Search papers, labs, and topics across Lattice.
Addressing the trade-off between interpretable claim verification and compute-heavy span-level error localization, Enoki uses an Open Information Extraction (OpenIE) framework to extract text-anchored relational tuples, verify them against evidence, and project unsupported facts directly back to input spans. This shared representation bypasses the expensive multi-call prompting and brittle alignment modules typical of hybrid hallucination detection systems. Across rule-, encoder-, and LLM-based backbones, the approach matches top claim-level verifiers at reduced inference cost while outperforming existing baselines on fine-grained span- and entity-level localization on the EnokiQA benchmark.
Text-anchored relational extraction eliminates the costly trade-off between claim verification and token-level error localization, matching state-of-the-art factuality detection at a fraction of the inference overhead.
Ensuring factuality remains a critical challenge for deploying LLMs in high-stakes settings. Existing hallucination detectors usually operate at a single level: claim-level methods provide interpretable factual units, while span-level methods localize unsupported text. Bridging these views is costly, as LLM-heavy pipelines require multiple decomposition and verification calls, and modular systems need additional claim-to-span alignment. We propose Enoki, an Open Information Extraction framework for multi-level hallucination detection. Enoki extracts text-anchored relational facts, verifies them against evidence, and projects unsupported facts back to hallucinated spans. This shared representation enables claim-level verification and span-level localization without requiring separate alignment. Enoki supports LLM-based, encoder-based, and rule-based extraction regimes, balancing accuracy and inference cost through a common interface. Experiments show that Enoki remains competitive with strong claim-level systems while using fewer resources and achieves superior performance on fine-grained span- and entity-level localization. We also release EnokiQA, a dual-granularity dataset with aligned claim-level verification and span-level localization annotations.
