SKG-Eval: Stateful Evaluation of Multi-Turn Dialogue via Incremental Semantic Knowledge Graphs
Avijit Shil Suman Samui, [1ex] Maulana Abul Kalam Azad University of Technology, West Bengal, India · 2026
Read SKG-Eval: Stateful Evaluation of Multi-Turn Dialogue via Incremental Semantic Knowledge Graphs by Avijit Shil Suman Samui, [1ex] Maulana Abul Kalam Azad University of Technology, West Bengal, India (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-16650.txt.
Description
Shil and Samui propose SKG-Eval, a framework for judging multi-turn dialogue by tracking a growing knowledge graph of entities, relations, and commitments. It scores three things: fit to the current prompt, fit to past context, and logical coherence, the last via a geometric contradiction engine that needs no language-model judge. The three signals fuse into a session score that stays fair across lengths, with explicit contradiction certificates. Across benchmarks it matches human judgments better and catches long-range inconsistencies more reliably than flat evaluators.
Catalog record
| CREATOR | Avijit Shil Suman Samui, [1ex] Maulana Abul Kalam Azad University of Technology, West Bengal, India |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 105917 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.16650