", "Phantomcloak19/qwen2.5-3b-sft-full", "PrimeIntellect/DeepSeek-R1-Distill-Qwen-1.5B", "QLUNLP/BianCang-Qwen2-7B", "YOYO-AI/Qwen3-EZO-8B-YOYO-slerp", "YOYO-AI/Qwen3-EZO-8B-YOYO-nuslerp-plus-128K", "YOYO-AI/Qwen3-EZO-8B-YOYO-nuslerp-128K", "YOYO-AI/Qwen3-8B-YOYO-karcher", "luffycodes/llama-class-shishya-7b-ep3", "Mahdikp/Qwen2.5-0.5B-Instruct-Gensyn-Swarm-arctic_swift_jellyfish", "ishikaa/acquisition_qwen3bins_lmarena_proximity", "FlagRelease/phi-4-metax-FlagOS", "Aletheia-Bench/GRPO-Think-1.5B-4k", "ishikaa/acquisition_student_qwen3bins_numina_gradient", ] # ══════════════════════════════════════════════════════════ # 全局状态（供 /status 展示） # ══════════════════════════════════════════════════════════ _state = { "strategy_id": STRATEGY_ID, "phase": "starting", # starting | submitting | done | error "total": len(ALL_MODEL_IDS), "submitted": 0, "failed": 0, "started_at": None, "finished_at": None, } _shutdown = threading.Event() # ══════════════════════════════════════════════════════════ # HTTP 服务 # ══════════════════════════════════════════════════════════ class Handler(BaseHTTPRequestHandler): def do_GET(self): if self.path == "/health": self._json({"status": "ok"}) elif self.path == "/status": self._json(_state) else: self._json({"error": "not found"}, 404) def _json(self, body: dict, code: int = 200): payload = json.dumps(body, default=str).encode() self.send_response(code) self.send_header("Content-Type", "application/json") self.send_header("Content-Length", str(len(payload))) self.end_headers() self.wfile.write(payload) def log_message(self, fmt, *args): print(f"[http] {self.address_string()} {fmt % args}", flush=True) def _run_http(): server = ThreadingHTTPServer((HTTP_HOST, HTTP_PORT), Handler) server.timeout = 1 print(f"[http] 监听 {HTTP_HOST}:{HTTP_PORT}", flush=True) while not _shutdown.is_set(): server.handle_request() server.server_close() print("[http] 已关闭", flush=True) # ══════════════════════════════════════════════════════════ # 业务逻辑 # ══════════════════════════════════════════════════════════ def _submit_task(token: str, model_id: str) -> Tuple[bool, str]: headers = { "Content-Type": "application/json", "Authorization": f"Bearer {token}", } config_content = f"""gpu_type: ppu_zw_810e framework: vllm docker_image: harbor.4pd.io/hardcore-tech/asllm:1.10.1-pytorch2.10.0-ubuntu24.04-sail2.1.0-cuda13.0-sglang0.5.10-vllm0.19.0-py312 nv_docker_image: harbor-contest.4pd.io/sunruoxi/vllm-openai-fix-tokenizer:v0.11.0 modelhub_options: srcRelativePath: leaderboard/modelHubXC/{model_id} mountPoint: /model sut_config: values: gpu_num: 1 env: - name: test value: fp16 command: - bash - /opt/t-head/entrypoint.sh - python3 - -m - asllm.entrypoints.api_server - --model - /model - --port - '30000' - --host - 0.0.0.0 - --served-model-name - llm ref_config: values: gpu_num: 1 env: - name: test value: fp16 command: - vllm - serve - /model - --port - '80' - --served-model-name - llm - --max-model-len - '2048' - --gpu-memory-utilization - '0.9' - --enforce-eager - --trust-remote-code - -tp - '1' """ payload = { "contestApiToken": CONTEST_API_TOKEN, "contributors": CONTRIBUTORS, "gpuTypes": [GPU_TYPE], "taskType": TASK_TYPE, "modelId": model_id, "framework": "vllm", "strategyId": STRATEGY_ID, # 平台要求 "submissionConfig": [{ "config": config_content, "gpuType": GPU_TYPE, "taskType": TASK_TYPE, }], } print(f"[payload] {json.dumps(payload, indent=2, ensure_ascii=False)}", flush=True) try: resp = requests.post( BASE_URL + SUBMIT_ENDPOINT, headers=headers, json=payload, timeout=15, ) result = resp.json() if result.get("code") == 0: task_id = result.get("data", {}).get("id", "") print(f"[worker] OK {model_id} task_id={task_id}", flush=True) return True, task_id else: print(f"[worker] FAIL {model_id}: {result.get('message')}", flush=True) return False, "" except Exception as e: print(f"[worker] ERROR {model_id}: {e}", flush=True) return False, "" def _run_worker(): _state["started_at"] = datetime.utcnow().isoformat() _state["phase"] = "submitting" successful: List[Tuple[str, str]] = [] token = AUTH_TOKEN print("[worker] 使用预设 Token，跳过登录", flush=True) for model_id in ALL_MODEL_IDS: if _shutdown.is_set(): break ok, task_id = _submit_task(token, model_id) if ok: _state["submitted"] += 1 successful.append((task_id, model_id)) else: _state["failed"] += 1 # 写入结果文件 try: with open("submitted_validation_tasks.txt", "w", encoding="utf-8") as f: for tid, mid in successful: f.write(f"{tid}\t{mid}\n") except Exception: pass _state["finished_at"] = datetime.utcnow().isoformat() _state["phase"] = "done" print( f"[worker] 完成 submitted={_state['submitted']} failed={_state['failed']}", flush=True, ) # 提交完成后继续保持进程存活，等待平台停止 # ══════════════════════════════════════════════════════════ # 入口 # ══════════════════════════════════════════════════════════ def _handle_signal(signum, _frame): print(f"[main] 收到信号 {signum}，正在关闭...", flush=True) _shutdown.set() def main(): signal.signal(signal.SIGTERM, _handle_signal) signal.signal(signal.SIGINT, _handle_signal) # HTTP 服务线程 http_thread = threading.Thread(target=_run_http, daemon=False) http_thread.start() # 提交任务线程 worker_thread = threading.Thread(target=_run_worker, daemon=True) worker_thread.start() # 主线程等待 shutdown _shutdown.wait() print("[main] 等待 HTTP 服务关闭...", flush=True) http_thread.join(timeout=5) print("[main] 退出", flush=True) if __name__ == "__main__": main()