No Single Best Model for Diversity: Learning a Router for Sample Diversity
Not on the map. Why?
Abstract (arXiv)
When posed with prompts that permit a large number of valid answers, comprehensively generating them is the first step towards satisfying a wide range of users. In this paper, we study methods to elicit a comprehensive set of valid responses. To evaluate this, we introduce diversity coverage, a metric that measures the total quality scores assigned to each unique answer in the predicted answer set relative to the best possible answer set with the same number of answers. Using this metric, we evaluate 18 LLMs, finding no single model dominates at generating diverse responses to a wide range of open-ended prompts. Yet, per each prompt, there exists a model that outperforms all other models significantly at generating a diverse answer set. Motivated by this finding, we introduce a router that predicts the best model for each query. On NB-WildChat, our trained router outperforms the single best model baseline (26.3% vs 23.8%). We further show generalization to an out-of-domain dataset (NB-Curated) as well as different answer-generation prompting strategies. Our work lays foundation for studying generating comprehensive answers when we have access to a suite of models.
Key facts. Cited by 2 papers in the map (data through 14 Oct 2026). Builds on 40 of its 53 references that are in the map. First posted on arXiv 2 Apr 2026.
Builds on
40 of 53 references.
References · 40
- 2018BERT: Pre-training of Deep Bidirectional Transformers for Language Understandingcited by 40,399
- 2021Training Verifiers to Solve Math Word Problemscited by 10,109
- 2020Measuring Massive Multitask Language Understandingcited by 7,963
- 2019The Curious Case of Neural Text Degenerationcited by 2,916
- 2025AlphaEvolve: A coding agent for scientific and algorithmic discoverycited by 867
- 2023Whose Opinions Do Language Models Reflect?cited by 773
- 2024RouteLLM: Learning to Route LLMs with Preference Datacited by 682
- 2023LLM-Blender: Ensembling Large Language Models with Pairwise Ranking and Generative Fusioncited by 564
- 2022Unnatural Instructions: Tuning Language Models with (Almost) No Human Laborcited by 397
- 2024Hybrid LLM: Cost-Efficient and Quality-Aware Query Routingcited by 372
- 2024Skywork-Reward: Bag of Tricks for Reward Modeling in LLMscited by 327
- 2023Does Writing with Language Models Reduce Content Diversity?cited by 192
28 more
- 2024Turning Up the Heat: Min-p Sampling for Creative and Coherent LLM Outputscited by 150
- 2023Routing to the Expert: Efficient Reward-guided Ensemble of Large Language Modelscited by 145
- 2025Artificial Hivemind: The Open-Ended Homogeneity of Language Models (and Beyond)cited by 141
- 2022Diverse Demonstrations Improve In-context Compositional Generalizationcited by 139
- 2024From Decoding to Meta-Generation: Inference-time Algorithms for Large Language Modelscited by 137
- 2020Evaluating the Evaluation of Diversity in Natural Language Generationcited by 129
- 2024Homogenization Effects of Large Language Models on Human Creative Ideationcited by 127
- 2024Modular Pluralism: Pluralistic Alignment via Multi-LLM Collaborationcited by 113
- 2022Locally Typical Samplingcited by 111
- 2025Verbalized Sampling: How to Mitigate Mode Collapse and Unlock LLM Diversitycited by 105
- 2019Do Massively Pretrained Language Models Make Better Storytellers?cited by 90
- 2024RouterDC: Query-Based Router by Dual Contrastive Learning for Assembling Large Language Modelscited by 88
- 2025Router-R1: Teaching LLMs Multi-Round Routing and Aggregation via Reinforcement Learningcited by 72
- 2025NoveltyBench: Evaluating Language Models for Humanlike Diversitycited by 66
- 2024Benchmarking Distributional Alignment of Large Language Modelscited by 61
- 2025Diverse Preference Optimizationcited by 52
- 2025Base Models Beat Aligned Models at Randomness and Creativitycited by 51
- 2024Forcing Diffuse Distributions out of Language Modelscited by 48
- 2024Benchmarking Language Model Creativity: A Case Study on Code Generationcited by 40
- 2025Modifying Large Language Model Post-Training for Diverse Creative Writingcited by 40
- 2025Multilingual Prompting for Improving LLM Generation Diversitycited by 25
- 2025Heterogeneous Swarms: Jointly Optimizing Model Roles and Weights for Multi-LLM Systemscited by 23
- 2023How Far Can We Extract Diverse Perspectives from Large Language Models?cited by 22
- 2025Spectrum Tuning: Post-Training for Distributional Coverage and In-Context Steerabilitycited by 20
- 2025Mode-Conditioning Unlocks Superior Test-Time Scalingcited by 8
- 2025The Role of Diversity in In-Context Learning for Large Language Modelscited by 6
- 2026Reaching Beyond the Mode: RL for Distributional Reasoning in Language Modelscited by 6
- 2026The Single-Multi Evolution Loop for Self-Improving Model Collaboration Systemscited by 2
Built on by
Citing papers per year: 2026: 2.
2 papers.