Research map / Label-efficient and robust learning / Domain adaptation and generalization
Conversational Model Adaptation via KL Divergence Regularization
Juncen Li, Ping Luo, Fen Lin, Bo Chen
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| KL-regularized adaptation instead of ft-RED | ft-RED · Target-domain test set | BLEU: 7.71 -> 10.53 | Improved |
| KL-regularized adaptation instead of t-RED | t-RED · Target-domain test set | BLEU: 4.36 -> 10.53 | Improved |
| KL-regularized adaptation instead of c-RED | c-RED · Target-domain test set | BLEU: 1.35 -> 10.53 | Improved |
| combine source and target data (c-RED) instead of target only (t-RED) | t-RED · Target-domain test set | BLEU: 4.36 -> 1.35 | Hurt |
| combine source and target data (c-RED) instead of target only (t-RED) | t-RED · Manual evaluation | mean score: 0.318 -> 0.429 | Improved |
| pre-train on source then fine-tune (ft-RED) instead of target only | t-RED · Target-domain test set | BLEU: 4.36 -> 7.71 | Improved |
| KL-regularized adaptation instead of ft-RED | ft-RED · Manual evaluation | mean score: 0.484 -> 0.530 | Improved |
| KL-regularized adaptation instead of c-RED | c-RED · Manual evaluation | mean score: 0.429 -> 0.530 | Improved |
| KL-regularized adaptation instead of t-RED | t-RED · Manual evaluation | Bad(0) share: 71.3% -> 54.7% | Improved |
Closest papers by meaning
- Generative Adapter: Contextualizing Language Models in Parameters with A Single Forward Pass (ICLR 2025 · similarity 0.87)
- A Statistical Framework for Ranking LLM-based Chatbots (ICLR 2025 · similarity 0.86)
- Exploring Conversational Adaptability: Assessing the Proficiency of Large Language Models in Dynamic Alignment with Updated User Intent (AAAI 2025 · similarity 0.86)
- ULTRAFEEDBACK: Boosting Language Models with Scaled AI Feedback (ICML 2024 · similarity 0.85)
- WildChat-50M: A Deep Dive Into the Role of Synthetic Data in Post-Training (ICML 2025 · similarity 0.85)
- LMSYS-Chat-1M: A Large-Scale Real-World LLM Conversation Dataset (ICLR 2024 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
