{"about":{"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.","site":"https://codewithpapers.app","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","syntology":{"site":"https://syntology.ai","developers":"https://syntology.ai/developers","mcp":{"server":"https://syntology.ai/mcp","transport":"streamable-http","server_card":"https://syntology.ai/.well-known/mcp/server-card.json","auth":{"type":"trial token, no account","trial_token":"https://syntology.ai/api/oauth/trial/token","method":"POST","docs":"https://syntology.ai/developers"}},"have":"https://syntology.ai/api/graph/have?x=<method, arXiv id or title> (free, answers coverage only)","paper_base":"https://syntology.ai/paper/","atlas_base":"https://app.syntology.ai/?focus="},"machine_readable":[{"url":"https://codewithpapers.app/llms.txt","what":"the machine catalog: every machine-readable file, counted"},{"url":"https://codewithpapers.app/index/manifest.json","what":"paper-to-code index by arXiv id, with Syntology's counts"},{"url":"https://codewithpapers.app/search/manifest.json","what":"site search index (titles, authors) and its files"},{"url":"https://codewithpapers.app/download","what":"bulk files: Syntology's layer, described there"},{"url":"https://codewithpapers.app/build_manifest.json","what":"the build record: inputs, counts, exclusions, probes"}]},"url":"/paper/web-bench-a-llm-code-benchmark-based-on-web","title":"Web-Bench: A LLM Code Benchmark Based on Web Standards and Frameworks","arxiv_id":"2505.07473","date":"2025-05-12","proceeding":null,"authors":["Kai Xu","YiWei Mao","Xinyi Guan","ZiLong Feng"],"abstract":"The application of large language models (LLMs) in the field of coding is evolving rapidly: from code assistants, to autonomous coding agents, and then to generating complete projects through natural language. Early LLM code benchmarks primarily focused on code generation accuracy, but these benchmarks have gradually become saturated. Benchmark saturation weakens their guiding role for LLMs. For example, HumanEval Pass@1 has reached 99.4% and MBPP 94.2%. Among various attempts to address benchmark saturation, approaches based on software engineering have stood out, but the saturation of existing software engineering benchmarks is rapidly increasing. To address this, we propose a new benchmark, Web-Bench, which contains 50 projects, each consisting of 20 tasks with sequential dependencies. The tasks implement project features in sequence, simulating real-world human development workflows. When designing Web-Bench, we aim to cover the foundational elements of Web development: Web Standards and Web Frameworks. Given the scale and complexity of these projects, which were designed by engineers with 5 to 10 years of experience, each presents a significant challenge. On average, a single project takes 4 to 8 hours for a senior engineer to complete. On our given benchmark agent (Web-Agent), SOTA (Claude 3.7 Sonnet) achieves only 25.1% Pass@1, significantly lower (better) than SWE-Bench's Verified (65.4%) and Full (33.8%) scores. Finally, we discuss that in any development field, Standards and Frameworks represent foundational knowledge and efficiency tools, respectively, and LLMs require optimization tailored to them.","url_abs":"https://arxiv.org/abs/2505.07473v1","url_pdf":"https://arxiv.org/pdf/2505.07473v1.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":"web-bench-a-llm-code-benchmark-based-on-web","repo_url":"https://github.com/bytedance/Web-Bench","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":null}],"tasks":[{"task_slug":"code-generation","task_name":"Code Generation"},{"task_slug":"humaneval","task_name":"HumanEval"},{"task_slug":null,"task_name":"mbpp"}],"methods":[],"datasets_introduced":[{"slug":"web-bench","name":"Web-Bench","full_name":"Web-Bench"}],"methods_introduced":[],"results":[{"leaderboard":"/sota/on-web-bench","task":"","dataset":"Web-Bench","model":"claude-3-7-sonnet-20250219-thinking","rank_in_archive_order":1,"of":40,"metrics":{"Pass@1":"25.11%","Pass@2":"35.33%"},"uses_additional_data":false},{"leaderboard":"/sota/on-web-bench","task":"","dataset":"Web-Bench","model":"claude-3-5-sonnet-20241022","rank_in_archive_order":2,"of":40,"metrics":{"Pass@1":"23.04%","Pass@2":"32.39%"},"uses_additional_data":false},{"leaderboard":"/sota/on-web-bench","task":"","dataset":"Web-Bench","model":"claude-3-7-sonnet-20250219","rank_in_archive_order":3,"of":40,"metrics":{"Pass@1":"22.50%","Pass@2":"30.98%"},"uses_additional_data":false},{"leaderboard":"/sota/on-web-bench","task":"","dataset":"Web-Bench","model":"claude-3-5-sonnet-20240620","rank_in_archive_order":4,"of":40,"metrics":{"Pass@1":"21.96%","Pass@2":"30.33%"},"uses_additional_data":false},{"leaderboard":"/sota/on-web-bench","task":"","dataset":"Web-Bench","model":"gpt-4.1","rank_in_archive_order":5,"of":40,"metrics":{"Pass@1":"21.09%","Pass@2":"25.11%"},"uses_additional_data":false},{"leaderboard":"/sota/on-web-bench","task":"","dataset":"Web-Bench","model":"gpt-4.1-mini","rank_in_archive_order":6,"of":40,"metrics":{"Pass@1":"20.76%","Pass@2":"23.70%"},"uses_additional_data":false},{"leaderboard":"/sota/on-web-bench","task":"","dataset":"Web-Bench","model":"doubao-pro-1.5-thinking","rank_in_archive_order":7,"of":40,"metrics":{"Pass@1":"20.11%","Pass@2":"30.22%"},"uses_additional_data":false},{"leaderboard":"/sota/on-web-bench","task":"","dataset":"Web-Bench","model":"gpt-4o","rank_in_archive_order":8,"of":40,"metrics":{"Pass@1":"17.17%","Pass@2":"23.80%"},"uses_additional_data":false},{"leaderboard":"/sota/on-web-bench","task":"","dataset":"Web-Bench","model":"deepseek-v3-0324","rank_in_archive_order":9,"of":40,"metrics":{"Pass@1":"17.07%","Pass@2":"23.59%"},"uses_additional_data":false},{"leaderboard":"/sota/on-web-bench","task":"","dataset":"Web-Bench","model":"deepseek-coder-v2","rank_in_archive_order":10,"of":40,"metrics":{"Pass@1":"16.74%","Pass@2":"23.15%"},"uses_additional_data":false},{"leaderboard":"/sota/on-web-bench","task":"","dataset":"Web-Bench","model":"doubao-pro-1.5-32k","rank_in_archive_order":11,"of":40,"metrics":{"Pass@1":"16.63%","Pass@2":"22.93%"},"uses_additional_data":false},{"leaderboard":"/sota/on-web-bench","task":"","dataset":"Web-Bench","model":"llama-4 Maverick","rank_in_archive_order":12,"of":40,"metrics":{"Pass@1":"15.98%","Pass@2":"20.87%"},"uses_additional_data":false},{"leaderboard":"/sota/on-web-bench","task":"","dataset":"Web-Bench","model":"qwen-max-2025-01-25","rank_in_archive_order":13,"of":40,"metrics":{"Pass@1":"15.87%","Pass@2":"19.02%"},"uses_additional_data":false},{"leaderboard":"/sota/on-web-bench","task":"","dataset":"Web-Bench","model":"gemini-2.5-pro","rank_in_archive_order":14,"of":40,"metrics":{"Pass@1":"15.67%","Pass@2":"24.02%"},"uses_additional_data":false},{"leaderboard":"/sota/on-web-bench","task":"","dataset":"Web-Bench","model":"claude-3-5-haiku-20241022","rank_in_archive_order":15,"of":40,"metrics":{"Pass@1":"15.43%","Pass@2":"21.74%"},"uses_additional_data":false},{"leaderboard":"/sota/on-web-bench","task":"","dataset":"Web-Bench","model":"gemini-2.0-flash","rank_in_archive_order":16,"of":40,"metrics":{"Pass@1":"15.33%","Pass@2":"20.87%"},"uses_additional_data":false},{"leaderboard":"/sota/on-web-bench","task":"","dataset":"Web-Bench","model":"gemini-2.0-flash-thinking","rank_in_archive_order":17,"of":40,"metrics":{"Pass@1":"14.89%","Pass@2":"19.24%"},"uses_additional_data":false},{"leaderboard":"/sota/on-web-bench","task":"","dataset":"Web-Bench","model":"gemini-pro-1.5","rank_in_archive_order":18,"of":40,"metrics":{"Pass@1":"14.78%","Pass@2":"20.87%"},"uses_additional_data":false},{"leaderboard":"/sota/on-web-bench","task":"","dataset":"Web-Bench","model":"deepseek-r1","rank_in_archive_order":19,"of":40,"metrics":{"Pass@1":"14.46%","Pass@2":"26.20%"},"uses_additional_data":false},{"leaderboard":"/sota/on-web-bench","task":"","dataset":"Web-Bench","model":"step-fun-2-16k","rank_in_archive_order":20,"of":40,"metrics":{"Pass@1":"13.70%","Pass@2":"15.87%"},"uses_additional_data":false},{"leaderboard":"/sota/on-web-bench","task":"","dataset":"Web-Bench","model":"o4-mini","rank_in_archive_order":21,"of":40,"metrics":{"Pass@1":"13.26%","Pass@2":"22.93%"},"uses_additional_data":false},{"leaderboard":"/sota/on-web-bench","task":"","dataset":"Web-Bench","model":"mistral-large-2411","rank_in_archive_order":22,"of":40,"metrics":{"Pass@1":"13.04%","Pass@2":"18.70%"},"uses_additional_data":false},{"leaderboard":"/sota/on-web-bench","task":"","dataset":"Web-Bench","model":"gemini-flash-1.5","rank_in_archive_order":23,"of":40,"metrics":{"Pass@1":"12.83%","Pass@2":"17.07%"},"uses_additional_data":false},{"leaderboard":"/sota/on-web-bench","task":"","dataset":"Web-Bench","model":"qwen-plus-2025-01-25","rank_in_archive_order":24,"of":40,"metrics":{"Pass@1":"11.85%","Pass@2":"15.11%"},"uses_additional_data":false},{"leaderboard":"/sota/on-web-bench","task":"","dataset":"Web-Bench","model":"grok-2-1212","rank_in_archive_order":25,"of":40,"metrics":{"Pass@1":"11.30%","Pass@2":"17.17%"},"uses_additional_data":false},{"leaderboard":"/sota/on-web-bench","task":"","dataset":"Web-Bench","model":"qwen-2.5-72b-instruct","rank_in_archive_order":26,"of":40,"metrics":{"Pass@1":"10.54%","Pass@2":"13.70%"},"uses_additional_data":false},{"leaderboard":"/sota/on-web-bench","task":"","dataset":"Web-Bench","model":"o1","rank_in_archive_order":27,"of":40,"metrics":{"Pass@1":"10.43%","Pass@2":"12.39%"},"uses_additional_data":false},{"leaderboard":"/sota/on-web-bench","task":"","dataset":"Web-Bench","model":"gemma-3-27b","rank_in_archive_order":28,"of":40,"metrics":{"Pass@1":"9.89%","Pass@2":"11.85%"},"uses_additional_data":false},{"leaderboard":"/sota/on-web-bench","task":"","dataset":"Web-Bench","model":"o3-mini","rank_in_archive_order":29,"of":40,"metrics":{"Pass@1":"9.13%","Pass@2":"14.24%"},"uses_additional_data":false},{"leaderboard":"/sota/on-web-bench","task":"","dataset":"Web-Bench","model":"gpt-4o-mini","rank_in_archive_order":30,"of":40,"metrics":{"Pass@1":"8.48%","Pass@2":"13.04%"},"uses_additional_data":false},{"leaderboard":"/sota/on-web-bench","task":"","dataset":"Web-Bench","model":"sense-chat-5","rank_in_archive_order":31,"of":40,"metrics":{"Pass@1":"8.48%","Pass@2":"12.72%"},"uses_additional_data":false},{"leaderboard":"/sota/on-web-bench","task":"","dataset":"Web-Bench","model":"minimax-text","rank_in_archive_order":32,"of":40,"metrics":{"Pass@1":"8.48%","Pass@2":"10.76%"},"uses_additional_data":false},{"leaderboard":"/sota/on-web-bench","task":"","dataset":"Web-Bench","model":"360-gpt2-o1","rank_in_archive_order":33,"of":40,"metrics":{"Pass@1":"8.26%","Pass@2":"14.46%"},"uses_additional_data":false},{"leaderboard":"/sota/on-web-bench","task":"","dataset":"Web-Bench","model":"GLM-4-0414","rank_in_archive_order":34,"of":40,"metrics":{"Pass@1":"7.50%","Pass@2":"9.02%"},"uses_additional_data":false},{"leaderboard":"/sota/on-web-bench","task":"","dataset":"Web-Bench","model":"gpt-4.1-nano","rank_in_archive_order":35,"of":40,"metrics":{"Pass@1":"7.07%","Pass@2":"12.28%"},"uses_additional_data":false},{"leaderboard":"/sota/on-web-bench","task":"","dataset":"Web-Bench","model":"llama-3.3","rank_in_archive_order":36,"of":40,"metrics":{"Pass@1":"6.63%","Pass@2":"9.57%"},"uses_additional_data":false},{"leaderboard":"/sota/on-web-bench","task":"","dataset":"Web-Bench","model":"moonshot-kimi-latest","rank_in_archive_order":37,"of":40,"metrics":{"Pass@1":"5.22%","Pass@2":"11.85%"},"uses_additional_data":false},{"leaderboard":"/sota/on-web-bench","task":"","dataset":"Web-Bench","model":"llama-4 Scout","rank_in_archive_order":38,"of":40,"metrics":{"Pass@1":"5.00%","Pass@2":"7.72%"},"uses_additional_data":false},{"leaderboard":"/sota/on-web-bench","task":"","dataset":"Web-Bench","model":"doubao-pro-1.5-32k-lite","rank_in_archive_order":39,"of":40,"metrics":{"Pass@1":"3.48%","Pass@2":"5.98%"},"uses_additional_data":false},{"leaderboard":"/sota/on-web-bench","task":"","dataset":"Web-Bench","model":"qwen-turbo-2024-11-01","rank_in_archive_order":40,"of":40,"metrics":{"Pass@1":"2.61%","Pass@2":"5.11%"},"uses_additional_data":false}],"syntology":{"syntology_url":"https://syntology.ai/paper/2505.07473","atlas_url":"https://app.syntology.ai/?focus=2505.07473","mcp":null,"developers":"https://syntology.ai/developers","held":true,"reason":"arxiv_id_in_atlas","checked_against":{"atlas_ids":549449,"atlas_ids_sha256":"afd72399dafd9906e9062927f70655952385a3d75488558e1f0484e950ab09b7","read_at":"2026-10-05T01:46:21Z"},"agent_calls":[{"tool":"get_citation_path","arguments":{"paper_1":"2505.07473"},"arguments_in":{"paper_2":"another paper's arXiv id or title"},"call":"get_citation_path(paper_1=\"2505.07473\", paper_2=\"…\")"},{"tool":"get_concepts_for_paper","arguments":{"arxiv_id":"2505.07473"},"call":"get_concepts_for_paper(arxiv_id=\"2505.07473\")"}],"by_repo":[{"repo":"bytedance/Web-Bench","kind":"official","samples":0,"ran":0,"instrument":0,"unverified":0}],"by_repo_is":"one row per repository in the archive's code links, Syntology's graph links or the samples; samples 0 means no sample is linked to this paper, a statement about Syntology's coverage, not about the repository: with harvested_for_other_papers present, Syntology harvested the repository for that many other papers, and without it, Syntology harvested nothing from it; placed_by_identical_code counts samples whose link records no repository, placed here because identical code was harvested from this repository for another paper; ran counts executions on a synthesized input and instrument counts failures of Syntology's instrument, not of the code"},"code_links_note":null,"arxiv_metadata":null,"syntology_extracted_results":null,"record_sha256":"375c8d18aa2717e38ad347816034778c8c947fe65b2e2e78c19f9e077d2b8d2e","record_changed_at":"2026-09-28","record_changed_at_basis":"first_hashed"}