DIAL: Direct Iterative Adversarial Learning for Realistic Multi-Turn Dialogue Simulation
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866913092685791232 |
|---|---|
| author | Zhu, Ziyi Tieleman, Olivier Stamatis, Caitlin A. Smyth, Luka Hull, Thomas D. Cahn, Daniel R. Chen, Jinghong Malgaroli, Matteo |
| author_facet | Zhu, Ziyi Tieleman, Olivier Stamatis, Caitlin A. Smyth, Luka Hull, Thomas D. Cahn, Daniel R. Chen, Jinghong Malgaroli, Matteo |
| contents | Realistic user simulation is crucial for training and evaluating multi-turn dialogue systems, yet creating simulators that accurately replicate human behavior remains a significant challenge. An effective simulator must expose the failure modes of the systems under evaluation. This work introduces Direct Iterative Adversarial Learning (DIAL), an adversarial framework that iteratively enhances user simulator realism through a competitive dynamic between a generator (user simulator) and a discriminator. When applied to mental health support, a domain characterized by diverse failure types and a critical dependence on realistic user behavior for failure detection, DIAL restores lexical diversity diminished by supervised fine-tuning and drastically reduces discriminator accuracy. The resulting simulator exhibits a strong correlation between simulated and real failure occurrence rates while maintaining low distributional divergence of failure modes. These findings indicate that DIAL is a promising method for developing realistic user simulators in multi-turn dialogue, facilitating reliable and cost-effective system evaluation prior to deployment. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2512_20773 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | DIAL: Direct Iterative Adversarial Learning for Realistic Multi-Turn Dialogue Simulation Zhu, Ziyi Tieleman, Olivier Stamatis, Caitlin A. Smyth, Luka Hull, Thomas D. Cahn, Daniel R. Chen, Jinghong Malgaroli, Matteo Computation and Language Realistic user simulation is crucial for training and evaluating multi-turn dialogue systems, yet creating simulators that accurately replicate human behavior remains a significant challenge. An effective simulator must expose the failure modes of the systems under evaluation. This work introduces Direct Iterative Adversarial Learning (DIAL), an adversarial framework that iteratively enhances user simulator realism through a competitive dynamic between a generator (user simulator) and a discriminator. When applied to mental health support, a domain characterized by diverse failure types and a critical dependence on realistic user behavior for failure detection, DIAL restores lexical diversity diminished by supervised fine-tuning and drastically reduces discriminator accuracy. The resulting simulator exhibits a strong correlation between simulated and real failure occurrence rates while maintaining low distributional divergence of failure modes. These findings indicate that DIAL is a promising method for developing realistic user simulators in multi-turn dialogue, facilitating reliable and cost-effective system evaluation prior to deployment. |
| title | DIAL: Direct Iterative Adversarial Learning for Realistic Multi-Turn Dialogue Simulation |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2512.20773 |