RECITATION just started out the blue

Yesterday I processed 3000 pdfs without issue. To date 24k. Today I’m running into a ton of RECITATION finish_reason errors. No idea why.

The documents are freely available on the FDA website

Its citing this patent application as the source

Again, freely available not copyrighted document which I’m not using.

Model: gemini-1.5-flash-002

Using generate_content to parse content from the documents. Haven’t changed my prompt in ages.

Hi @Martin_Phee,

If you see the model stops generating output due to the RECITATION reason, this means the model output may resemble certain data or text derived from copyrighted sources within the foundational training data of the LLM. To counter this, try to make prompt / context as unique as possible and use a higher temperature.