SKG-Eval models dialogues as incremental semantic knowledge graphs and fuses local relevance, historical consistency, and geometric contradiction signals into a length-invariant session score that correlates better with humans on long-range issues.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
SKG-Eval: Stateful Evaluation of Multi-Turn Dialogue via Incremental Semantic Knowledge Graphs
SKG-Eval models dialogues as incremental semantic knowledge graphs and fuses local relevance, historical consistency, and geometric contradiction signals into a length-invariant session score that correlates better with humans on long-range issues.