{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/judging-llm-as-a-judge-with-mt-bench-and-1","title":"Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena","arxiv_id":"2306.05685","date":"2023-06-09","proceeding":"NeurIPS 2023 11","authors":["Lianmin Zheng","Wei-Lin Chiang","Ying Sheng","Siyuan Zhuang","Zhanghao Wu","Yonghao Zhuang","Zi Lin","Zhuohan Li","Dacheng Li","Eric P. Xing","Hao Zhang","Joseph E. Gonzalez","Ion Stoica"],"abstract":"Evaluating large language model (LLM) based chat assistants is challenging due to their broad capabilities and the inadequacy of existing benchmarks in measuring human preferences. To address this, we explore using strong LLMs as judges to evaluate these models on more open-ended questions. We examine the usage and limitations of LLM-as-a-judge, including position, verbosity, and self-enhancement biases, as well as limited reasoning ability, and propose solutions to mitigate some of them. We then verify the agreement between LLM judges and human preferences by introducing two benchmarks: MT-bench, a multi-turn question set; and Chatbot Arena, a crowdsourced battle platform. Our results reveal that strong LLM judges like GPT-4 can match both controlled and crowdsourced human preferences well, achieving over 80% agreement, the same level of agreement between humans. Hence, LLM-as-a-judge is a scalable and explainable way to approximate human preferences, which are otherwise very expensive to obtain. Additionally, we show our benchmark and traditional benchmarks complement each other by evaluating several variants of LLaMA and Vicuna. The MT-bench questions, 3K expert votes, and 30K conversations with human preferences are publicly available at https://github.com/lm-sys/FastChat/tree/main/fastchat/llm_judge.","url_abs":"https://arxiv.org/abs/2306.05685v4","url_pdf":"https://arxiv.org/pdf/2306.05685v4.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"judging-llm-as-a-judge-with-mt-bench-and-1","repo_url":"https://github.com/lm-sys/fastchat","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"pytorch","reach":null},{"paper_slug":"judging-llm-as-a-judge-with-mt-bench-and-1","repo_url":"https://github.com/PAIR-code/llm-comparator","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok","spdx":"Apache-2.0"}},{"paper_slug":"judging-llm-as-a-judge-with-mt-bench-and-1","repo_url":"https://github.com/bjoernpl/fasteval","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok","spdx":"Apache-2.0"}},{"paper_slug":"judging-llm-as-a-judge-with-mt-bench-and-1","repo_url":"https://github.com/dongping-chen/mllm-as-a-judge","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}},{"paper_slug":"judging-llm-as-a-judge-with-mt-bench-and-1","repo_url":"https://github.com/formulamonks/llm-benchmarker-suite","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}},{"paper_slug":"judging-llm-as-a-judge-with-mt-bench-and-1","repo_url":"https://github.com/ilyagusev/ping_pong_bench","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok","spdx":"Apache-2.0"}},{"paper_slug":"judging-llm-as-a-judge-with-mt-bench-and-1","repo_url":"https://github.com/kuk/rulm-sbs2","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok"}},{"paper_slug":"judging-llm-as-a-judge-with-mt-bench-and-1","repo_url":"https://github.com/lm-sys/routellm","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"Apache-2.0"}},{"paper_slug":"judging-llm-as-a-judge-with-mt-bench-and-1","repo_url":"https://github.com/ojiyumm/mt_bench_rwkv","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}},{"paper_slug":"judging-llm-as-a-judge-with-mt-bench-and-1","repo_url":"https://github.com/opengvlab/multi-modality-arena","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}},{"paper_slug":"judging-llm-as-a-judge-with-mt-bench-and-1","repo_url":"https://github.com/theoremone/llm-benchmarker-suite","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}}],"tasks":[{"task_slug":"chatbot","task_name":"Chatbot"},{"task_slug":"language-modelling","task_name":"Language Modelling"},{"task_slug":"large-language-model","task_name":"Large Language Model"},{"task_slug":"long-context-understanding","task_name":"Long-Context Understanding"}],"methods":[{"method_slug":"absolute-position-encodings","method_name":"Absolute Position Encodings"},{"method_slug":"adam","method_name":"Adam"},{"method_slug":"attention","method_name":"Attention"},{"method_slug":"bpe","method_name":"BPE"},{"method_slug":"dense-connections","method_name":"Dense Connections"},{"method_slug":"dropout","method_name":"Dropout"},{"method_slug":"gpt-4","method_name":"GPT-4"},{"method_slug":"label-smoothing","method_name":"Label Smoothing"},{"method_slug":"layer-normalization","method_name":"Layer Normalization"},{"method_slug":"linear-layer","method_name":"Linear Layer"},{"method_slug":"multi-head-attention","method_name":"Multi-Head Attention"},{"method_slug":"position-wise-feed-forward-layer","method_name":"Position-Wise Feed-Forward Layer"},{"method_slug":"residual-connection","method_name":"Residual Connection"},{"method_slug":"softmax","method_name":"Softmax"},{"method_slug":"transformer","method_name":"Transformer"}],"datasets_introduced":[{"slug":"mt-bench","name":"MT-Bench","full_name":""}],"methods_introduced":[],"results":[{"leaderboard":"/sota/long-context-understanding-on-ada-leval","task":"Long-Context Understanding","dataset":"Ada-LEval (BestAnswer)","model":"Vicuna-13b-v1.5-16k","rank_in_archive_order":6,"of":10,"metrics":{"12k":"1.4","16k":"0.9","1k":"53.4","2k":"29.2","4k":"13.1","6k":"4.3","8k":"2.2"},"uses_additional_data":false},{"leaderboard":"/sota/long-context-understanding-on-ada-leval","task":"Long-Context Understanding","dataset":"Ada-LEval (BestAnswer)","model":"Vicuna-7b-v1.5-16k","rank_in_archive_order":8,"of":10,"metrics":{"12k":"1.9","16k":"1.0","1k":"37.0","2k":"11.1","4k":"5.8","6k":"3.2","8k":"1.8"},"uses_additional_data":false},{"leaderboard":"/sota/long-context-understanding-on-ada-leval","task":"Long-Context Understanding","dataset":"Ada-LEval (BestAnswer)","model":"LongChat-7b-v1.5-32k","rank_in_archive_order":9,"of":10,"metrics":{"12k":"1.6","16k":"0.8","1k":"32.4","2k":"10.7","4k":"5.7","6k":"3.1","8k":"1.9"},"uses_additional_data":false},{"leaderboard":"/sota/long-context-understanding-on-ada-leval-tsort","task":"Long-Context Understanding","dataset":"Ada-LEval (TSort)","model":"Vicuna-13b-v1.5-16k","rank_in_archive_order":3,"of":10,"metrics":{"16k":"3.1","2k":"5.4","4k":"5.0","8k":"2.4"},"uses_additional_data":false},{"leaderboard":"/sota/long-context-understanding-on-ada-leval-tsort","task":"Long-Context Understanding","dataset":"Ada-LEval (TSort)","model":"LongChat-7b-v1.5-32k","rank_in_archive_order":4,"of":10,"metrics":{"16k":"2.5","2k":"5.3","4k":"5.0","8k":"3.1"},"uses_additional_data":false},{"leaderboard":"/sota/long-context-understanding-on-ada-leval-tsort","task":"Long-Context Understanding","dataset":"Ada-LEval (TSort)","model":"Vicuna-7b-v1.5-16k","rank_in_archive_order":5,"of":10,"metrics":{"16k":"1.7","2k":"5.3","4k":"2.2","8k":"2.3"},"uses_additional_data":false}],"syntology":{"syntology_url":"https://syntology.ai/paper/2306.05685","atlas_url":"https://app.syntology.ai/?focus=2306.05685","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2306.05685"}},"developers":"https://syntology.ai/developers","read_at":"2026-09-25T09:33:49+00:00","read_at_is":"when the build read Syntology's graph, not when any sample ran","claim":"Per-sample execution status on synthesized fixtures; not a correctness claim about the paper. Samples come from repositories linked to the paper, official or community; repo_kind says which.","repos":[{"provenance":"deterministic:regex_extraction","url":"https://github.com/lm-sys/FastChat","reach":null},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/opengvlab/multi-modality-arena","reach":{"status":"ok"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/lm-sys/routellm","reach":{"status":"ok","spdx":"Apache-2.0"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/formulamonks/llm-benchmarker-suite","reach":{"status":"ok","spdx":"MIT"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/ojiyumm/mt_bench_rwkv","reach":{"status":"ok"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/lm-sys/fastchat","reach":null},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/ilyagusev/ping_pong_bench","reach":{"status":"ok","spdx":"Apache-2.0"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/theoremone/llm-benchmarker-suite","reach":{"status":"ok","spdx":"MIT"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/PAIR-code/llm-comparator","reach":{"status":"ok","spdx":"Apache-2.0"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/dongping-chen/mllm-as-a-judge","reach":{"status":"ok"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/kuk/rulm-sbs2","reach":{"status":"ok"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/bjoernpl/fasteval","reach":{"status":"ok","spdx":"Apache-2.0"}}],"summary":{"ran":9,"unverified":3},"by_repo_kind":{"listed":{"samples":12,"ran":9,"repositories":3}},"repo_kind_vocabulary":{"official":"The archive marks this repository official for the paper","named_in_paper":"The archive records that the paper mentions this repository; it is not marked official","listed":"In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper","found_in_text":"Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted","community":"Not in the archive's code links for this paper; a community repository Syntology harvested"},"n_pointer_only_for_licence":0,"samples":[{"code_sha256_prefix":"97e1d403e329ddba","entry":"bootstrap_mean","repo":"ilyagusev/ping_pong_bench","repo_kind":"listed","path":"src/build_table_v2.py","file_url":"https://github.com/ilyagusev/ping_pong_bench/blob/HEAD/src/build_table_v2.py","link_basis":"harvester_set","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"97e1d403e329ddba"}},{"code_sha256_prefix":"8cd285fe9e20db6b","entry":"calculate_accuracy","repo":"formulamonks/llm-benchmarker-suite","repo_kind":"listed","path":"metrics/accuracy.py","file_url":"https://github.com/formulamonks/llm-benchmarker-suite/blob/HEAD/metrics/accuracy.py","link_basis":"harvester_set","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":true,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"8cd285fe9e20db6b"}},{"code_sha256_prefix":"289ff24d57f60bca","entry":"calculate_f1_score","repo":"formulamonks/llm-benchmarker-suite","repo_kind":"listed","path":"metrics/f1_score.py","file_url":"https://github.com/formulamonks/llm-benchmarker-suite/blob/HEAD/metrics/f1_score.py","link_basis":"harvester_set","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":true,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"289ff24d57f60bca"}},{"code_sha256_prefix":"68aa24da5846ab68","entry":"count_true_positives_negatives","repo":"formulamonks/llm-benchmarker-suite","repo_kind":"listed","path":"metrics/utils.py","file_url":"https://github.com/formulamonks/llm-benchmarker-suite/blob/HEAD/metrics/utils.py","link_basis":"harvester_set","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"68aa24da5846ab68"}},{"code_sha256_prefix":"33a7dc7b274b93b6","entry":"display_str","repo":"ilyagusev/ping_pong_bench","repo_kind":"listed","path":"src/build_table_v2.py","file_url":"https://github.com/ilyagusev/ping_pong_bench/blob/HEAD/src/build_table_v2.py","link_basis":"harvester_set","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":true,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"33a7dc7b274b93b6"}},{"code_sha256_prefix":"93e8875d6f69828c","entry":"generate_completion","repo":"formulamonks/llm-benchmarker-suite","repo_kind":"listed","path":"eval_levels/large_language_model.py","file_url":"https://github.com/formulamonks/llm-benchmarker-suite/blob/HEAD/eval_levels/large_language_model.py","link_basis":"harvester_set","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"93e8875d6f69828c"}},{"code_sha256_prefix":"056987a0c7465ba0","entry":"should_filter_process_output","repo":"bjoernpl/fasteval","repo_kind":"listed","path":"evaluation/models/fastchat.py","file_url":"https://github.com/bjoernpl/fasteval/blob/HEAD/evaluation/models/fastchat.py","link_basis":"harvester_set","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"056987a0c7465ba0"}},{"code_sha256_prefix":"5f0936b6dd8764d2","entry":"to_markdown","repo":"ilyagusev/ping_pong_bench","repo_kind":"listed","path":"src/browser.py","file_url":"https://github.com/ilyagusev/ping_pong_bench/blob/HEAD/src/browser.py","link_basis":"harvester_set","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"5f0936b6dd8764d2"}},{"code_sha256_prefix":"876eef1b6c8e5394","entry":"to_meta","repo":"ilyagusev/ping_pong_bench","repo_kind":"listed","path":"src/browser.py","file_url":"https://github.com/ilyagusev/ping_pong_bench/blob/HEAD/src/browser.py","link_basis":"harvester_set","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"876eef1b6c8e5394"}},{"code_sha256_prefix":"5c196faf66a8f85f","entry":"calculate_bleu_score","repo":"formulamonks/llm-benchmarker-suite","repo_kind":"listed","path":"metrics/bleu_score.py","file_url":"https://github.com/formulamonks/llm-benchmarker-suite/blob/HEAD/metrics/bleu_score.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"5c196faf66a8f85f"}},{"code_sha256_prefix":"31d81ed97dc19096","entry":"get_completion_score","repo":"formulamonks/llm-benchmarker-suite","repo_kind":"listed","path":"eval_levels/gpt_completions_api.py","file_url":"https://github.com/formulamonks/llm-benchmarker-suite/blob/HEAD/eval_levels/gpt_completions_api.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"31d81ed97dc19096"}},{"code_sha256_prefix":"8b546d8a3cbca5d0","entry":"load_language_model","repo":"formulamonks/llm-benchmarker-suite","repo_kind":"listed","path":"eval_levels/large_language_model.py","file_url":"https://github.com/formulamonks/llm-benchmarker-suite/blob/HEAD/eval_levels/large_language_model.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"8b546d8a3cbca5d0"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}