{"id":12329,"date":"2026-09-16T17:58:56","date_gmt":"2026-09-16T09:58:56","guid":{"rendered":"https:\/\/infernews.com\/blog\/mind2dialogue-training-human-aware-language-models-through-shared-state-user-simulation\/"},"modified":"2026-09-16T17:58:56","modified_gmt":"2026-09-16T09:58:56","slug":"mind2dialogue-training-human-aware-language-models-through-shared-state-user-simulation","status":"publish","type":"post","link":"https:\/\/infernews.com\/blog\/mind2dialogue-training-human-aware-language-models-through-shared-state-user-simulation\/","title":{"rendered":"Mind2Dialogue\uff1a\u8b93\u6a21\u578b\u5b78\u6703\u63a8\u65b7\u7528\u6236\u5167\u5fc3\u72c0\u614b"},"content":{"rendered":"\n<figure class=\"wp-block-image size-large\"><img decoding=\"async\" src=\"https:\/\/infernews.com\/blog\/wp-content\/uploads\/2026\/09\/idea-promotion.jpg\" alt=\"idea-promotion\"><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\">\u73fe\u6642\u5f88\u591a\u804a\u5929\u6a21\u578b\u5728\u591a\u8f2a\u5c0d\u8a71\u4e2d\u8868\u73fe\u4e0d\u932f\uff0c\u4f46\u4e00\u5230\u300c\u9019\u500b\u7528\u6236\u5176\u5be6\u60f3\u8981\u4ec0\u9ebc\u3001\u5fc3\u60c5\u5982\u4f55\u3001\u8ddf\u4e0a\u4e00\u53e5\u8a71\u80cc\u5f8c\u7684\u8108\u7d61\u300d\u5c31\u5bb9\u6613\u5931\u53bb\u7dda\u7d22\u3002Mind2Dialogue \u76f4\u63a5\u5728\u8a13\u7df4\u968e\u6bb5\u628a\u9019\u500b\u7f3a\u53e3\u88dc\u4e0a\uff1a\u6a21\u64ec\u5668\u8ddf\u4e00\u500b Oracle \u52a9\u7406\u5171\u7528\u540c\u4e00\u500b\u6703\u96a8\u5c0d\u8a71\u6f14\u5316\u7684\u7528\u6236\u72c0\u614b\uff0cOracle \u5beb\u51fa\u7684\u56de\u8986\u5c31\u6210\u70ba\u5b78\u751f\u6a21\u578b\u7684\u8a13\u7df4\u76ee\u6a19\uff1b\u5b78\u751f\u6a21\u578b\u5728\u8a13\u7df4\u6642\u53ea\u770b\u5230 persona \u8207\u5c0d\u8a71\u7d00\u9304\uff0c\u8981\u81ea\u884c\u9084\u539f\u80cc\u5f8c\u72c0\u614b\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u6574\u500b\u6d41\u7a0b\u7531\u516d\u500b\u968e\u6bb5\u7d44\u6210\uff1a\u898f\u6a21\u5316\u7522\u751f persona \u7a2e\u5b50\u60c5\u5883\u3001\u6279\u6b21\u751f\u6210\u5c0d\u8a71\u3001\u7528 LLM \u4f5c\u8a55\u5be9\u904e\u6ffe\u4f4e\u8cea\u6a23\u672c\u3001\u5f9e\u5c0d\u8a71\u62bd\u53d6 QA \u5c0d\u3001\u8a9e\u6599\u6539\u5beb\u589e\u5ee3\uff0c\u6700\u5f8c\u505a\u5fae\u8abf\u8207\u8a55\u4f30\u3002\u8a2d\u8a08\u4e0a\u523b\u610f\u628a\u6a21\u64ec\u8005\u3001Oracle\u3001\u8a55\u5be9\u5206\u6210\u4e0d\u540c\u6a21\u578b\u7aef\u9ede\uff0c\u907f\u514d\u81ea\u6211\u8a55\u5206\uff0c\u4e5f\u5bb9\u8a31\u628a\u8f03\u5f37\u6a21\u578b\u653e\u5728 Oracle \u4f4d\u7f6e\u3002\u60c5\u5883\u985e\u578b\u6db5\u84cb\u7d42\u751f\u95dc\u4fc2\u3001\u9ad8\u983b\u4e92\u52d5\u3001\u60c5\u7dd2\u4e8b\u4ef6\u8207\u654f\u611f\u8b70\u984c\uff0c\u4f7f\u8a13\u7df4\u8a9e\u6599\u8cbc\u8fd1\u771f\u5be6\u9577\u671f\u966a\u4f34\u7684\u5834\u666f\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u5728 Qwen2.5-7B-Instruct \u4e0a\uff0c\u500b\u4eba\u5316\u6307\u6a19 PrefEval-Gen \u6bd4\u57fa\u790e\u6a21\u578b\u9ad8 33.4 \u5206\uff0cPersonaMem-v2 \u9ad8 10.0 \u5206\uff0c\u4e26\u5728\u540c\u9aa8\u67b6\u4e0b\u58d3\u904e\u52a0\u8a18\u61b6\u6a21\u7d44\u8207\u5176\u4ed6\u500b\u4eba\u5316\u65b9\u6cd5\u3002Qwen \u8207 Llama-3.1-8B-Instruct \u5728 BigToM \u7b49 Theory-of-Mind \u57fa\u6e96\u4e0a\u4e5f\u540c\u6b65\u9032\u6b65\uff0c\u4ee3\u8868\u500b\u4eba\u5316\u8207\u5fc3\u667a\u7406\u8ad6\u53ef\u80fd\u5171\u4eab\u540c\u4e00\u7a2e\u300c\u5f9e\u89c0\u5bdf\u63a8\u56de\u672a\u89c0\u5bdf\u72c0\u614b\u300d\u7684\u80fd\u529b\u3002OLMo-3-8B-Instruct \u5247\u672a\u898b\u8f49\u79fb\uff0c\u986f\u793a\u6548\u679c\u4ecd\u53d7\u9aa8\u67b6\u9810\u8a13\u7df4\u5f71\u97ff\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u96a8\u8a13\u7df4\u8cc7\u6599\u91cf\u7531\u516b\u5206\u4e4b\u4e00\u64f4\u5230\u5168\u91cf\uff0c\u4e09\u500b\u9aa8\u67b6\u7684\u500b\u4eba\u5316\u5206\u6578\u90fd\u7a69\u6b65\u4e0a\u5347\uff0c\u547c\u61c9\u300c\u8cc7\u6599\u898f\u6a21\u8db3\u5920\uff0c\u5b78\u751f\u624d\u80fd\u5b78\u6703\u72c0\u614b\u9084\u539f\u300d\u9019\u500b\u6838\u5fc3\u5047\u8a2d\u3002\u5c0d\u505a\u5c0d\u8a71\u7cfb\u7d71\u3001\u89d2\u8272 AI\u3001AI \u966a\u4f34\u6216\u5ba2\u670d\u52a9\u624b\u7684\u7814\u7a76\u5718\u968a\uff0c\u9019\u5957\u7ba1\u7dda\u53ef\u4f5c\u70ba\u4e0d\u9700\u8981\u9010\u53e5\u4eba\u5de5\u6a19\u8a3b\u7684\u500b\u4eba\u5316\u8cc7\u6599\u751f\u6210\u8d77\u9ede\uff1b\u5c0d\u95dc\u5fc3\u8a55\u6e2c\u7684\u7814\u7a76\u8005\uff0c\u5167\u9644 LLM-as-judge \u8207\u6d88\u878d\u8a2d\u5b9a\uff08full\u3001no_privilege\u3001no_state\u3001latent \u7b49\uff09\u4fbf\u65bc\u8ffd\u6eaf\u4e0d\u540c\u76e3\u7763\u8a0a\u865f\u7684\u8ca2\u737b\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>\u91cd\u9ede\u6458\u8981\uff1a<\/strong><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>\u5171\u4eab\u72c0\u614b\u6a21\u64ec<\/strong>\uff1a\u6a21\u64ec\u5668\u8207 Oracle \u5171\u7528\u6f14\u5316\u4e2d\u7684\u7528\u6236\u72c0\u614b\uff0cOracle \u56de\u8986\u6210\u70ba\u5b78\u751f\u8a13\u7df4\u76ee\u6a19<\/li>\n<li><strong>\u96f6\u4eba\u5de5\u6a19\u8a3b\u7ba1\u7dda<\/strong>\uff1a\u516d\u968e\u6bb5\u6d41\u7a0b\u5f9e persona \u7a2e\u5b50\u5230\u5fae\u8abf\u7686\u53ef\u6279\u6b21\u57f7\u884c<\/li>\n<li><strong>\u500b\u4eba\u5316\u8207 ToM \u540c\u6b65\u63d0\u5347<\/strong>\uff1aQwen \u8207 Llama \u5728 PrefEval-Gen\u3001PersonaMem-v2\u3001BigToM \u4e0a\u5747\u6709\u660e\u986f\u6539\u5584<\/li>\n<li><strong>OLMo \u672a\u898b\u8f49\u79fb<\/strong>\uff1a\u6548\u679c\u4f9d\u8cf4\u9aa8\u67b6\u9810\u8a13\u7df4\u7279\u6027\uff0cOLMo-3-8B-Instruct \u5728 ToM \u4e0a\u53cd\u800c\u4e0b\u8dcc<\/li>\n<li><strong>\u8cc7\u6599\u8d8a\u591a\u8d8a\u597d<\/strong>\uff1a\u5f9e 1\/8 \u5230\u5168\u91cf\u8a9e\u6599\uff0c\u4e09\u500b\u9aa8\u67b6\u7684\u500b\u4eba\u5316\u5206\u6578\u6301\u7e8c\u4e0a\u5347<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\"><a href=\"https:\/\/wannabeyourfriend.github.io\/mind2dialogue\/\" rel=\"noopener noreferrer\"><strong>\u9805\u76ee\u4e3b\u9801<\/strong><\/a> \u00b7 <a href=\"https:\/\/github.com\/wannabeyourfriend\/mind2dialogue\" rel=\"noopener noreferrer\"><strong>GitHub<\/strong><\/a> \u00b7 <a href=\"https:\/\/huggingface.co\/datasets\/wannabeyourfriend-hf\/mind2dialogue\" rel=\"noopener noreferrer\"><strong>\u6578\u64da\u96c6<\/strong><\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u9019\u500b\u9805\u76ee\u7528\u4e00\u5957\u5171\u4eab\u72c0\u614b\u7684\u5c0d\u8a71\u6a21\u64ec\u6d41\u7a0b\uff0c\u8b93\u6a21\u578b\u5f9e\u7d14\u5c0d\u8a71\u6b77\u53f2\u63a8\u56de\u7528\u6236\u7684\u4fe1\u5ff5\u3001\u76ee\u6a19\u548c\u60c5\u7dd2\uff0c\u5f9e\u800c\u63d0\u5347\u500b\u4eba\u5316\u8207\u5fc3\u7406\u7406\u8ad6\u80fd\u529b\u3002<\/p>\n","protected":false},"author":8,"featured_media":12328,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"wpai_generated_summary":"","footnotes":""},"categories":[133,116,127,185,40,199],"tags":[],"class_list":["post-12329","post","type-post","status-publish","format-standard","hentry","category-133","category-agentic","category-127","category-qwen","category-llama","category-dataset-"],"_links":{"self":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/12329","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/users\/8"}],"replies":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/comments?post=12329"}],"version-history":[{"count":0,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/12329\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media\/12328"}],"wp:attachment":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media?parent=12329"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/categories?post=12329"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/tags?post=12329"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}