ConceptFlow Score

FlowScore

A normalized, interpretable metric for dialogue flow. It abstracts a dialogue into a sequence of intents and actions and performs sequence-level comparison to measure whether the conversation proceeds in the correct order, without relying on lexical or semantic embedding similarity.

Timeline (last 90 days)

Oct 16

FlowScore significantly outperformed all baseline methods across seven encoders in dialogue-level evaluation.

Unverified50%

Oct 16

FlowScore is a normalized interpretable score that facilitates direct comparison of dialogues of different lengths and domains.

Unverified50%

All Facts (2)

Unverified

FlowScore significantly outperformed all baseline methods across seven encoders in dialogue-level evaluation.

50%Unverified

FlowScore is a normalized interpretable score that facilitates direct comparison of dialogues of different lengths and domains.

50%

Source Articles

DialFork: A New Benchmark for Evaluating Procedural Hallucination in Conversational AgentsOct 16