Conformal Language Modeling
This work addresses the need for trustworthy and statistically rigorous outputs in generative language models, particularly for applications like medical report generation, though it is incremental in extending conformal prediction to LMs.
The authors tackled the problem of generating reliable prediction sets for language models by proposing a conformal prediction method that calibrates sampling and rejection rules to ensure at least one acceptable answer with high probability while maintaining small set sizes. They demonstrated the approach on tasks like question answering and summarization, achieving statistical guarantees for correctness.
We propose a novel approach to conformal prediction for generative language models (LMs). Standard conformal prediction produces prediction sets -- in place of single predictions -- that have rigorous, statistical performance guarantees. LM responses are typically sampled from the model's predicted distribution over the large, combinatorial output space of natural language. Translating this process to conformal prediction, we calibrate a stopping rule for sampling different outputs from the LM that get added to a growing set of candidates until we are confident that the output set is sufficient. Since some samples may be low-quality, we also simultaneously calibrate and apply a rejection rule for removing candidates from the output set to reduce noise. Similar to conformal prediction, we prove that the sampled set returned by our procedure contains at least one acceptable answer with high probability, while still being empirically precise (i.e., small) on average. Furthermore, within this set of candidate responses, we show that we can also accurately identify subsets of individual components -- such as phrases or sentences -- that are each independently correct (e.g., that are not "hallucinations"), again with statistical guarantees. We demonstrate the promise of our approach on multiple tasks in open-domain question answering, text summarization, and radiology report generation using different LM variants.