LLM cliché highlighter is a tool designed to identify and highlight sentences in text that match known clichés commonly produced by large language models. Users can paste text directly into the interface or load text from a URL, and the tool will automatically analyze the content as it is entered. Sentences that fit recognized cliché patterns are highlighted for easy identification.
One notable feature is its ability to detect chain patterns, such as phrases following the structure "no X, no Y," and display a badge that counts the number of items in the chain. Hovering over a highlighted section reveals which specific cliché pattern was matched. The tool also provides options to load an example, clear the current input, or display only the highlighted matches, making it easier to focus on the detected patterns.
For those interested in automated testing, the tool includes self-tests that can be run headlessly using Node.js. A provided one-liner command reads the tool's HTML file, extracts the relevant implementation and test code, evaluates it, and outputs a pass/fail tally, with the process exit code reflecting the test results.
It is delivered as a web-based application with additional capabilities for command-line testing using Node.js.
In the LLM eval & observability space, LLM cliché highlighter takes a focused approach. It focuses on identifying repetitive or formulaic language generated by large language models in text. It is built as a consumer product for AI researchers and writers. LLM cliché highlighter costs nothing to use. The product ships for the web.
Simon Willison builds and maintains LLM cliché highlighter, and the product first shipped in 2024. Key capabilities include cliché detection, text highlighting, and pattern matching.
Latest indexed changes and source events
LLM Cliché Highlighter verified by the PulseGate indexer
Other apps tracked under the same category.
