AIToday
Large Language ModelsTop Companies' AI MovesTop Companies AIPublished: Sep 3, 2026, 06:31 JST1 min read

Dentsu and SB Intuitions Build 116K Ad Copy Evaluation Dataset

Dentsu and SB Intuitions Build 116K Ad Copy Evaluation Dataset

3 Key Points

  1. What happened

    Dentsu, Dentsu Digital, and SB Intuitions announced that they have built a dataset of over 110,000 ad copy evaluations by 64 Dentsu copywriters, as part of their joint research on a Japanese copywriting-focused generative AI.

  2. Why it matters

    In tested categories, the AI's automated evaluation showed a positive correlation with human copywriter judgments, suggesting it can learn copywriting expertise. However, the correlation dropped for untrained categories, revealing that prompt adjustment alone is not enough for broad application.

  3. What to watch

    The research paper, titled "Verification of LLM-as-a-Judge for Japanese catch copy evaluation," will be presented at FIT2026, held from September 2 to 4, 2026, in Kitakyushu.

Ask the AI about this article →

Summaries like this, in your inbox every morning.

Context & Analysis

The joint research combines Dentsu's creative expertise and Dentsu Digital's AI solutions with SB Intuitions' domestic generative AI "Sarashina," built on SoftBank's AI computing infrastructure. This collaboration aims to create a generative AI that understands Japanese nuances and can produce emotionally resonant ad copy. The dataset and systematic evaluation criteria are early steps toward practical automated evaluation, but the drop in performance for untrained industries highlights the need for more data and model refinement. The findings will be shared with the academic community at FIT2026.

FAQ
What does LLM-as-a-Judge mean in this context?
It refers to using a generative AI model to automatically evaluate Japanese ad copy, incorporating the copywriters' evaluation criteria.
Where will the research be presented?
The research will be presented as a paper at the 25th Information Science and Technology Forum (FIT2026), held from September 2 to 4, 2026.
Top Companies AIRead Original Article

Get the latest Large Language Models news every morning

For example, today's edition would include:

  • Wired writer unleashes rogue AI agent on home network — finds vulnerabilities, hacks PCWIRED AI · 3m ago
  • Databricks Adaptive Instructed-Retriever doubles speed vs Claude Sonnet 5SiliconANGLE AI · 3h ago
  • Lightbits launches Inferra to boost AI inference performanceSiliconANGLE AI · 3h ago

AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.

Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. Q&As are published on this page for other readers too.

Related Articles

Next articleMicrosoft IQ and Copilot Cowork/Scout: AI agents move from chat to autonomous execution