IdeaLens: Detecting AI Ideas in Long-form Writing
IdeaLens detects whether a document's ideas came from a human or AI, independent of the prose.
IdeaLens detects whether a document's ideas originated with a human or an AI, regardless of who wrote the words, by converting text into outlines that pair discourse roles with brief paraphrases. It is trained on 1 million FineWeb documents using silver labels from the Pangram prose detector. In a controlled study its AI flag rate drops from 95% to 7% as models write from increasingly detailed human plans, while Pangram 4 still flags 92%; on 50 human-written stories based on AI plans it flags 68% versus 8% for Pangram 4. The authors report strong detection at low false-positive rates across 19 benchmarks and release models and labeled datasets.
- Represents documents as role-and-paraphrase outlines to focus on ideas.
- Trained on 1 million FineWeb documents with Pangram silver labels.
- AI flag rate falls from 95% to 7% as human plans grow more detailed.
- Flags 68% of 50 human stories from AI plans, versus 8% for Pangram 4.
- Maintains strong detection at low false positives across 19 benchmarks.
From the original article
1 min read at arxiv.orgWhile modern AI detectors identify who wrote the words, emerging policies on AI use increasingly hinge on a different question: who came up with the ideas? We introduce IdeaLens, a detector that identifies whether a document's ideas came from a human or AI (idea provenance), regardless of who wrote its words.
By Rishanth Rajendhran · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.