sui-1: Grounded and Verifiable Long-Form Summarization
This addresses a critical limitation in compliance-sensitive domains like government and legal analysis by enabling verifiable summaries, though it is incremental as it builds on existing summarization methods with a focus on citations.
The paper tackles the problem of large language models generating unfaithful summaries that are hard to verify, presenting sui-1, a 24B parameter model that produces abstractive summaries with inline citations, which significantly outperforms open-weight baselines, including models with 3x more parameters.
Large language models frequently generate plausible but unfaithful summaries that users cannot verify against source text, a critical limitation in compliance-sensitive domains such as government and legal analysis. We present sui-1, a 24B parameter model that produces abstractive summaries with inline citations, enabling users to trace each claim to its source sentence. Our synthetic data pipeline combines chain-of-thought prompting with multi-stage verification, generating over 22,000 high-quality training examples across five languages from diverse sources including parliamentary documents, web text, and Wikipedia. Evaluation shows sui-1 significantly outperforms all tested open-weight baselines, including models with 3x more parameters. These results demonstrate that task-specific training substantially outperforms scale alone for citation-grounded summarization. Model weights and an interactive demo are publicly available.