# Conversational Derailment Forecasting on CGA-CMV Large (randomly selected sample) Source: [Conversational Derailment Forecasting on CGA-CMV Large (randomly selected sample)]() Collection index: [llms.txt]() Data updated: 2026-05-29T08:23:00.411Z Evaluation of model performance on forecasting conversational derailment using the CGA-CMV Large dataset sample, assessed via Accuracy, F1, Precision, Recall, FPR, and H metrics. A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage. | Scope | Recorded value | | --- | --- | | Task | Conversational Derailment Forecasting | | Dataset (including any stated split) | CGA-CMV Large (randomly selected sample) | | Metric | Accuracy (Acc) | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 3 of 3 | | Coverage | All results currently returned for this benchmark | Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol. ## Metrics - [Accuracy (Acc)](): Higher is better; unit: PERCENT. - [False Positive Rate (FPR)](): Lower is better; unit: PERCENT. - [Precision (P)](): Higher is better; unit: PERCENT. - [Recall (R)](): Higher is better; unit: PERCENT. - [F1 Score (F1)](): Higher is better; unit: PERCENT. - [H Metric (H)](): Higher is better; unit: SCALAR. ## Recorded results | Method | Accuracy (Acc) | Result date | Report paper | | --- | --- | --- | --- | | Human | 70 | Not recorded | 6a191dabdbd7150506f828f2 | | SOTA | 70 | Not recorded | 6a191dabdbd7150506f828f2 | | Human | 62.2 | Not recorded | 6a191dabdbd7150506f828f2 | ### Human Accuracy (Acc): 70. Result ID: 6a194b6bfbc373592c1391b7. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Round":"2"}. This record does not assert independent reproduction. ### SOTA Accuracy (Acc): 70. Result ID: 6a194b6bfbc373592c1391b8. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"aggregation":"averaged across 5 random seeds"}. This record does not assert independent reproduction. ### Human Accuracy (Acc): 62.2. Result ID: 6a194b6bfbc373592c1391b6. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Round":"1"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.