{"id":12782,"date":"2024-06-10T10:39:06","date_gmt":"2024-06-10T10:39:06","guid":{"rendered":"https:\/\/dailyai.com\/?p=12782"},"modified":"2024-06-10T10:39:06","modified_gmt":"2024-06-10T10:39:06","slug":"natural-plan-benchmarking-llms-on-natural-language-planning","status":"publish","type":"post","link":"https:\/\/dailyai.com\/pt\/2024\/06\/natural-plan-benchmarking-llms-on-natural-language-planning\/","title":{"rendered":"NATURAL PLAN: Avalia\u00e7\u00e3o comparativa de LLMs no planeamento de linguagem natural"},"content":{"rendered":"<p><strong>Os investigadores do Google DeepMind desenvolveram o NATURAL PLAN, um par\u00e2metro de refer\u00eancia para avaliar a capacidade dos LLM de planearem tarefas reais com base em instru\u00e7\u00f5es de linguagem natural.<\/strong><\/p>\n<p>A pr\u00f3xima evolu\u00e7\u00e3o da IA \u00e9 fazer com que ela saia dos limites de uma plataforma de conversa\u00e7\u00e3o e assuma fun\u00e7\u00f5es de agente para concluir tarefas em plataformas em nosso nome. Mas isso \u00e9 mais dif\u00edcil do que parece.<\/p>\n<p>Planear tarefas como marcar uma reuni\u00e3o ou compilar um itiner\u00e1rio de f\u00e9rias pode parecer simples para n\u00f3s. Os seres humanos s\u00e3o bons a raciocinar atrav\u00e9s de v\u00e1rios passos e a prever se um curso de a\u00e7\u00e3o vai ou n\u00e3o atingir o objetivo desejado.<\/p>\n<p>Pode ser f\u00e1cil, mas at\u00e9 os melhores modelos de IA t\u00eam dificuldades em planear. Poder\u00edamos compar\u00e1-los para ver qual \u00e9 o melhor LLM em termos de planeamento?<\/p>\n<p>O benchmark NATURAL PLAN testa os LLMs em 3 tarefas de planeamento:<\/p>\n<ul>\n<li><strong>Planeamento da viagem<\/strong> - Planear um itiner\u00e1rio de viagem com restri\u00e7\u00f5es de voo e de destino<\/li>\n<li><strong>Planeamento de reuni\u00f5es<\/strong> - Agendar reuni\u00f5es com v\u00e1rios amigos em diferentes locais<\/li>\n<li><strong>Programa\u00e7\u00e3o do calend\u00e1rio<\/strong> - Agendamento de reuni\u00f5es de trabalho entre v\u00e1rias pessoas, tendo em conta os hor\u00e1rios existentes e v\u00e1rios condicionalismos<\/li>\n<\/ul>\n<p>A experi\u00eancia come\u00e7ou com um est\u00edmulo de poucos disparos, em que os modelos receberam 5 exemplos de est\u00edmulos e as correspondentes respostas correctas. Em seguida, foram-lhes dadas instru\u00e7\u00f5es de planeamento de dificuldade vari\u00e1vel.<\/p>\n<p>Eis um exemplo de um pedido e de uma solu\u00e7\u00e3o fornecidos como exemplo para os modelos:<\/p>\n<figure id=\"attachment_12784\" aria-describedby=\"caption-attachment-12784\" style=\"width: 1342px\" class=\"wp-caption aligncenter\"><img loading=\"lazy\" decoding=\"async\" class=\"wp-image-12784 size-full\" src=\"https:\/\/dailyai.com\/wp-content\/uploads\/2024\/06\/NATURAL-PLAN-Prompt-example.png\" alt=\"\" width=\"1342\" height=\"808\" srcset=\"https:\/\/dailyai.com\/wp-content\/uploads\/2024\/06\/NATURAL-PLAN-Prompt-example.png 1342w, https:\/\/dailyai.com\/wp-content\/uploads\/2024\/06\/NATURAL-PLAN-Prompt-example-300x181.png 300w, https:\/\/dailyai.com\/wp-content\/uploads\/2024\/06\/NATURAL-PLAN-Prompt-example-1024x617.png 1024w, https:\/\/dailyai.com\/wp-content\/uploads\/2024\/06\/NATURAL-PLAN-Prompt-example-768x462.png 768w, https:\/\/dailyai.com\/wp-content\/uploads\/2024\/06\/NATURAL-PLAN-Prompt-example-18x12.png 18w, https:\/\/dailyai.com\/wp-content\/uploads\/2024\/06\/NATURAL-PLAN-Prompt-example-60x36.png 60w\" sizes=\"auto, (max-width: 1342px) 100vw, 1342px\" \/><figcaption id=\"caption-attachment-12784\" class=\"wp-caption-text\">Um exemplo de um pedido e de uma solu\u00e7\u00e3o utilizados na experi\u00eancia de Planeamento de Viagens. Fonte: arXiv<\/figcaption><\/figure>\n<h2>Resultados<\/h2>\n<p>Os investigadores testaram o GPT-3.5, GPT-4, <a href=\"https:\/\/dailyai.com\/pt\/2024\/05\/everything-you-need-to-know-about-openais-new-flagship-model-gpt-4o\/\">GPT-4o<\/a>, Gemini 1.5 Flash, e <a href=\"https:\/\/dailyai.com\/pt\/2024\/02\/google-plays-another-ai-card-in-the-form-of-gemini-1-5-pro\/\"><span class=\"noTranslate\" data-no-translation=\"\">Gemini<\/span> 1.5 Pro<\/a>nenhum dos quais teve um desempenho muito bom nestes testes.<\/p>\n<p>No entanto, os resultados devem ter ca\u00eddo bem no escrit\u00f3rio da DeepMind, uma vez que o Gemini 1.5 Pro ficou em primeiro lugar.<\/p>\n<figure id=\"attachment_12785\" aria-describedby=\"caption-attachment-12785\" style=\"width: 1302px\" class=\"wp-caption aligncenter\"><img loading=\"lazy\" decoding=\"async\" class=\"wp-image-12785 size-full\" src=\"https:\/\/dailyai.com\/wp-content\/uploads\/2024\/06\/NATURAL-PLAN-results.png\" alt=\"\" width=\"1302\" height=\"204\" srcset=\"https:\/\/dailyai.com\/wp-content\/uploads\/2024\/06\/NATURAL-PLAN-results.png 1302w, https:\/\/dailyai.com\/wp-content\/uploads\/2024\/06\/NATURAL-PLAN-results-300x47.png 300w, https:\/\/dailyai.com\/wp-content\/uploads\/2024\/06\/NATURAL-PLAN-results-1024x160.png 1024w, https:\/\/dailyai.com\/wp-content\/uploads\/2024\/06\/NATURAL-PLAN-results-768x120.png 768w, https:\/\/dailyai.com\/wp-content\/uploads\/2024\/06\/NATURAL-PLAN-results-18x3.png 18w, https:\/\/dailyai.com\/wp-content\/uploads\/2024\/06\/NATURAL-PLAN-results-60x9.png 60w\" sizes=\"auto, (max-width: 1302px) 100vw, 1302px\" \/><figcaption id=\"caption-attachment-12785\" class=\"wp-caption-text\">Resultados do benchmark NATURAL PLAN. Fonte: arXiv<\/figcaption><\/figure>\n<p>Como era de esperar, os resultados pioraram exponencialmente com as solicita\u00e7\u00f5es mais complexas, em que o n\u00famero de pessoas ou cidades foi aumentado. Por exemplo, veja como a precis\u00e3o diminuiu rapidamente \u00e0 medida que mais pessoas foram adicionadas ao teste de planeamento de reuni\u00f5es.<\/p>\n<figure id=\"attachment_12786\" aria-describedby=\"caption-attachment-12786\" style=\"width: 1330px\" class=\"wp-caption aligncenter\"><img loading=\"lazy\" decoding=\"async\" class=\"wp-image-12786 size-full\" src=\"https:\/\/dailyai.com\/wp-content\/uploads\/2024\/06\/NATURAL-PLANNING-results-vs-complexity.png\" alt=\"\" width=\"1330\" height=\"530\" srcset=\"https:\/\/dailyai.com\/wp-content\/uploads\/2024\/06\/NATURAL-PLANNING-results-vs-complexity.png 1330w, https:\/\/dailyai.com\/wp-content\/uploads\/2024\/06\/NATURAL-PLANNING-results-vs-complexity-300x120.png 300w, https:\/\/dailyai.com\/wp-content\/uploads\/2024\/06\/NATURAL-PLANNING-results-vs-complexity-1024x408.png 1024w, https:\/\/dailyai.com\/wp-content\/uploads\/2024\/06\/NATURAL-PLANNING-results-vs-complexity-768x306.png 768w, https:\/\/dailyai.com\/wp-content\/uploads\/2024\/06\/NATURAL-PLANNING-results-vs-complexity-18x7.png 18w, https:\/\/dailyai.com\/wp-content\/uploads\/2024\/06\/NATURAL-PLANNING-results-vs-complexity-60x24.png 60w\" sizes=\"auto, (max-width: 1330px) 100vw, 1330px\" \/><figcaption id=\"caption-attachment-12786\" class=\"wp-caption-text\">A precis\u00e3o dos resultados no teste de Planeamento de Reuni\u00f5es degradou-se exponencialmente \u00e0 medida que as instru\u00e7\u00f5es se tornaram mais complexas. Fonte: arXiv<\/figcaption><\/figure>\n<p>Poder\u00e1 a solicita\u00e7\u00e3o de m\u00faltiplos disparos resultar numa maior precis\u00e3o? Os resultados da investiga\u00e7\u00e3o indicam que sim, mas apenas se o modelo tiver uma janela de contexto suficientemente grande.<\/p>\n<p>A janela de contexto maior do Gemini 1.5 Pro permite-lhe aproveitar mais exemplos no contexto do que os modelos GPT.<\/p>\n<p>Os investigadores descobriram que, no planeamento de viagens, o aumento do n\u00famero de disparos de 1 para 800 melhora a precis\u00e3o do Gemini Pro 1.5 de 2,7% para 39,9%.<\/p>\n<p><a href=\"https:\/\/arxiv.org\/pdf\/2406.04520\" target=\"_blank\" rel=\"noopener\">O jornal<\/a> Estes resultados mostram a promessa de um planeamento em contexto, em que as capacidades de contexto longo permitem aos LLM tirar partido de mais contexto para melhorar o planeamento\".<\/p>\n<p>Um resultado estranho foi o facto de o GPT-4o ser realmente mau no planeamento de viagens. Os investigadores descobriram que tinha dificuldade em \"compreender e respeitar a conetividade dos voos e as restri\u00e7\u00f5es de datas de viagem\".<\/p>\n<p>Outro resultado estranho foi o facto de a auto-corre\u00e7\u00e3o ter levado a uma queda significativa do desempenho de todos os modelos. Quando os modelos foram convidados a verificar o seu trabalho e a fazer correc\u00e7\u00f5es, cometeram mais erros.<\/p>\n<p>Curiosamente, os modelos mais fortes, como o GPT-4 e o Gemini 1.5 Pro, sofreram perdas maiores do que o GPT-3.5 quando se auto-corrigiram.<\/p>\n<p>A IA ag\u00eantica \u00e9 uma perspetiva empolgante e j\u00e1 estamos a ver alguns casos de utiliza\u00e7\u00e3o pr\u00e1tica em <a href=\"https:\/\/dailyai.com\/pt\/2024\/05\/ai-agents-multimodal-phi-3-unveiled-at-microsoft-build-2024\/\">Microsoft <span class=\"noTranslate\" data-no-translation=\"\">Copilot<\/span> agentes<\/a>.<\/p>\n<p>Mas os resultados dos testes de refer\u00eancia do NATURAL PLAN mostram que ainda temos um longo caminho a percorrer at\u00e9 que a IA possa lidar com um planeamento mais complexo.<\/p>\n<p>Os investigadores da DeepMind conclu\u00edram que \"o PLANO NATURAL \u00e9 muito dif\u00edcil de resolver pelos modelos mais avan\u00e7ados\".<\/p>\n<p>Parece que a IA ainda n\u00e3o vai substituir os agentes de viagens e os assistentes pessoais.<\/p>","protected":false},"excerpt":{"rendered":"<p>Os investigadores da Google DeepMind desenvolveram o NATURAL PLAN, um par\u00e2metro de refer\u00eancia para avaliar a capacidade dos LLM para planear tarefas reais com base em instru\u00e7\u00f5es de linguagem natural. A pr\u00f3xima evolu\u00e7\u00e3o da IA \u00e9 fazer com que ela saia dos limites de uma plataforma de conversa\u00e7\u00e3o e assuma fun\u00e7\u00f5es de agente para realizar tarefas em plataformas em nosso nome. Mas isso \u00e9 mais dif\u00edcil do que parece. Planear tarefas como a marca\u00e7\u00e3o de uma reuni\u00e3o ou a compila\u00e7\u00e3o de um itiner\u00e1rio de f\u00e9rias pode parecer simples para n\u00f3s. Os seres humanos s\u00e3o bons a raciocinar atrav\u00e9s de v\u00e1rios passos e a prever se um curso de a\u00e7\u00e3o ir\u00e1 ou n\u00e3o atingir o objetivo desejado. Pode achar que<\/p>","protected":false},"author":6,"featured_media":12787,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[84],"tags":[147,118],"class_list":["post-12782","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-industry","tag-deepmind","tag-llms"],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v27.4 - https:\/\/yoast.com\/product\/yoast-seo-wordpress\/ -->\n<title>NATURAL PLAN: Benchmarking LLMs on natural language planning | DailyAI<\/title>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/dailyai.com\/pt\/2024\/06\/natural-plan-benchmarking-llms-on-natural-language-planning\/\" \/>\n<meta property=\"og:locale\" content=\"pt_PT\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"NATURAL PLAN: Benchmarking LLMs on natural language planning | DailyAI\" \/>\n<meta property=\"og:description\" content=\"Google DeepMind researchers developed NATURAL PLAN, a benchmark for evaluating the capability of LLMs to plan real-world tasks based on natural language prompts. The next evolution of AI is to have it leave the confines of a chat platform and take on agentic roles to complete tasks across platforms on our behalf. But that\u2019s harder than it sounds. Planning tasks like scheduling a meeting or compiling a holiday itinerary might seem simple for us. Humans are good at reasoning through multiple steps and predicting whether a course of action will accomplish the desired objective or not. You might find that\" \/>\n<meta property=\"og:url\" content=\"https:\/\/dailyai.com\/pt\/2024\/06\/natural-plan-benchmarking-llms-on-natural-language-planning\/\" \/>\n<meta property=\"og:site_name\" content=\"DailyAI\" \/>\n<meta property=\"article:published_time\" content=\"2024-06-10T10:39:06+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/dailyai.com\/wp-content\/uploads\/2024\/06\/Planning.webp\" \/>\n\t<meta property=\"og:image:width\" content=\"1792\" \/>\n\t<meta property=\"og:image:height\" content=\"1024\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/webp\" \/>\n<meta name=\"author\" content=\"Eugene van der Watt\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:creator\" content=\"@DailyAIOfficial\" \/>\n<meta name=\"twitter:site\" content=\"@DailyAIOfficial\" \/>\n<meta name=\"twitter:label1\" content=\"Escrito por\" \/>\n\t<meta name=\"twitter:data1\" content=\"Eugene van der Watt\" \/>\n\t<meta name=\"twitter:label2\" content=\"Tempo estimado de leitura\" \/>\n\t<meta name=\"twitter:data2\" content=\"4 minutos\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\\\/\\\/schema.org\",\"@graph\":[{\"@type\":\"NewsArticle\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/2024\\\/06\\\/natural-plan-benchmarking-llms-on-natural-language-planning\\\/#article\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/2024\\\/06\\\/natural-plan-benchmarking-llms-on-natural-language-planning\\\/\"},\"author\":{\"name\":\"Eugene van der Watt\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/#\\\/schema\\\/person\\\/7ce525c6d0c79838b7cc7cde96993cfa\"},\"headline\":\"NATURAL PLAN: Benchmarking LLMs on natural language planning\",\"datePublished\":\"2024-06-10T10:39:06+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/2024\\\/06\\\/natural-plan-benchmarking-llms-on-natural-language-planning\\\/\"},\"wordCount\":606,\"publisher\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/#organization\"},\"image\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/2024\\\/06\\\/natural-plan-benchmarking-llms-on-natural-language-planning\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/dailyai.com\\\/wp-content\\\/uploads\\\/2024\\\/06\\\/Planning.webp\",\"keywords\":[\"DeepMind\",\"LLMS\"],\"articleSection\":[\"Industry\"],\"inLanguage\":\"pt-PT\"},{\"@type\":\"WebPage\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/2024\\\/06\\\/natural-plan-benchmarking-llms-on-natural-language-planning\\\/\",\"url\":\"https:\\\/\\\/dailyai.com\\\/2024\\\/06\\\/natural-plan-benchmarking-llms-on-natural-language-planning\\\/\",\"name\":\"NATURAL PLAN: Benchmarking LLMs on natural language planning | DailyAI\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/2024\\\/06\\\/natural-plan-benchmarking-llms-on-natural-language-planning\\\/#primaryimage\"},\"image\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/2024\\\/06\\\/natural-plan-benchmarking-llms-on-natural-language-planning\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/dailyai.com\\\/wp-content\\\/uploads\\\/2024\\\/06\\\/Planning.webp\",\"datePublished\":\"2024-06-10T10:39:06+00:00\",\"breadcrumb\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/2024\\\/06\\\/natural-plan-benchmarking-llms-on-natural-language-planning\\\/#breadcrumb\"},\"inLanguage\":\"pt-PT\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\\\/\\\/dailyai.com\\\/2024\\\/06\\\/natural-plan-benchmarking-llms-on-natural-language-planning\\\/\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"pt-PT\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/2024\\\/06\\\/natural-plan-benchmarking-llms-on-natural-language-planning\\\/#primaryimage\",\"url\":\"https:\\\/\\\/dailyai.com\\\/wp-content\\\/uploads\\\/2024\\\/06\\\/Planning.webp\",\"contentUrl\":\"https:\\\/\\\/dailyai.com\\\/wp-content\\\/uploads\\\/2024\\\/06\\\/Planning.webp\",\"width\":1792,\"height\":1024},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/2024\\\/06\\\/natural-plan-benchmarking-llms-on-natural-language-planning\\\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\\\/\\\/dailyai.com\\\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"NATURAL PLAN: Benchmarking LLMs on natural language planning\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/#website\",\"url\":\"https:\\\/\\\/dailyai.com\\\/\",\"name\":\"DailyAI\",\"description\":\"Your Daily Dose of AI News\",\"publisher\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/#organization\"},\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\\\/\\\/dailyai.com\\\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"pt-PT\"},{\"@type\":\"Organization\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/#organization\",\"name\":\"DailyAI\",\"url\":\"https:\\\/\\\/dailyai.com\\\/\",\"logo\":{\"@type\":\"ImageObject\",\"inLanguage\":\"pt-PT\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/#\\\/schema\\\/logo\\\/image\\\/\",\"url\":\"https:\\\/\\\/dailyai.com\\\/wp-content\\\/uploads\\\/2023\\\/06\\\/Daily-Ai_TL_colour.png\",\"contentUrl\":\"https:\\\/\\\/dailyai.com\\\/wp-content\\\/uploads\\\/2023\\\/06\\\/Daily-Ai_TL_colour.png\",\"width\":4501,\"height\":934,\"caption\":\"DailyAI\"},\"image\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/#\\\/schema\\\/logo\\\/image\\\/\"},\"sameAs\":[\"https:\\\/\\\/x.com\\\/DailyAIOfficial\",\"https:\\\/\\\/www.linkedin.com\\\/company\\\/dailyaiofficial\\\/\",\"https:\\\/\\\/www.youtube.com\\\/@DailyAIOfficial\"]},{\"@type\":\"Person\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/#\\\/schema\\\/person\\\/7ce525c6d0c79838b7cc7cde96993cfa\",\"name\":\"Eugene van der Watt\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"pt-PT\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/wp-content\\\/uploads\\\/2023\\\/07\\\/Eugine_Profile_Picture-96x96.png\",\"url\":\"https:\\\/\\\/dailyai.com\\\/wp-content\\\/uploads\\\/2023\\\/07\\\/Eugine_Profile_Picture-96x96.png\",\"contentUrl\":\"https:\\\/\\\/dailyai.com\\\/wp-content\\\/uploads\\\/2023\\\/07\\\/Eugine_Profile_Picture-96x96.png\",\"caption\":\"Eugene van der Watt\"},\"description\":\"Eugene comes from an electronic engineering background and loves all things tech. When he takes a break from consuming AI news you'll find him at the snooker table.\",\"sameAs\":[\"www.linkedin.com\\\/in\\\/eugene-van-der-watt-16828119\"],\"url\":\"https:\\\/\\\/dailyai.com\\\/pt\\\/author\\\/eugene\\\/\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"PLANO NATURAL: Avalia\u00e7\u00e3o comparativa de LLMs no planeamento de linguagem natural | DailyAI","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/dailyai.com\/pt\/2024\/06\/natural-plan-benchmarking-llms-on-natural-language-planning\/","og_locale":"pt_PT","og_type":"article","og_title":"NATURAL PLAN: Benchmarking LLMs on natural language planning | DailyAI","og_description":"Google DeepMind researchers developed NATURAL PLAN, a benchmark for evaluating the capability of LLMs to plan real-world tasks based on natural language prompts. The next evolution of AI is to have it leave the confines of a chat platform and take on agentic roles to complete tasks across platforms on our behalf. But that\u2019s harder than it sounds. Planning tasks like scheduling a meeting or compiling a holiday itinerary might seem simple for us. Humans are good at reasoning through multiple steps and predicting whether a course of action will accomplish the desired objective or not. You might find that","og_url":"https:\/\/dailyai.com\/pt\/2024\/06\/natural-plan-benchmarking-llms-on-natural-language-planning\/","og_site_name":"DailyAI","article_published_time":"2024-06-10T10:39:06+00:00","og_image":[{"width":1792,"height":1024,"url":"https:\/\/dailyai.com\/wp-content\/uploads\/2024\/06\/Planning.webp","type":"image\/webp"}],"author":"Eugene van der Watt","twitter_card":"summary_large_image","twitter_creator":"@DailyAIOfficial","twitter_site":"@DailyAIOfficial","twitter_misc":{"Escrito por":"Eugene van der Watt","Tempo estimado de leitura":"4 minutos"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"NewsArticle","@id":"https:\/\/dailyai.com\/2024\/06\/natural-plan-benchmarking-llms-on-natural-language-planning\/#article","isPartOf":{"@id":"https:\/\/dailyai.com\/2024\/06\/natural-plan-benchmarking-llms-on-natural-language-planning\/"},"author":{"name":"Eugene van der Watt","@id":"https:\/\/dailyai.com\/#\/schema\/person\/7ce525c6d0c79838b7cc7cde96993cfa"},"headline":"NATURAL PLAN: Benchmarking LLMs on natural language planning","datePublished":"2024-06-10T10:39:06+00:00","mainEntityOfPage":{"@id":"https:\/\/dailyai.com\/2024\/06\/natural-plan-benchmarking-llms-on-natural-language-planning\/"},"wordCount":606,"publisher":{"@id":"https:\/\/dailyai.com\/#organization"},"image":{"@id":"https:\/\/dailyai.com\/2024\/06\/natural-plan-benchmarking-llms-on-natural-language-planning\/#primaryimage"},"thumbnailUrl":"https:\/\/dailyai.com\/wp-content\/uploads\/2024\/06\/Planning.webp","keywords":["DeepMind","LLMS"],"articleSection":["Industry"],"inLanguage":"pt-PT"},{"@type":"WebPage","@id":"https:\/\/dailyai.com\/2024\/06\/natural-plan-benchmarking-llms-on-natural-language-planning\/","url":"https:\/\/dailyai.com\/2024\/06\/natural-plan-benchmarking-llms-on-natural-language-planning\/","name":"PLANO NATURAL: Avalia\u00e7\u00e3o comparativa de LLMs no planeamento de linguagem natural | DailyAI","isPartOf":{"@id":"https:\/\/dailyai.com\/#website"},"primaryImageOfPage":{"@id":"https:\/\/dailyai.com\/2024\/06\/natural-plan-benchmarking-llms-on-natural-language-planning\/#primaryimage"},"image":{"@id":"https:\/\/dailyai.com\/2024\/06\/natural-plan-benchmarking-llms-on-natural-language-planning\/#primaryimage"},"thumbnailUrl":"https:\/\/dailyai.com\/wp-content\/uploads\/2024\/06\/Planning.webp","datePublished":"2024-06-10T10:39:06+00:00","breadcrumb":{"@id":"https:\/\/dailyai.com\/2024\/06\/natural-plan-benchmarking-llms-on-natural-language-planning\/#breadcrumb"},"inLanguage":"pt-PT","potentialAction":[{"@type":"ReadAction","target":["https:\/\/dailyai.com\/2024\/06\/natural-plan-benchmarking-llms-on-natural-language-planning\/"]}]},{"@type":"ImageObject","inLanguage":"pt-PT","@id":"https:\/\/dailyai.com\/2024\/06\/natural-plan-benchmarking-llms-on-natural-language-planning\/#primaryimage","url":"https:\/\/dailyai.com\/wp-content\/uploads\/2024\/06\/Planning.webp","contentUrl":"https:\/\/dailyai.com\/wp-content\/uploads\/2024\/06\/Planning.webp","width":1792,"height":1024},{"@type":"BreadcrumbList","@id":"https:\/\/dailyai.com\/2024\/06\/natural-plan-benchmarking-llms-on-natural-language-planning\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/dailyai.com\/"},{"@type":"ListItem","position":2,"name":"NATURAL PLAN: Benchmarking LLMs on natural language planning"}]},{"@type":"WebSite","@id":"https:\/\/dailyai.com\/#website","url":"https:\/\/dailyai.com\/","name":"DailyAI","description":"A sua dose di\u00e1ria de not\u00edcias sobre IA","publisher":{"@id":"https:\/\/dailyai.com\/#organization"},"potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/dailyai.com\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"pt-PT"},{"@type":"Organization","@id":"https:\/\/dailyai.com\/#organization","name":"DailyAI","url":"https:\/\/dailyai.com\/","logo":{"@type":"ImageObject","inLanguage":"pt-PT","@id":"https:\/\/dailyai.com\/#\/schema\/logo\/image\/","url":"https:\/\/dailyai.com\/wp-content\/uploads\/2023\/06\/Daily-Ai_TL_colour.png","contentUrl":"https:\/\/dailyai.com\/wp-content\/uploads\/2023\/06\/Daily-Ai_TL_colour.png","width":4501,"height":934,"caption":"DailyAI"},"image":{"@id":"https:\/\/dailyai.com\/#\/schema\/logo\/image\/"},"sameAs":["https:\/\/x.com\/DailyAIOfficial","https:\/\/www.linkedin.com\/company\/dailyaiofficial\/","https:\/\/www.youtube.com\/@DailyAIOfficial"]},{"@type":"Person","@id":"https:\/\/dailyai.com\/#\/schema\/person\/7ce525c6d0c79838b7cc7cde96993cfa","name":"Eugene van der Watt","image":{"@type":"ImageObject","inLanguage":"pt-PT","@id":"https:\/\/dailyai.com\/wp-content\/uploads\/2023\/07\/Eugine_Profile_Picture-96x96.png","url":"https:\/\/dailyai.com\/wp-content\/uploads\/2023\/07\/Eugine_Profile_Picture-96x96.png","contentUrl":"https:\/\/dailyai.com\/wp-content\/uploads\/2023\/07\/Eugine_Profile_Picture-96x96.png","caption":"Eugene van der Watt"},"description":"Eugene vem de uma forma\u00e7\u00e3o em engenharia eletr\u00f3nica e adora tudo o que \u00e9 tecnologia. Quando faz uma pausa no consumo de not\u00edcias sobre IA, pode encontr\u00e1-lo \u00e0 mesa de snooker.","sameAs":["www.linkedin.com\/in\/eugene-van-der-watt-16828119"],"url":"https:\/\/dailyai.com\/pt\/author\/eugene\/"}]}},"_links":{"self":[{"href":"https:\/\/dailyai.com\/pt\/wp-json\/wp\/v2\/posts\/12782","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/dailyai.com\/pt\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/dailyai.com\/pt\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/dailyai.com\/pt\/wp-json\/wp\/v2\/users\/6"}],"replies":[{"embeddable":true,"href":"https:\/\/dailyai.com\/pt\/wp-json\/wp\/v2\/comments?post=12782"}],"version-history":[{"count":3,"href":"https:\/\/dailyai.com\/pt\/wp-json\/wp\/v2\/posts\/12782\/revisions"}],"predecessor-version":[{"id":12789,"href":"https:\/\/dailyai.com\/pt\/wp-json\/wp\/v2\/posts\/12782\/revisions\/12789"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/dailyai.com\/pt\/wp-json\/wp\/v2\/media\/12787"}],"wp:attachment":[{"href":"https:\/\/dailyai.com\/pt\/wp-json\/wp\/v2\/media?parent=12782"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/dailyai.com\/pt\/wp-json\/wp\/v2\/categories?post=12782"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/dailyai.com\/pt\/wp-json\/wp\/v2\/tags?post=12782"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}