Dialogue flow prediction on OnCoCo 60 categories (5-fold CV)
10.2Macro-F1GB-large
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| GB-largelambda_tm=0.5, Backbone=History, Context length=42026.04 | 10.2 | 17.1 | 34.2 | 89.4 | 17.5 | |
| GB-baselambda_tm=0.5, Backbone=BERT, Context length=82026.04 | 9.8 | 15.8 | 35.4 | 93.3 | 14.3 | |
| MGB-1Blambda_tm=0.2, Backbone=History, Context length=82026.04 | 9.7 | 15.9 | 31.4 | 82.1 | 23.5 | |
| EB-610Mlambda_tm=0.5, Backbone=History, Context length=42026.04 | 9.7 | 16 | 33.4 | 93.9 | 10.9 | |
| EB-210Mlambda_tm=0.5, Backbone=History, Context length=122026.04 | 9.5 | 15.8 | 33.1 | 93.7 | 12.5 | |
| GPT-5-miniEvaluation protocol=zero-shot2026.04 | 9.1 | 13.2 | 25.4 | 55 | — | |
| MGB-134Mlambda_tm=0.5, Backbone=History, Context length=42026.04 | 8.6 | 14.7 | 35.8 | 94.2 | 12.2 | |
| gpt-oss-120bEvaluation protocol=zero-shot2026.04 | 7.2 | 10.8 | 17.4 | 40 | — | |
| GELECTRalambda_tm=1.0, Backbone=History, Context length=42026.04 | 7 | 12.8 | 33.5 | 93.3 | 12.1 | |
| TM OnlyDescription=Transition Matrix Baseline2026.04 | 5.6 | 11.5 | 31.5 | 100 | — | |
| Tanaka (2019)lambda_tm=1.52026.04 | 0.8 | 3.1 | 23.5 | 80.2 | 20.9 | |
| Simple RNNlambda_tm=0.52026.04 | 0.3 | 1.4 | 20.8 | 75.1 | 19.8 |