Beyond Task-Oriented and Chitchat Dialogues: Proactive and Transition-Aware Conversational Agents
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866912703780487168 |
|---|---|
| author | Yoon, Yejin Son, Yuri So, Namyoung Kim, Minseo Cho, Minsoo Park, Chanhee Lee, Seungshin Kim, Taeuk |
| author_facet | Yoon, Yejin Son, Yuri So, Namyoung Kim, Minseo Cho, Minsoo Park, Chanhee Lee, Seungshin Kim, Taeuk |
| contents | Conversational agents have traditionally been developed for either task-oriented dialogue (TOD) or open-ended chitchat, with limited progress in unifying the two. Yet, real-world conversations naturally involve fluid transitions between these modes. To address this gap, we introduce TACT (TOD-And-Chitchat Transition), a dataset designed for transition-aware dialogue modeling that incorporates structurally diverse and integrated mode flows. TACT supports both user- and agent-driven mode switches, enabling robust modeling of complex conversational dynamics. To evaluate an agent's ability to initiate and recover from mode transitions, we propose two new metrics -- Switch and Recovery. Models trained on TACT outperform baselines in both intent detection and mode transition handling. Moreover, applying Direct Preference Optimization (DPO) to TACT-trained models yields additional gains, achieving 75.74\% joint mode-intent accuracy and a 70.1\% win rate against GPT-4o in human evaluation. These results demonstrate that pairing structurally diverse data with DPO enhances response quality and transition control, paving the way for more proactive and transition-aware conversational agents. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2511_08835 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Beyond Task-Oriented and Chitchat Dialogues: Proactive and Transition-Aware Conversational Agents Yoon, Yejin Son, Yuri So, Namyoung Kim, Minseo Cho, Minsoo Park, Chanhee Lee, Seungshin Kim, Taeuk Computation and Language Artificial Intelligence Conversational agents have traditionally been developed for either task-oriented dialogue (TOD) or open-ended chitchat, with limited progress in unifying the two. Yet, real-world conversations naturally involve fluid transitions between these modes. To address this gap, we introduce TACT (TOD-And-Chitchat Transition), a dataset designed for transition-aware dialogue modeling that incorporates structurally diverse and integrated mode flows. TACT supports both user- and agent-driven mode switches, enabling robust modeling of complex conversational dynamics. To evaluate an agent's ability to initiate and recover from mode transitions, we propose two new metrics -- Switch and Recovery. Models trained on TACT outperform baselines in both intent detection and mode transition handling. Moreover, applying Direct Preference Optimization (DPO) to TACT-trained models yields additional gains, achieving 75.74\% joint mode-intent accuracy and a 70.1\% win rate against GPT-4o in human evaluation. These results demonstrate that pairing structurally diverse data with DPO enhances response quality and transition control, paving the way for more proactive and transition-aware conversational agents. |
| title | Beyond Task-Oriented and Chitchat Dialogues: Proactive and Transition-Aware Conversational Agents |
| topic | Computation and Language Artificial Intelligence |
| url | https://arxiv.org/abs/2511.08835 |