{"id":593,"date":"2026-06-26T15:20:13","date_gmt":"2026-06-26T07:20:13","guid":{"rendered":"http:\/\/www.imlip.org\/home\/?p=593"},"modified":"2026-06-29T16:56:12","modified_gmt":"2026-06-29T08:56:12","slug":"imlip2026%e7%89%b9%e9%82%80%e6%8a%a5%e5%91%8a-%e9%bb%84%e8%be%89%ef%bc%9arethinking-evaluation-paradigmsllms-lrms-and-agentic-frameworks-for-machine-translation-and-beyond","status":"publish","type":"post","link":"http:\/\/www.imlip.org\/home\/2026\/06\/26\/imlip2026%e7%89%b9%e9%82%80%e6%8a%a5%e5%91%8a-%e9%bb%84%e8%be%89%ef%bc%9arethinking-evaluation-paradigmsllms-lrms-and-agentic-frameworks-for-machine-translation-and-beyond\/","title":{"rendered":"IMLIP2026\u7279\u9080\u62a5\u544a | \u9ec4\u8f89\uff1aRethinking Evaluation Paradigms:LLMs, LRMs, and Agentic Frameworks for Machine Translation and Beyond"},"content":{"rendered":"\t\t
\n\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t
\n
\n
\n
\n
\n
\n
\n

\u00a0<\/span><\/p>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n

\n
\n

\u4f1a\u8bae\u7b80\u4ecb<\/span><\/span><\/strong><\/p>\n<\/section>\n<\/section>\n

\n
\n
\n
\n

\u00a0<\/span><\/p>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n

\n
\n
<\/section>\n

\"\"<\/p>\n

\n

\u00a0<\/p>\n

\u56fd\u9645\u591a\u8bed\u79cd\u667a\u80fd\u4fe1\u606f\u5904\u7406\u4f1a\u8bae\uff08IMLIP\uff09\u662f\u591a\u8bed\u79cd\u667a\u80fd\u4fe1\u606f\u5904\u7406\u9886\u57df\u7684\u56fd\u9645\u9876\u7ea7\u5b66\u672f\u4f1a\u8bae\uff0c\u662f\u5177\u6709\u56fd\u9645\u5f71\u54cd\u529b\u7684\u591a\u8bed\u79cd\u667a\u80fd\u4fe1\u606f\u5904\u7406\u9886\u57df\u4ea7\u5b66\u7814\u7684\u9ad8\u7aef\u4f1a\u8bae\uff0c\u65e8\u5728\u4e3a\u56fd\u5185\u5916\u5b66\u8005\u63d0\u4f9b\u5b66\u672f\u4ea4\u6d41\u4e0e\u5408\u4f5c\u7814\u7a76\u7684\u5e73\u53f0\uff0c\u4fc3\u8fdb\u6211\u56fd\u5c11\u6570\u6c11\u65cf\u3001\u201c\u4e00\u5e26\u4e00\u8def\u201d\u6cbf\u7ebf\u56fd\u5bb6\u7b49\u8bed\u8a00\u5b66\u7814\u7a76\u548c\u81ea\u7136\u8bed\u8a00\u5904\u7406\u7684\u5b66\u672f\u7814\u7a76\u3002<\/span><\/p>\n

2026\u5e74\u7b2c\u4e09\u5c4a\u56fd\u9645\u591a\u8bed\u79cd\u667a\u80fd\u4fe1\u606f\u5904\u7406\u4f1a\u8bae\uff08IMLIP 2026\uff09\u7531\u4e2d\u56fd\u4eba\u5de5\u667a\u80fd\u5b66\u4f1a\u4e3b\u529e\uff0c\u4e2d\u56fd\u4eba\u5de5\u667a\u80fd\u5b66\u4f1a\u591a\u8bed\u79cd\u667a\u80fd\u4fe1\u606f\u5904\u7406\u4e13\u59d4\u4f1a\u3001\u6606\u660e\u7406\u5de5\u5927\u5b66\u5171\u540c\u627f\u529e\uff0c\u5c06\u4e8e2026\u5e74<\/span>7\u670824-26\u65e5\u5728<\/span>\u4e91\u5357\u6606\u660e<\/span><\/strong>\u53ec\u5f00\uff0c\u9884\u8ba1\u53c2\u4f1a\u4eba\u6570300-500\u4eba\u3002<\/span><\/p>\n

\u4f1a\u8bae\u5c06\u9080\u8bf7\u9662\u58eb\u3001\u56fd\u9645\u77e5\u540d\u8bed\u8a00\u5b66\u5bb6\u4e0e\u591a\u8bed\u79cd\u5904\u7406\u6280\u672f\u4e13\u5bb6\u7b49\u56fd\u5185\u5916\u77e5\u540d\u4e13\u5bb6\u517110\u4f59\u540d\u4f5c\u7279\u9080\u62a5\u544a\uff0c\u4ee5\u671f\u4fc3\u8fdb\u591a\u8bed\u79cd\u8bed\u8a00\u5b66\u7814\u7a76\u3001\u667a\u80fd\u4fe1\u606f\u5904\u7406\u5b66\u672f\u754c\u3001\u5927\u6570\u636e\u548c\u4eba\u5de5\u667a\u80fd\u7b49\u9886\u57df\u7684\u4ea7\u4e1a\u754c\u548c\u5e7f\u5927\u7231\u597d\u8005\u4e4b\u95f4\u7684\u4ea4\u6d41\u3002<\/span><\/p>\n

\u00a0<\/span><\/p>\n

\u542f\u5e55\u5728\u5373\uff0c\u7cbe\u5f69\u53ef\u671f\uff01<\/span><\/span><\/strong><\/span><\/p>\n<\/section>\n<\/section>\n<\/section>\n

\n
\n
\n
\n
\n

\u00a0<\/span><\/p>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n

\n
\n

\u62a5\u544a\u9898\u76ee<\/span><\/span><\/strong><\/p>\n<\/section>\n<\/section>\n

\n
\n
\n
\n

\u00a0<\/span><\/p>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n

\n
\n

Rethinking Evaluation Paradigms:LLMs, LRMs, and Agentic Frameworks for Machine Translation and Beyond<\/span><\/span><\/strong><\/span><\/p>\n

\u00a0<\/span><\/p>\n

Abstract:<\/strong><\/p>\n

Large Language Models (LLMs) are emerging as promising evaluators for complex linguistic tasks such as machine translation, especially as translation demands become more culturally and contextually nuanced. An analysis comparing foundation models with fine-tuned systems shows that foundation models possess strong multilingual reasoning abilities and can effectively assess translation quality across diverse languages. Building on these insights, the alignment between human and model-based judgments is examined, revealing a persistent gap in consistency and interpretability. To address this, the Large Reasoning Model (LRM) framework is explored as a means of extending reasoning depth and contextual understanding in evaluation. The study further investigates agentic frameworks grounded in MQM principles and generalizes the Model-as-a-Judge concept to the training of reinforcement learning systems. Overall, the findings demonstrate that reasoning-based evaluation frameworks offer a path toward more transparent, scalable, and human-aligned methods for assessing translation and other language tasks.<\/p>\n<\/section>\n<\/section>\n

\n
\n
\n
\n
\n

\u00a0<\/p>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n

\n
\n

\u4e13\u5bb6\u4ecb\u7ecd<\/span><\/strong><\/p>\n<\/section>\n<\/section>\n

\n
\n
\n
\n

\u00a0<\/p>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n

\n
\n
\n
\"\"<\/section>\n<\/section>\n
\n

Derek F. Wong<\/span><\/span><\/strong><\/span><\/p>\n

\u9ec4\u8f89<\/span><\/span><\/strong><\/span><\/p>\n

Full Professor at the University of Macau (UM), where he leads the Natural Language Processing and Chinese\u2013Portuguese Machine Translation Laboratory (NLP2CT Lab). He serves on the boards and committees of CIPS, CCF, and AFNLP, and holds editorial roles with IEEE\/ACM TASLP, ACM TALLIP, TACL, and the ACL Rolling Review. His research has earned multiple honors, including the Macao Science and Technology Awards (2012, 2022), the FST Research Excellence Award, and both the Outstanding Academic Staff Incentive (2022) and Teaching Excellence Award (2024) from UM. He has also contributed to major NLP conferences such as ACL, NeurIPS, ICML, IJCAI, AAAI, EMNLP, NAACL, COLING, AACL, and IJCNLP in various program leadership roles.<\/p>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n

\"\"<\/section>\n
\u5b98\u7f51\u94fe\u63a5\uff1ahttp:\/\/www.imlip.org\/home\/IMLIP2026\/imlip\/imlip.html<\/section>\n<\/div>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t","protected":false},"excerpt":{"rendered":"

\u00a0 \u4f1a\u8bae\u7b80\u4ecb \u00a0 \u00a0 \u56fd\u9645\u591a\u8bed\u79cd\u667a\u80fd\u4fe1\u606f\u5904\u7406\u4f1a\u8bae\uff08IMLIP\uff09\u662f\u591a\u8bed\u79cd\u667a\u80fd\u4fe1\u606f\u5904\u7406\u9886\u57df\u7684\u56fd\u9645\u9876\u7ea7\u5b66\u672f\u4f1a\u8bae\uff0c\u662f\u5177 … \u9605\u8bfb\u66f4\u591a<\/a><\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_eb_attr":"","footnotes":""},"categories":[9,17,12],"tags":[],"class_list":["post-593","post","type-post","status-publish","format-standard","hentry","category-9","category-17","category-12"],"_links":{"self":[{"href":"http:\/\/www.imlip.org\/home\/wp-json\/wp\/v2\/posts\/593","targetHints":{"allow":["GET"]}}],"collection":[{"href":"http:\/\/www.imlip.org\/home\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"http:\/\/www.imlip.org\/home\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"http:\/\/www.imlip.org\/home\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"http:\/\/www.imlip.org\/home\/wp-json\/wp\/v2\/comments?post=593"}],"version-history":[{"count":9,"href":"http:\/\/www.imlip.org\/home\/wp-json\/wp\/v2\/posts\/593\/revisions"}],"predecessor-version":[{"id":673,"href":"http:\/\/www.imlip.org\/home\/wp-json\/wp\/v2\/posts\/593\/revisions\/673"}],"wp:attachment":[{"href":"http:\/\/www.imlip.org\/home\/wp-json\/wp\/v2\/media?parent=593"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"http:\/\/www.imlip.org\/home\/wp-json\/wp\/v2\/categories?post=593"},{"taxonomy":"post_tag","embeddable":true,"href":"http:\/\/www.imlip.org\/home\/wp-json\/wp\/v2\/tags?post=593"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}