KO | EN

GitLite

Minimalist Open Source Repository Search Engine

gitlite — search

#developer-tools #codex #typescript #ai-agents #claude-code #dsh #dsh-plugin #open-source #cli #deepseek-harness #macos #ai-agent

Awesome-LLM-Ensemble

★ 253 Open GitHub ↗

A curated list of Awesome-LLM-Ensemble papers for the survey "Harnessing Multiple Large Language Models: A Survey on LLM Ensemble"

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

<h1 align="center">🔥🔥🔥 Awesome-LLM-Ensemble

"Harnessing Multiple Large Language Models: A Survey on LLM Ensemble" (IJCAI Survey 2026) </h2>

<a href="https://zhijunchen-ai.github.io/">Zhijun Chen</a>,
 <a href="https://scholar.google.com/citations?user=JEdD0o8AAAAJ&hl">Xiaodong Lu</a>,
 <a href="https://scholar.google.com/citations?user=6v6JfdsAAAAJ&hl=zh-en">Jingzheng Li</a>,
 <a href="https://scholar.google.com/citations?user=mzXg1s8AAAAJ&hl=zh-en">Pengpeng Chen</a>,
 <a href="https://lizhuoran-nlp.github.io/">Zhuoran Li</a>,
 <a href="https://scholar.google.com/citations?user=buUlnJUAAAAJ&hl=zh-en">Kai Sun</a>,
 <a href="https://luoyk1999.github.io/">Yuankai Luo</a>,
 <a href="https://scholar.google.com/citations?hl=zh-CN&user=PnDqlPkAAAAJ&view_op=list_works&sortby=pubdate">Qianren Mao</a>,
 <a href="https://scholar.google.com/citations?user=H3pKglUAAAAJ&hl=zh-en">Ming Li</a>, 
 <a href="https://www.researchgate.net/scientific-contributions/Likang-Xiao-2257895822">Likang Xiao</a>,
 <a href="https://scholar.google.co.uk/citations?user=fnfg9S0AAAAJ&hl=en">Dingqi Yang</a>,
 <a href="https://scholar.google.com/citations?user=Be21PkYAAAAJ">Xiao Huang</a>,
 <a href="https://www.banyikun.com/">Yikun Ban</a>,
 <a href="https://scholar.google.com/citations?user=HWOWCdcAAAAJ&hl=zh-CN">Hailong Sun</a>,
 <a href="https://scholar.google.com/citations?user=D0lL1r0AAAAJ&hl=zh-CN/">Philip S. Yu</a>


 <img src="fig/logo.png" alt="Logo" width="=80%">

<h5 align="center">If you like our project, please give it a star ⭐ to show your support！Thank you:)

</h5>

# 📣 Notices
> 🔥🔥🔥 This is a collection of papers on ***LLM Ensemble***. 
> <a href='https://arxiv.org/abs/2502.18036'>
> <img src='https://img.shields.io/badge/Arxiv-2502.18036-b31b1b.svg?logo=arXiv'>
> </a>
> <a href="https://junchenzhi.github.io/LLM-Ensemble/">
> <img src="https://img.shields.io/badge/Website-Visit%20Now-purple" alt="Website">
> </a>
> <a href="https://mp.weixin.qq.com/s/yVWzHQmr_KyyOfY5k3ivOw">
> <img src="https://img.shields.io/badge/Blog-(Chinese)-orange" alt="Blog (Chinese)">
> </a>

> **[2026-05] Accepted by IJCAI Survey 2026!**    
>

> 🔥🔥🔥 **[2026-04] We have updated our arXiv paper with a new version! Stay tuned for our journal-style paper in recent months.**    
>

> **[Always] [Add your papers in this repo]** **Thank you to all the papers that have cited our survey.**      
>We will add all related citing papers to this GitHub repo, in a timely manner, to help increase the visibility of your contributions.

> **[Always] [Maintain]** **We will make this list updated frequently!**     
> If you found any missed/new paper, please don't hesitate to contact [us](zhijunchen@buaa.edu.cn) or Pull requests.

&nbsp;

## 🍀 Citation
If you find this survey useful, please consider citing our paper:

```
@article{chen2025harnessing,
  title={Harnessing Multiple Large Language Models: A Survey on LLM Ensemble},
  author={Chen, Zhijun and Lu, Xiaodong and Li, Jingzheng and Chen, Pengpeng and Li, Zhuoran and Sun, Kai and Luo, Yuankai and Mao, Qianren and Li, Ming and Xiao, Likang and Yang, Dingqi and Huang, Xiao and Ban, Yikun and Sun, Hailong and Yu, Philip S},
  journal={arXiv preprint arXiv:2502.18036},
  year={2025}
}
```

&nbsp; 
&nbsp;  
&nbsp;

- [Contents](#Awesome-LLM-Ensemble)
  - [1. LLM Ensemble and Taxonomy](#1-llm-ensemble-and-taxonomy)
    - [1.1 LLM Ensemble](#11-llm-ensemble)
    - [1.2 Taxonomy](#12-taxonomy)
  - [2. Papers](#2-papers)
    - [2.1 Ensemble Before Inference](#21-ensemble-before-inference)
      - [2.1.1 (a,1) Pre-Trained Router](#211-a1-pre-trained-router)
      - [2.1.2 (a,2) Non pre-trained router](#212-a2-non-pre-trained-router)
    - [2.2 Ensemble During Inference](#22-ensemble-during-inference)
      - [2.2.1 (b,1) Token-Level Ensemble](#221-b1-token-level-ensemble)
      - [2.2.2 (b,2) Span-Level Ensemble](#222-b2-span-level-ensemble)
      - [2.2.3 (b,3) Process-Level Ensemble](#223-b3-process-level-ensemble)
    - [2.3 Ensemble After Inference](#23-ensemble-after-inference)
      - [2.3.1 (c,1) Non Cascade](#231-c1-non-cascade)
      - [2.3.2 (c,2) Cascade](#232-c2-cascade)
    - [2.4 Others: Benchmarks, Applications, Systems and Related Surveys](#24-others-benchmarks-applications-systems-and-related-surveys)
      - [2.4.1 Benchmarks](#241-benchmarks)
      - [2.4.2 Applications](#242-applications)
      - [2.4.3 Systems](#243-systems)
      - [2.4.4 Related Surveys](#244-related-surveys)
  - [3 Others: Some public implementations of the LLM Ensemble methods](#3-others-some-public-implementations-of-the-llm-ensemble-methods)
  - [4 Others: Some other related interesting papers](#4-others-some-other-related-interesting-papers)
      - [4.1 Test-Time Scaling](#41-test-time-scaling)
      - [4.2 LLM Collaboration and Others](#42-llm-collaboration-and-others)
  - [5 Summarization](#5-summarization)

# 1. LLM Ensemble and Taxonomy

## 1.1 LLM Ensemble
**Paper Abstract**:

LLM Ensemble---which involves the comprehensive use of multiple large language models (LLMs), each aimed at handling user queries during the downstream inference, to benefit from their individual strengths---has gained substantial attention recently. 
The widespread availability of LLMs, coupled with their varying strengths and out-of-the-box usability, has profoundly advanced the field of LLM Ensemble. 
This paper presents the first systematic review of recent developments in LLM Ensemble. 
First, we introduce our taxonomy of LLM Ensemble and discuss several related research problems. 
Then, we provide a more in-depth classification of methods under the broad categories of ``ensemble-before-inference, ensemble-during-inference, ensemble-after-inference'', and review all relevant methods. Finally, we introduce related benchmarks and applications, summarize existing studies, and suggest several future research directions. 
A curated list of papers  on LLM Ensemble is available at https://github.com/junchenzhi/Awesome-LLM-Ensemble.

## 1.2 Taxonomy

**Figure 1: Illustration of LLM Ensemble Taxonomy.** (Note that for *(b) ensemble-during-inference* paradigm, there is also a *process-level ensemble* approach that we have not represented in the figure, mainly because that this approach is instantiated by a single method.)
</div>

**Figure 2: Taxonomy of All LLM Ensemble Methods. (Please note that this figure may not be fully updated to include all the papers listed below.)**
</div>

- ***(a) Ensemble before inference.***  
Since the ensemble-before-inference methods require routing a query to the most suitable LLM before LLM inference, the core of such methods lies in predicting the utility of candidate models for a given query under certain preferences (e.g., performance or cost). Based on how they formulate the utility of candidate LLMs, we divide existing methods into two categories:
  - ***(a1) Discrete utility*** methods, discretize the model utility into categorical labels;
  - ***(a2) Continuous utility*** methods model LLM utility as real-valued variables, such as response length or performance scores. This formulation enables a fine-grained characterization of model behavior, capturing subtle performance differences obscured by categorical definitions.

- ***(b) Ensemble during inference.***  
As the most granular form of ensemble among the three broad categories, this type of approach encompasses: 
  - ***(b1) Token-level ensemble*** methods, which integrate the token-level outputs of multiple models at the finest granularity of decoding;
  - ***(b2) Span-level ensemble*** methods, which conduct ensemble at the level of a sequence fragment (e.g., a span of four words); 
  - ***(b3) Process-level ensemble*** methods, which select the optimal reasoning process step-by-step within the reasoning chain for a given complex reasoning task. 
Note that for these ensemble-during-inference methods, the aggregated text segments will be concatenated with the previous text and fed again to models.

- ***(c) Ensemble after inference.***  
These methods can be classified into two categories:
  - ***(c1) Non cascade*** methods, which perform ensemble using multiple complete responses contributed from all LLM candidates;
  - ***(c2) Cascade*** methods, which consider both performance and inference costs, progressively reasoning through a chain of LLM candidates largely sorted by model size to find the most suitable inference response.

&nbsp; 
&nbsp;  
&nbsp;

# 2. Papers

## 2.1 Ensemble Before Inference

Figure 3: Summary analysis of the key attributes of ensemble-before-inference methods. (Please note that this table may not be fully updated to include all the papers listed below.)
</div>

### 2.1.1 (a,1) Discrete utility methods

| Date | Name | Title | Paper/Github |
|:---:|:---:|:---|:---:|
| 2026-06 | <code>ACRouter</code> | Agent-as-a-Router: Agentic Model Routing for Coding Tasks | <a href="https://arxiv.org/abs/2606.22902"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/LanceZPF/agent-as-a-router?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/LanceZPF/agent-as-a-router) |
| 2026-04 | <code>DialRouter</code> | From Myopic Selection to Long-Horizon Awareness: Sequential LLM Routing for Multi-Turn Dialogue | <a href="https://arxiv.org/abs/2604.12385"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2025-10 | <code>DiSRouter</code> | DISROUTER: Distributed Self-Routing for LLM Selections | <a href="https://arxiv.org/abs/2510.19208"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2025-06 | <code>TagRouter</code> | TAGROUTER: Learning Route to LLMs through Tags for Open-Domain Text Generation Tasks | <a href="https://arxiv.org/abs/2506.12473"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2025-06 | <code>Router-R1</code> | Router-R1: Teaching LLMs Multi-Round Routing and Aggregation via Reinforcement Learning | <a href="https://arxiv.org/abs/2506.09033"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/ulab-uiuc/Router-R1?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/ulab-uiuc/Router-R1) |
| 2025-06 | <code>RadialRouter</code> | RadialRouter: Structured Representation for Efficient and Robust Large Language Models Routing | <a href="https://www.arxiv.org/abs/2506.03880"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2025-05 | <code>RTR</code> | Route to Reason: Adaptive Routing for LLM and Reasoning Strategy Selection | <a href="https://arxiv.org/abs/2505.19435"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/goodmanpzh/Route-To-Reason?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/goodmanpzh/Route-To-Reason) |
| 2024-12 | <code>Bench-CoE</code> | Bench-CoE: a Framework for Collaboration of Experts from Benchmark | <a href="https://arxiv.org/abs/2412.04167"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/ZhangXJ199/Bench-CoE?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/ZhangXJ199/Bench-CoE) |
| 2024-10 | <code>GraphRouter</code> | GraphRouter: A Graph-based Router for LLM Selections | <a href="https://arxiv.org/abs/2410.03834"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/ulab-uiuc/GraphRouter?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/ulab-uiuc/GraphRouter) |
| 2024-09 | <code>Eagle</code> | Eagle: Efficient Training-Free Router for Multi-LLM Inference | <a href="https://arxiv.org/abs/2409.15518"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2024-08 | <code>SelectLLM</code> | SelectLLM: Query-Aware Efficient Selection Algorithm for Large Language Models | <a href="https://arxiv.org/abs/2408.08545"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2024-06 | <code>RouteLLM</code> | RouteLLM: Learning to Route LLMs with Preference Data | <a href="https://arxiv.org/abs/2406.18665"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/lm-sys/RouteLLM?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/lm-sys/RouteLLM) |
| 2024-05 | <code>LLM Routing Lessons</code> | Harnessing the Power of Multiple Minds: Lessons Learned from LLM Routing | <a href="https://arxiv.org/abs/2405.00467"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/kvadityasrivatsa/llm-routing?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/kvadityasrivatsa/llm-routing) |
| 2024-04 | <code>Hybrid-LLM</code> | Hybrid LLM: Cost-Efficient and Quality-Aware Query Routing | <a href="https://arxiv.org/abs/2404.14618"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2024-03 | <code>ETR</code> | An Expert is Worth One Token: Synergizing Multiple Expert LLMs as Generalist via Expert Token Routing | <a href="https://arxiv.org/abs/2403.16854"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/zjunet/ETR?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/zjunet/ETR) |
| 2024-01 | <code>Routoo</code> | Routoo: Learning to Route to Large Language Models Effectively | <a href="https://arxiv.org/abs/2401.13979"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2024 | <code>RouterDC</code> | RouterDC: Query-Based Router by Dual Contrastive Learning for Assembling Large Language Models | <a href="https://proceedings.neurips.cc/paper_files/paper/2024/hash/7a641b8ec86162fc875fb9f6456a542f-Abstract-Conference.html"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/shuhao02/RouterDC?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/shuhao02/RouterDC) |
| 2023-11 | <code>ZOOTER</code> | Routing to the Expert: Efficient Reward-guided Ensemble of Large Language Models | <a href="https://arxiv.org/abs/2311.08692"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2023-08 | <code>FORC</code> | Fly-Swat or Cannon? Cost-Effective Language Model Choice via Meta-Modeling | <a href="https://arxiv.org/abs/2308.06077"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/epfl-dlab/forc?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/epfl-dlab/forc) |
| 2023 | <code>Benchmark Routing</code> | LLM Routing with Benchmark Datasets | <a href="https://openreview.net/forum?id=k9EfAJhFZc"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=openreview&logoColor=white" height="18"></a> - |

### 2.1.2 (a,2) Continuous utility methods

| Date | Name | Title | Paper/Github |
|:---:|:---:|:---|:---:|
| 2026-06 | <code>FLARE</code> | FLARE: Fine-Grained Length-Aware Routing for Resource-Efficient Heterogeneous LLM Serving | <a href="https://aclanthology.org/2026.acl-long.1018/"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=acl&logoColor=white" height="18"></a> - |
| 2026-06 | <code>-</code> | Conformal LLM Routing with Distribution-Free Safety Guarantees | <a href="https://aclanthology.org/2026.acl-srw.70/"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=acl&logoColor=white" height="18"></a> - |
| 2026-06 | <code>SCOPE</code> | SCOPE: Cost-Efficient Model Selection for Compound AI Systems under Quality Constraints | <a href="https://arxiv.org/abs/2606.00774"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2026-03 | <code>MonoRouter</code> | MonoRouter: Enforcing Monotonicity in Item Response Theory for LLM Routing | <a href="https://www.semanticscholar.org/paper/4d269df75f97ae82ba63bd248e1d7d9adf30e0a0"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=semantic-scholar&logoColor=white" height="18"></a> - |
| 2026-03 | <code>ReLope</code> | ReLope: KL-Regularized LoRA Probes for Multimodal LLM Routing | <a href="https://arxiv.org/abs/2603.24787"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/Spinozaaa/ReLope?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/Spinozaaa/ReLope) |
| 2026-03 | <code>-</code> | Multi-LLM Query Optimization | <a href="https://arxiv.org/abs/2603.24617"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2026-02 | <code>RACER</code> | RACER: Risk-Aware Calibrated Efficient Routing for Large Language Models | <a href="https://arxiv.org/abs/2603.06616"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2026-02 | <code>-</code> | Don't Always Pick the Highest-Performing Model: An Information Theoretic View of LLM Ensemble Selection | <a href="https://arxiv.org/abs/2602.08003"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2025-11 | <code>EPIC</code> | EPIC: Ensemble Planning with Contrastive learning | <a href="https://arxiv.org/abs/2511.00521"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/nguyenngocbaocmt02/EPIC?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/nguyenngocbaocmt02/EPIC) |
| 2025-10 | <code>WebRouter</code> | WebRouter: Query-specific Router via Variational Information Bottleneck for Cost-sensitive Web Agent | <a href="https://arxiv.org/abs/2510.11221"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2025-10 | <code>LLMRank</code> | LLMRank: Understanding LLM Strengths for Model Routing | <a href="https://arxiv.org/abs/2510.01234"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2025-05 | <code>Avengers</code> | The Avengers: A Simple Recipe for Uniting Smaller Language Models to Challenge Proprietary Giants | <a href="https://arxiv.org/abs/2505.19797"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/ZhangYiqun018/Avengers?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/ZhangYiqun018/Avengers) |
| 2025-05 | <code>InferenceDynamics</code> | InferenceDynamics: Efficient Routing Across LLMs through Structured Capability and Knowledge Profiling | <a href="https://arxiv.org/abs/2505.16303"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2025-05 | <code>kNN Router</code> | Rethinking Predictive Modeling for LLM Routing: When Simple kNN Beats Complex Learned Routers | <a href="https://arxiv.org/abs/2505.12601"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2025 | <code>RELM</code> | Co-optimizing Recommendation and Evaluation for LLM Selection | <a href="https://openreview.net/pdf?id=gWi4ZcPQRl"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=openreview&logoColor=white" height="18"></a> - |
| 2025-02 | <code>LLM Bandit</code> | LLM Bandit: Cost-Efficient LLM Generation via Preference-Conditioned Dynamic Routing | <a href="https://arxiv.org/abs/2502.02743"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2024-12 | <code>PickLLM</code> | PickLLM: Context-Aware RL-Assisted Large Language Model Routing | <a href="https://arxiv.org/abs/2412.12170"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2024-08 | <code>TO-Router</code> | TensorOpera Router: A Multi-Model Router for Efficient LLM Inference | <a href="https://arxiv.org/abs/2408.12320"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2024-07 | <code>MetaLLM</code> | MetaLLM: A High-performant and Cost-efficient Dynamic Framework for Wrapping LLMs | <a href="https://arxiv.org/abs/2407.10834"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/mail-research/MetaLLM-wrapper?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/mail-research/MetaLLM-wrapper/) |
| 2024-06 | <code>HomoRouter</code> | Query Routing for Homogeneous Tools: An Instantiation in the RAG Scenario | <a href="https://arxiv.org/abs/2406.12429"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2024-01 | <code>Blending</code> | Blending Is All You Need: Cheaper, Better Alternative to Trillion-Parameters LLM | <a href="https://arxiv.org/abs/2401.02994"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |

## 2.2 Ensemble During Inference

Figure 4: Summary analysis of the key attributes of ensemble-during-inference methods. (Please note that this table may not be fully updated to include all the papers listed below.)
</div>

### 2.2.1 (b,1) Token-Level Ensemble

| Date | Name | Title | Paper/Github |
|:---:|:---:|---|:---:|
| 2026-06 | `DLLG` | DLLG: Dynamic Logit-Level Gating of LLM Experts | <a href="https://arxiv.org/abs/2606.04378"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2026-05 | `WASH` | Linear Ensembles Wash Away Watermarks: On the Fragility of Distributional Perturbations in LLMs | <a href="https://arxiv.org/abs/2605.30501"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2026-05 | `ME` | Rethinking LLM Ensembling from the Perspective of Mixture Models | <a href="https://arxiv.org/abs/2605.00419"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/Kamichanw/Mixture-model-like-Ensemble?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/Kamichanw/Mixture-model-like-Ensemble) |
| 2026-04 | `SpecFed` | SpecFed: Accelerating Federated LLM Inference with Speculative Decoding and Compressed Transmission | <a href="https://arxiv.org/abs/2604.25777"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2026-01 | `FusionRoute` | Token-Level LLM Collaboration via FusionRoute | <a href="https://arxiv.org/abs/2601.05106"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2025-10 | `SAFE` | When to Ensemble: Identifying Token-Level Points for Stable and Fast LLM Ensembling | <a href="https://arxiv.org/pdf/2510.15346"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2025-10 | `CoRe` | Harnessing Consistency for Robust Test-Time LLM Ensemble | <a href="https://www.arxiv.org/abs/2510.13855"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2025-05 | `Transformer Copilot` | Transformer Copilot: Learning from The Mistake Log in LLM Fine-tuning | <a href="https://arxiv.org/abs/2505.16270"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/jiaruzouu/TransformerCopilot?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/jiaruzouu/TransformerCopilot) |
| 2025-02 | `ABE` | Token-level Ensembling of Models with Different Vocabularies | <a href="https://arxiv.org/abs/2502.21265"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/mjpost/abe?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/mjpost/abe) |
| 2025-02 | `CITER` | CITER: Collaborative Inference for Efficient Large Language Model Decoding with Token-Level Routing | <a href="https://arxiv.org/abs/2502.01976"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/aiming-lab/CITER?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/aiming-lab/CITER) |
| 2024-10 | `UniTe` | Determine-Then-Ensemble: Necessity of Top-k Union for Large Language Model Ensembling | <a href="https://arxiv.org/abs/2410.03777"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2024-06 | `GaC` | Breaking the Ceiling of the LLM Community by Treating Token Generation as a Classification for Ensembling | <a href="https://arxiv.org/abs/2406.12585"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/yaoching0/GaC?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/yaoching0/GaC) |
| 2024-04 | `DeePEn` | Ensemble Learning for Heterogeneous Large Language Models with Deep Parallel Collaboration | <a href="https://arxiv.org/abs/2404.12715"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/OrangeInSouth/DeePEn?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/OrangeInSouth/DeePEn) |
| 2024-04 | `PackLLM` | Pack of LLMs: Model Fusion at Test-Time via Perplexity Optimization | <a href="https://arxiv.org/abs/2404.11531"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/cmavro/PackLLM?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/cmavro/PackLLM) |
| 2024-04 | `EVA` | Bridging the Gap between Different Vocabularies for LLM Ensemble | <a href="https://arxiv.org/abs/2404.09492"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/xydaytoy/EVA?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/xydaytoy/EVA) |
| 2024-02 | `-` | Purifying large language models by ensembling a small language model | <a href="https://arxiv.org/abs/2402.14845"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |

### 2.2.2 (b,2) Span-Level Ensemble

| Date | Name | Title | Paper/Github |
|:---:|:---:|---|:---:|
| 2026-01 | `AdaFuse` | AdaFuse: Adaptive Ensemble Decoding with Test-Time Scaling for LLMs | <a href="https://arxiv.org/abs/2601.06022"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/CCM0111/AdaFuse?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/CCM0111/AdaFuse) |
| 2025-06 | `RLAE` | RLAE: Reinforcement Learning-Assisted Ensemble for LLMs | <a href="https://arxiv.org/abs/2506.00439"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2025-02 | `Speculative Ensemble` | Speculative Ensemble: Fast Large Language Model Ensemble via Speculation | <a href="https://arxiv.org/abs/2502.01662"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/Kamichanw/Speculative-Ensemble?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/Kamichanw/Speculative-Ensemble/) |
| 2024-12 | `SpecFuse` | SpecFuse: Ensembling Large Language Models via Next-Segment Prediction | <a href="https://arxiv.org/abs/2412.07380"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2024-09 | `SweetSpan` | Hit the Sweet Spot! Span-Level Ensemble for Large Language Models | <a href="https://arxiv.org/abs/2409.18583"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2024-07 | `Cool-Fusion` | Cool-Fusion: Fuse Large Language Models without Training | <a href="https://arxiv.org/abs/2407.19807"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |

### 2.2.3 (b,3) Process-Level Ensemble

| Date | Name | Title | Paper/Github |
|:---:|:---:|---|:---:|
| 2026-05 | `-` | Empirical Study of Process Reward-Guided Model Ensemble for Edge Reasoning | <a href="https://www.semanticscholar.org/paper/cb402251297a6349231d9f48995c264c3117eb4a"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=semantic-scholar&logoColor=white" height="18"></a> - |
| 2026-03 | `RouteGoT` | RouteGoT: Node-Adaptive Routing for Cost-Efficient Graph of Thoughts Reasoning | <a href="https://arxiv.org/abs/2603.05818"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2025-11 | `CBS` | Collaborative Beam Search: Enhancing LLM Reasoning via Collective Consensus | <a href="https://aclanthology.org/2025.emnlp-main.574/"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2024-12 | `LE-MCTS` | Ensembling Large Language Models with Process Reward-Guided Tree Search for Better Complex Reasoning | <a href="https://arxiv.org/abs/2412.15797"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |

&nbsp;

## 2.3 Ensemble After Inference

Figure 5: Summary analysis of the key attributes of ensemble-after-inference methods. (Please note that this table may not be fully updated to include all the papers listed below.)
</div>

### 2.3.1 (c,1) Non Cascade

| Date | Name | Title | Paper/Github |
|:---:|:---:|---|:---:|
| 2026-05 | `MCA` | Mixture of Complementary Agents for Robust LLM Ensemble | <a href="https://arxiv.org/abs/2605.24048"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2026-03 | `SCoOP` | SCoOP: Semantic Consistent Opinion Pooling for Uncertainty Quantification in Multiple Vision-Language Model Systems | <a href="https://arxiv.org/abs/2603.23853"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/chungenyu6/SCoOP?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/chungenyu6/SCoOP) |
| 2025-12 | `LLM-PeerReview` | Scoring, Reasoning, and Selecting the Best! Ensembling Large Language Models via a Peer-Review Process | <a href="https://arxiv.org/abs/2512.23213"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/zeyuji/LLM-PeerReview?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/zeyuji/LLM-PeerReview) |
| 2025-10 | `LLMartini` | LLMartini: Seamless and Interactive Leveraging of Multiple LLMs through Comparison and Composition | <a href="https://arxiv.org/abs/2510.19252"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2025-10 | `-` | Beyond Consensus: Mitigating the Agreeableness Bias in LLM Judge Evaluations | <a href="https://arxiv.org/abs/2510.11822"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/ai-cet/paper-arxiv-llm-judge-calibration?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/ai-cet/paper-arxiv-llm-judge-calibration) |
| 2025-10 | `OW/ISP` | Beyond Majority Voting: LLM Aggregation by Leveraging Higher-Order Information | <a href="https://www.arxiv.org/abs/2510.01499"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2025-09 | `FLAME` | Explainable Fault Localization for Programming Assignments via LLM-Guided Annotation | <a href="https://arxiv.org/pdf/2509.25676v1"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/FLAME-FL/FLAME?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/FLAME-FL/FLAME) |
| 2025-09 | `CARGO` | CARGO: A Framework for Confidence-Aware Routing of Large Language Models | <a href="https://arxiv.org/abs/2509.14899"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2025-07 | `LENS` | LENS: Learning Ensemble Confidence from Neural States for Multi-LLM Answer Integration | <a href="https://arxiv.org/abs/2507.23167"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2025-05 | `EL4NER` | EL4NER: Ensemble Learning for Named Entity Recognition via Multiple Small-Parameter Large Language Models | <a href="https://arxiv.org/abs/2505.23038"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2025-04 | `WiSE-FT` | Weight Ensembling Improves Reasoning in Language Models | <a href="https://arxiv.org/abs/2504.10478"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2025-03 | `Symbolic-MoE` | Symbolic Mixture-of-Experts: Adaptive Skill-based Routing for Heterogeneous Reasoning | <a href="https://arxiv.org/abs/2503.05641"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/dinobby/Symbolic-MoE?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/dinobby/Symbolic-MoE/) |
| 2025-01 | `DFPE` | DFPE: A Diverse Fingerprint Ensemble for Enhancing LLM Performance | <a href="https://arxiv.org/abs/2501.17479"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/nivgold/DFPE?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/nivgold/DFPE) |
| 2025-01 | `DMoA` | Balancing Act: Diversity and Consistency in Large Language Model Ensembles | <a href="https://openreview.net/pdf?id=Dl6nkKKvlX"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=openreview&logoColor=white" height="18"></a> - |
| 2024-12 | `Smoothie` | Smoothie: Label Free Language Model Routing | <a href="https://arxiv.org/abs/2412.04692"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/HazyResearch/smoothie?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/HazyResearch/smoothie) |
| 2024-10 | `LLM-Forest` | LLM-Forest: Ensemble Learning of LLMs with Graph-Augmented Prompts for Data Imputation | <a href="https://arxiv.org/abs/2410.21520"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/Xinrui17/LLM-Forest?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/Xinrui17/LLM-Forest) |
| 2024-10 | `LLM-TOPLA` | LLM-TOPLA: Efficient LLM Ensemble by Maximising Diversity | <a href="https://arxiv.org/abs/2410.03953"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/git-disl/llm-topla?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/git-disl/llm-topla) |
| 2024-10 | `MLKF` | Two Heads are Better than One: Zero-shot Cognitive Reasoning via Multi-LLM Knowledge Fusion | <a href="https://dl.acm.org/doi/abs/10.1145/3627673.3679744"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2024-08 | `URG` | URG: A Unified Ranking and Generation Method for Ensembling Language Models | <a href="https://aclanthology.org/2024.findings-acl.261/"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2024-02 | `Agent-Forest` | More Agents Is All You Need | <a href="https://arxiv.org/abs/2402.05120"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/MoreAgentsIsAllYouNeed/AgentForest?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/MoreAgentsIsAllYouNeed/AgentForest) |
| 2023-06 | `LLM-Blender` | LLM-Blender: Ensembling Large Language Models with Pairwise Ranking and Generative Fusion | <a href="https://arxiv.org/abs/2306.02561"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/yuchenlin/LLM-Blender?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/yuchenlin/LLM-Blender) |
| 2023-05 | `MoRE` | Getting MoRE out of Mixture of Language Model Reasoning Experts | <a href="https://arxiv.org/abs/2305.14628"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/NoviScl/MoRE?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/NoviScl/MoRE) |

### 2.3.2 (c,2) Cascade

| Date | Name | Title | Paper/Github |
|:---:|:---:|---|:---:|
| 2026-5 | `-` | Is Escalation Worth It? A Decision-Theoretic Characterization of LLM Cascades | <a href="https://arxiv.org/abs/2605.06350"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2025-12 | `RoBoN` | RoBoN: Routed Online Best-of-n for Test-Time Scaling with Multiple LLMs | <a href="https://www.arxiv.org/abs/2512.05542"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/j-geuter/RoBoN?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/j-geuter/RoBoN) |
| 2025-09 | `-` | Semantic Agreement Enables Efficient Open-Ended LLM Cascades | <a href="https://www.arxiv.org/abs/2509.21837"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2025-04 | `EMAFusionTM` | EMAFusionTM: A Self-Optimizing System for Seamless LLM Selection and Integration | <a href="https://arxiv.org/abs/2504.10681"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2025-04 | `ModelSwitch` | Do We Truly Need So Many Samples? Multi-LLM Repeated Sampling Efficiently Scales Test-Time Compute | <a href="https://arxiv.org/abs/2504.00762"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/JianhaoChen-nju/ModelSwitch?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/JianhaoChen-nju/ModelSwitch) |
| 2024-12 | `DER` | Dynamic Ensemble Reasoning for LLM Experts | <a href="https://arxiv.org/abs/2412.07448"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2024-10 | `Cascade Routing` | A Unified Approach to Routing and Cascading for LLMs | <a href="https://arxiv.org/abs/2410.10347"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/eth-sri/cascade-routing?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/eth-sri/cascade-routing) |
| 2024-04 | `-` | Language Model Cascades: Token-level uncertainty and beyond | <a href="https://arxiv.org/abs/2404.10136"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2023-10 | `AutoMix` | AutoMix: Automatically Mixing Language Models | <a href="https://arxiv.org/abs/2310.12963"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/automix-llm/automix?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/automix-llm/automix) |
| 2023-10 | `neural caching` | Cache & Distil: Optimising API Calls to Large Language Models | <a href="https://arxiv.org/abs/2310.13561"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/guillemram97/neural-caching?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/guillemram97/neural-caching) |
| 2023-10 | `-` | Large Language Model Cascades with Mixture of Thoughts Representations for Cost-efficient Reasoning | <a href="https://arxiv.org/abs/2310.03094"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/MurongYue/LLM_MoT_cascade?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/MurongYue/LLM_MoT_cascade) |
| 2023-10 | `EcoAssistant` | EcoAssistant: Using LLM Assistant More Affordably and Accurately | <a href="https://arxiv.org/abs/2310.03046"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/JieyuZ2/EcoAssistant?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/JieyuZ2/EcoAssistant) |
| 2023-05 | `FrugalGPT` | FrugalGPT: How to Use Large Language Models While Reducing Cost and Improving Performance | <a href="https://arxiv.org/abs/2305.05176"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2023-01 | `-` | When Does Confidence-Based Cascade Deferral Suffice? | <a href="https://proceedings.neurips.cc/paper_files/paper/2023/hash/1f09e1ee5035a4c3fe38a5681cae5815-Abstract-Conference.html"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2022-10 | `Model Cascading` | Model Cascading: Towards Jointly Improving Efficiency and Accuracy of NLP Systems | <a href="https://arxiv.org/abs/2210.05528"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |

&nbsp;

## 2.4 Others: Benchmarks, Applications, Systems and Related Surveys

### 2.4.1 Benchmarks

| Date | Name | Title | Paper/Github |
|:---:|:-------------:|---|:---:|
| 2026-01 | `-` | LLMRouterBench: A Massive Benchmark and Unified Framework for LLM Routing | <a href="https://arxiv.org/abs/2601.07206"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/ynulihao/LLMRouterBench?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/ynulihao/LLMRouterBench) |
| 2025-12 | `-` | Scoring, Reasoning, and Selecting the Best! Ensembling Large Language Models via a Peer-Review Process | <a href="https://arxiv.org/abs/2512.23213"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/zeyuji/LLM-PeerReview?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/zeyuji/LLM-PeerReview) |
| 2025-09 | `-` | RouterArena: An Open Platform for Comprehensive Comparison of LLM Routers | <a href="https://arxiv.org/abs/2510.00202"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/RouteWorks/RouterArena?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/RouteWorks/RouterArena) |
| 2025-07 | `-` | FusionFactory: Fusing LLM Capabilities with Multi-LLM Log Data | <a href="https://arxiv.org/abs/2507.10540"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/ulab-uiuc/FusionFactory?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/ulab-uiuc/FusionFactory) |
| 2025-03 | `RouterEval` | RouterEval: A Comprehensive Benchmark for Routing LLMs to Explore Model-level Scaling Up in LLMs | <a href="https://arxiv.org/abs/2503.10657"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/MilkThink-Lab/RouterEval?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/MilkThink-Lab/RouterEval) |
| 2024-03 | `RouterBench` | RouterBench: A Benchmark for Multi-LLM Routing System | <a href="https://arxiv.org/abs/2403.12031"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/withmartian/routerbench?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/withmartian/routerbench) |
| 2023-06 | `MixInstruct` | LLM-BLENDER: Ensembling Large Language Models with Pairwise Ranking and Generative Fusion | <a href="https://arxiv.org/abs/2306.02561"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/yuchenlin/LLM-Blender?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/yuchenlin/LLM-Blender) |

### 2.4.2 Applications

Beyond the methods presented before, the concept of LLM Ensemble has found applications in a variety of more specialized tasks and domains.
Here we give some examples:

| Date | Name | Title | Paper/Github |
|:---:|:---:|---|:---:|
| 2026-06 | `BEATS` | BEATS: Bootstrapping E-commerce Attribute Taxonomies for Search through Iterative Human-AI Collaboration | <a href="https://arxiv.org/abs/2606.04909"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2026-05 | `-` | Extracting Social Determinants of Health From Electronic Health Records: Development and Comparison of Rule-Based and Large Language Model Methods | <a href="https://medinform.jmir.org/2026/1/e89534"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=doi&logoColor=white" height="18"></a> - |
| 2026 | `-` | Scalable code generation with large language models: an open-source ensemble and reinforcement learning-based selector | <a href="https://scholar.google.com/scholar?q=Scalable+code+generation+with+large+language+models+an+open-source+ensemble+and+reinforcement+learning-based+selector"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=google-scholar&logoColor=white" height="18"></a> - |
| 2026-04 | `-` | Pneumonia Detection in Paediatric Chest X-Rays using Ensembled Large Language Models | <a href="https://www.semanticscholar.org/paper/8721e6e3a45b99b3831e6dcd780bc3c5edf53e1b"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=semantic-scholar&logoColor=white" height="18"></a> - |
| 2026-02 | `-` | Ensemble Strategies for LLM-based Text Summarization: A Timing-Based Framework | <a href="https://www.semanticscholar.org/paper/86520973121f614fd8951f44e3ce4b77461f666e"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=semantic-scholar&logoColor=white" height="18"></a> - |
| 2026-01 | `-` | Next Generation Active Learning: Mixture of LLMs in the Loop | <a href="https://arxiv.org/abs/2601.15773"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2025-09 | `FLAME` | Explainable Fault Localization for Programming Assignments via LLM-Guided Annotation | <a href="https://arxiv.org/pdf/2509.25676v1"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/FLAME-FL/FLAME?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/FLAME-FL/FLAME) |
| 2025-05 | `Expert Orchestration` | Beyond Monoliths: Expert Orchestration for More Capable, Democratic, and Safe Large Language Models | <a href="https://arxiv.org/abs/2506.00051"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2025-04 | `Consensus Entropy` | Consensus Entropy: Harnessing Multi-VLM Agreement for Self-Verifying and Self-Improving OCR | <a href="https://arxiv.org/abs/2504.11101"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2024-11 | `BWRS` | Bayesian Calibration of Win Rate Estimation with LLM Evaluators | <a href="https://arxiv.org/abs/2411.04424"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/yale-nlp/bay-calibration-llm-evaluators?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/yale-nlp/bay-calibration-llm-evaluators) |
| 2024-06 | `FuseGen` | FuseGen: PLM Fusion for Data-generation based Zero-shot Learning | <a href="https://arxiv.org/abs/2406.12527"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/LindaLydia/FuseGen?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/LindaLydia/FuseGen) |
| 2024-05 | `-` | PromptMind Team at MEDIQA-CORR 2024: Improving Clinical Text Correction with Error Categorization and LLM Ensembles | <a href="https://arxiv.org/abs/2405.08373"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2024-02 | `-` | LLM-Ensemble: Optimal Large Language Model Ensemble Method for E-commerce Product Attribute Value Extraction | <a href="https://arxiv.org/abs/2403.00863"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2023-11 | `-` | On Preserving the Knowledge of Long Clinical Texts | <a href="https://arxiv.org/abs/2311.01571"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2023-10 | `Ensemble-Instruct` | Ensemble-Instruct: Generating Instruction-Tuning Data with a Heterogeneous Mixture of LMs | <a href="https://arxiv.org/abs/2310.13961"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/IBM/ensemble-instruct?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/IBM/ensemble-instruct) |

### 2.4.3 Systems

| Date | Name | Title | Paper/Github |
|:---:|:---:|---|:---:|
| 2026-06 | `SMCS` | A Scalable Multi-LLM Collaboration System with Retrieval-based Selection and Exploration-Exploitation-Driven Enhancement | <a href="https://aclanthology.org/2026.acl-long.579/"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=acl&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/magent4aci/SMCS?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/magent4aci/SMCS) |
| 2025-12 | `TADEL` | TADEL: Task-Aware Dynamic Ensemble of Lightweight LLMs for Improved Inference Accuracy in Edge AI | <a href="https://www.semanticscholar.org/paper/ecdcbc6d1954ed220c782e7943e611b7299a712f"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=semantic-scholar&logoColor=white" height="18"></a> - |
| 2025-10 | `LLMartini` | LLMartini: Seamless and Interactive Leveraging of Multiple LLMs through Comparison and Composition | <a href="https://www.arxiv.org/abs/2510.01499"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |

### 2.4.4 Related Surveys

| Date | Title | Paper/Github |
|:---:|---|:---:|
| 2026-02 | Dynamic Model Routing and Cascading for Efficient LLM Inference: A Survey | <a href="https://arxiv.org/abs/2603.04445"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2025-07 | Toward Edge General Intelligence with Multiple-Large Language Model (Multi-LLM): Architecture, Trust, and Orchestration | <a href="https://arxiv.org/abs/2507.00672"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2025-06 | Towards Efficient Multi-LLM Inference: Characterization and Analysis of LLM Routing and Hierarchical Techniques | <a href="https://www.arxiv.org/abs/2506.06579"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2025-05 | A Survey on Collaborative Mechanisms Between Large and Small Language Models | <a href="https://arxiv.org/abs/2505.07460"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2025-03 | A Survey on Test-Time Scaling in Large Language Models: What, How, Where, and How Well | <a href="https://arxiv.org/abs/2503.24235"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/testtimescaling/testtimescaling.github.io?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/testtimescaling/testtimescaling.github.io) |
| 2025-02 | Doing More with Less – Implementing Routing Strategies in Large Language Model-Based Systems: An Extended Survey | <a href="https://arxiv.org/abs/2502.00409"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2025-02 | Doing More with Less: A Survey on Routing Strategies for Resource Optimisation in Large Language Model-Based Systems | <a href="https://arxiv.org/abs/2502.00409"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2024-08 | Model Merging in LLMs, MLLMs, and Beyond: Methods, Theories, Applications and Opportunities | <a href="https://arxiv.org/abs/2408.07666"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/EnnengYang/Awesome-Model-Merging-Methods-Theories-Applications?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/EnnengYang/Awesome-Model-Merging-Methods-Theories-Applications) |
| 2024-08 | A Survey on Model MoErging: Recycling and Routing Among Specialized Experts for Collaborative Learning | <a href="https://arxiv.org/abs/2408.07057"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2024-07 | Merge, Ensemble, and Cooperate! A Survey on Collaborative Strategies in the Era of Large Language Models | <a href="https://arxiv.org/abs/2407.06089"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2023-09 | Deep Model Fusion: A Survey | <a href="https://arxiv.org/abs/2309.15698"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2023-02 | A comprehensive review on ensemble deep learning: Opportunities and challenges | <a href="https://www.sciencedirect.com/science/article/pii/S1319157823000228"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |

&nbsp; 
&nbsp;  
&nbsp;

## 3 Others: Some public implementations of the LLM Ensemble methods

| Date | Title |                                                                                  Github                                                                                   |
|:---:|---|:-------------------------------------------------------------------------------------------------------------------------------------------------------------------------:|
| 2025 | Ensemble-Hub | [![GITHUB Stars](https://img.shields.io/github/stars/Fzkuji/Ensemble-Hub?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/Fzkuji/Ensemble-Hub) |

&nbsp; 
&nbsp;  
&nbsp;

## 4 Others: Some other related interesting papers 
Here we briefly list some related papers, which are either discovered by us or suggested by the authors to this repository. 
They mainly focus on **LLM Collaboration**.

### 4.1 Test-Time Scaling

| Date | Name | Title | Paper/Github |
|:---:|:---:|---|:---:|
| 2026-05 | `-` | Agentic Systems as Boosting Weak Reasoning Models | <a href="https://arxiv.org/abs/2605.14163"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2025-12 | `-` | The Law of Multi-Model Collaboration: Scaling Limits of Model Ensembling for Large Language Models | <a href="https://arxiv.org/abs/2512.23340"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2025-10 | `-` | Stable LLM Ensemble: Interaction between Example Representativeness and Diversity | <a href="https://arxiv.org/abs/2510.13143"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |

### 4.2 LLM Collaboration and Others

| Date | Name | Title | Paper/Github |
|:---:|:---:|---|:---:|
| 2026-07 | `CanonicalMerge` | Cache Merging as a Convergent Replicated State for Multi-Agent Latent Reasoning | <a href="https://arxiv.org/abs/2607.01308"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2026-06 | `-` | Understanding Input-Output Uncertainty in LLM Routing Supervision | <a href="https://scholar.google.com/scholar?q=Understanding+Input-Output+Uncertainty+in+LLM+Routing+Supervision"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=google-scholar&logoColor=white" height="18"></a> - |
| 2026-06 | `PerspectiveGap` | PerspectiveGap: A Benchmark for Multi-Agent Orchestration Prompting | <a href="https://arxiv.org/abs/2606.08878"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/WhymustIhaveaname/PerspectiveGap?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/WhymustIhaveaname/PerspectiveGap) |
| 2026-04 | `Tandem` | Tandem: Riding Together with Large and Small Language Models for Efficient Reasoning | <a href="https://arxiv.org/abs/2604.23623"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/Applied-Machine-Learning-Lab/ACL2026_Tandem?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/Applied-Machine-Learning-Lab/ACL2026_Tandem) |
| 2026-03 | `CCPO` | Counterfactual Credit Policy Optimization for Multi-Agent Collaboration | <a href="https://arxiv.org/abs/2603.21563"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/bhai114/ccpo?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/bhai114/ccpo) |
| 2026-03 | `ARE` | Adaptive Robust Estimator for Multi-Agent Reinforcement Learning | <a href="https://arxiv.org/abs/2603.21574"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2025-12 | `Consensus-Driven Reasoning` | Towards Responsible and Explainable AI Agents with Consensus-Driven Reasoning | <a href="https://arxiv.org/abs/2512.21699"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2025-12 | `CogER` | Beyond Fast and Slow: Cognitive-Inspired Elastic Reasoning for Large Language Models | <a href="https://arxiv.org/abs/2512.15089"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2025-12 | `WISE` | WISE: Weighted Iterative Society-of-Experts for Robust Multimodal Multi-Agent Debate | <a href="https://arxiv.org/abs/2512.02405"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2025-10 | `ColMAD` | Towards Scalable Oversight with Collaborative Multi-Agent Debate in Error Detection | <a href="https://arxiv.org/abs/2510.20963"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2025-10 | `AdCo` | Adaptive Coopetition: Leveraging Coarse Verifier Signals for Resilient Multi-Agent LLM Reasoning | <a href="https://arxiv.org/abs/2510.18179"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/AdCo-Research/adaptive-coopetition?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/AdCo-Research/adaptive-coopetition) |
| 2025-10 | `-` | Harmonizing Diverse Models: A Layer-wise Merging Strategy for Consistent Generation | <a href="https://arxiv.org/abs/2510.14915"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2025-09 | `MoT` | Mixture of Thoughts: Learning to Aggregate What Experts Think, Not Just What They Say | <a href="https://arxiv.org/abs/2509.21164"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/jacobfa/mot?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/jacobfa/mot) |
| 2025-09 | `Best-of-∞` | Best-of-∞ -- Asymptotic Performance of Test-Time LLM Ensembling | <a href="https://arxiv.org/abs/2509.21091"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/jkomiyama/BoInf-code-publish?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/jkomiyama/BoInf-code-publish) |
| 2025-08 | `SLC` | Small-Large Collaboration: Training-efficient Concept Personalization for Large VLM | <a href="https://arxiv.org/abs/2508.07260"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/Hhankyangg/SLC?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/Hhankyangg/SLC) |
| 2025-03 | `GENOME` | Nature-Inspired Population-Based Evolution of Large Language Models | <a href="https://arxiv.org/abs/2503.01155"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/ZhangYiqun018/GENOME?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/ZhangYiqun018/GENOME) |
| 2025-02 | `Heter-MAD` | If Multi-Agent Debate is the Answer, What is the Question? | <a href="https://arxiv.org/abs/2502.08788"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> - |
| 2024-10 | `LLM-Forest` | LLM-Forest: Ensemble Learning of LLMs with Graph-Augmented Prompts for Data Imputation | <a href="https://arxiv.org/abs/2410.21520"><img src="https://img.shields.io/badge/PAPER-A42C25?style=for-the-badge&logo=arxiv&logoColor=white" height="18"></a> [![GITHUB Stars](https://img.shields.io/github/stars/Xinrui17/LLM-Forest?style=flat-square&logo=github&label=GITHUB&color=black)](https://github.com/Xinrui17/LLM-Forest) |

&nbsp; 
&nbsp;  
&nbsp;

## 5 Summarization

Figure 6: Summary analysis of the key attributes of LLM Ensemble approaches.
</div>

&nbsp; 
&nbsp;  
&nbsp;

---

&nbsp; 
<div align=center><img src="./fig/logobeihang.png" width="60%">

</div>

&nbsp;

Awesome-LLM-Ensemble

Explore Similar Repositories

Was this content helpful?

Recent Feedback

Download README