Towards Verifiable Text Generation with Symbolic References
This addresses the need for verifiable text generation in high-stakes applications, offering an incremental improvement by reducing human verification effort.
The paper tackles the problem of hallucinations in LLMs by proposing symbolically grounded generation (SymGen), which prompts LLMs to interleave output text with symbolic references to conditioning data, enabling easier manual verification; results show that LLMs can produce accurate references while maintaining fluency and factuality, and a human study confirms that annotations streamline verification.
LLMs are vulnerable to hallucinations, and thus their outputs generally require laborious human verification for high-stakes applications. To this end, we propose symbolically grounded generation (SymGen) as a simple approach for enabling easier manual validation of an LLM's output. SymGen prompts an LLM to interleave its regular output text with explicit symbolic references to fields present in some conditioning data (e.g., a table in JSON format). The references can be used to display the provenance of different spans of text in the generation, reducing the effort required for manual verification. Across a range of data-to-text and question-answering experiments, we find that LLMs are able to directly output text that makes use of accurate symbolic references while maintaining fluency and factuality. In a human study we further find that such annotations can streamline human verification of machine-generated text. Our code will be available at http://symgen.github.io.