Atomic Calibration of LLMs in Long-Form Generations
This addresses the challenge of trustworthiness in LLMs for real-world applications by improving confidence calibration in long-form outputs, though it is incremental as it builds on prior work on calibration.
The paper tackled the problem of hallucinations in large language models (LLMs) during long-form generation by introducing atomic calibration, which evaluates factuality at a fine-grained level by decomposing responses into atomic claims, and found that LLMs exhibit poorer calibration at this level.
Large language models (LLMs) often suffer from hallucinations, posing significant challenges for real-world applications. Confidence calibration, as an effective indicator of hallucination, is thus essential to enhance the trustworthiness of LLMs. Prior work mainly focuses on short-form tasks using a single response-level score (macro calibration), which is insufficient for long-form outputs that may contain both accurate and inaccurate claims. In this work, we systematically study atomic calibration, which evaluates factuality calibration at a fine-grained level by decomposing long responses into atomic claims. We further categorize existing confidence elicitation methods into discriminative and generative types, and propose two new confidence fusion strategies to improve calibration. Our experiments demonstrate that LLMs exhibit poorer calibration at the atomic level during long-form generation. More importantly, atomic calibration uncovers insightful patterns regarding the alignment of confidence methods and the changes of confidence throughout generation. This sheds light on future research directions for confidence estimation in long-form generation.