Research map / Large language models / Language, safety and interpretability
Post-training Large Language Models for Diverse High-Quality Responses
Yilei Chen, Souradip Chakraborty, Lorenz Wolf, Ioannis Paschalidis, Aldo Pacchiano
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 |
|---|---|---|---|
| add DPP determinant diversity term (DQO) | PPO · Dolly | distinct-4: 0.64 -> 0.69 | Improved |
| add DPP determinant diversity term (DQO) | PPO · Dolly | pass@10: 8.39 -> 8.74 | Improved |
| add DPP determinant diversity term (DQO) | GRPO · GSM8K | distinct-4: 0.32 -> 0.42 | Improved |
| add DPP determinant diversity term (DQO) | GRPO · GSM8K | pass@1: 76.8 -> 76.3 | No clear effect |
| add DPP determinant diversity term (DQO) | GRPO · GSM8K | pass@10: 87.9 -> 91.2 | Improved |
| use DQO instead of GRPO-likelihood | GRPO-likelihood · GSM8K | pass@1: 50.9 -> 76.3 | Improved |
| use DQO instead of GRPO-entropy | GRPO-entropy · Dolly | pass@1: 4.71 -> 5.92 | Improved |
| use DQO instead of GRPO-entropy | GRPO-entropy · GSM8K | pass@10: 92.6 -> 91.2 | Hurt |
| increase diversity weight alpha from 1.0 to 2.0 (ablation) | DQO · Dolly | distinct-4 and pass@1: distinct-4 0.69 -> 0.82, pass@1 5.92 -> 5.42 | Mixed |
| increase group size k from 2 to 8 (ablation) | DQO · Dolly | distinct-4: 0.62 -> 0.79 | Improved |
| replace pairwise distance with determinant diversity (ablation) | DQO · city recommendation | diversity of recommended cities | Improved |
| use outcome reward instead of reward model with diversity objective | DQO · GSM8K | reward hacking | Hurt |
Closest papers by meaning
- Jointly Reinforcing Diversity and Quality in Language Model Generations (NeurIPS 2026 · similarity 0.92)
- Diversity-Aware Policy Optimization for Large Language Model Reasoning (NeurIPS 2025 · similarity 0.90)
- Diversity as a Reward: Fine-Tuning LLMs on a Mixture of Domain-Undetermined Data (NeurIPS 2025 · similarity 0.90)
- SetPO: Set-Level Policy Optimization for Diversity-Preserving LLM Reasoning (ICML 2026 · similarity 0.89)
- Understanding the Effects of RLHF on LLM Generalisation and Diversity (ICLR 2024 · similarity 0.88)
- Towards Scalable Data Diversification for Language Model Pretraining via Leverage Score Sampling (NeurIPS 2026 · similarity 0.88)
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.
