AI TrackerFrontier desk · AEST
Back to news
normalaitech

SAUCE: Summary Analysis Using Conversation Entailment

arXiv cs.CL · 5 Nov 2026, 03:25 AEST

SAUCE is a reference-free, fact-based evaluation pipeline designed for cross-lingual conversational speech summarization. It measures summary accuracy and fact coverage through entailment between conversation and text, providing an explainable alternative to traditional LLM evaluators.

Key facts

Read aloud

SAUCE is a reference-free, fact-based evaluation pipeline designed for cross-lingual conversational speech summarization. It measures summary accuracy and fact coverage through entailment between conversation and text, providing an explainable alternative to traditional L L M evaluators. With the growing need for evaluating Large Language Models L L M s and their applications to speech, challenges persist in summarizing and evaluating conversations that lack a clear end goal. We introduce SAUCE -- a reference-free, fact-based evaluation pipeline for cross-lingual conversational speech summarization. Crucially, unlike black-box L L M evaluators or dense embedding metrics, SAUCE is inherently explainable: it maps summary scores to discrete, verifiable facts, allowing users to pinpoint exact hallucinations or omissions. Preliminary investigations show SAUCE strongly align with human judgment.

Sources in this cluster