---
title: "Latent Agents — paper by Yi et al. (2026) — summary and ideas"
description: "The authors propose Internalized Multi-Agent Debate (IMAD), which compresses this process into one… 41 references, cited by 2 papers here."
authors: ["John Seon Keun Yi", "Aaron Mueller", "Dokyun Lee"]
year: 2026
venue: "Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)"
source: "https://aclanthology.org/2026.acl-long.709/"
doi: "10.18653/v1/2026.acl-long.709"
licence: "CC BY 4.0 (ACL Anthology)"
licence_url: "https://creativecommons.org/licenses/by/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/yi-2026-latent-agents-post-training-procedure/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/yi-2026-latent-agents-post-training-procedure/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2026](https://md.research.elamur.ai/free/en/papers/2026/page/8/) / Latent Agents: A Post-Training Procedure for Internalized Multi-Agent Debate

# Latent Agents: A Post-Training Procedure for Internalized Multi-Agent Debate

**Authors:** John Seon Keun Yi, [Aaron Mueller](https://md.research.elamur.ai/free/en/authors/aaron-mueller/) and Dokyun Lee · **Year:** 2026 · **Venue:** [Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)](https://md.research.elamur.ai/free/en/venues/acl/) ([ACL 2026](https://md.research.elamur.ai/free/en/venues/acl/2026/)) · **Publisher:** [Association for Computational Linguistics](https://md.research.elamur.ai/free/en/publishers/association-for-computational-linguistics/) · **Licence:** [CC BY 4.0 (ACL Anthology)](https://creativecommons.org/licenses/by/4.0/)

**Links:** [Publisher's page](https://aclanthology.org/2026.acl-long.709/) · [PDF](https://aclanthology.org/2026.acl-long.709.pdf) · [DOI 10.18653/v1/2026.acl-long.709](https://doi.org/10.18653/v1/2026.acl-long.709) · [OpenAlex](https://openalex.org/W7166864388)

**Citations:** 0 citations (OpenAlex, 2026-10-14) · 2 in our library, of the 149,439 papers whose bibliographies we read

## Summary
Multi-agent debate, where several LLM instances argue over multiple rounds before settling on an answer, improves factual accuracy. However, it is expensive because long transcripts must be produced first. The authors propose Internalized Multi-Agent Debate (IMAD), which compresses this process into one model. They collect debates among GPT-3.5 turbo agents on synthetic arithmetic problems and mark them with structure tags. A single model is then fine-tuned on the full transcripts, so it learns to write the entire debate on its own. Next, a GRPO stage pushes the debate into the model's hidden computation. It rewards correct answers that appear within a token budget that shrinks over training, while the weight on the formatting reward is gradually reduced. The internalized models match or beat explicit debate while using up to 93% fewer tokens. Activation-steering analysis shows that each internalized agent occupies a distinct direction in activation space. Because of this, deliberately injected malicious agents can be suppressed with less loss of general capability than when steering base models. The work therefore matters for both efficiency and safety.

## Key points
- **Debate data:** 944 debate traces were generated with 3 agents over 2 rounds, using GPT-3.5 turbo as the agent model.
  - Each problem is an arithmetic expression built from six random two-digit numbers.
  - Debates without a final-round majority consensus were discarded.
  - Tags such as \<|Agent 1|>, \<|Round 1|>, \<|Consensus|> and \<|endofdebate|> mark the debate structure.
- **Two-stage training:**
  - SFT uses next-token prediction on whole debate traces, not only on final answers.
  - GRPO then optimizes a reward that combines a decaying format reward with a correctness reward. That reward only counts if the right answer appears within the first *l* tokens, and *l* is annealed downward during training.
- **Efficiency result:** IMAD performs on par with or better than explicit multi-agent debate across several models and benchmarks, including GSM8K, with up to 93% fewer tokens.
- **Interpretability and control:**
  - Difference-in-means steering vectors reveal separable agent subspaces; without structure tags, the separation is weaker.
  - Negative steering suppresses injected "evil" traits fully and hallucination traits partially, while keeping output coherence.
- **Limitations:**
  - The method is tested only on arithmetic data with a fixed debate format.
  - It depends on successful structure learning; LLaMA was the most reliable model at reproducing the debate format.
  - Trait scores come from an LLM judge, which a human study found to agree closely with human ratings.
  - Gains appear mainly in models of 7B parameters or more.
  - Code: github.com/johnsk95/latent_agents.

*Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.*

## Institutions

- [Boston University](https://md.research.elamur.ai/free/en/universities/boston-university/) — University, Boston, United States · 26 works · 4,022 citations

📚 Shelves (4): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 16,667 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 16,194 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 16,039 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 15,891 works

## Cite

Cite as: John Seon Keun Yi, Aaron Mueller and Dokyun Lee. 2026. Latent Agents: A Post-Training Procedure for Internalized Multi-Agent Debate. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers).

The paper itself, not this page. BibTeX from the library's record of the work.

```bibtex
@inproceedings{yi2026latent,
  title = {{Latent Agents: A Post-Training Procedure for Internalized Multi-Agent Debate}},
  author = {John Seon Keun Yi, Aaron Mueller and Dokyun Lee},
  booktitle = {Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)},
  year = {2026},
  doi = {10.18653/v1/2026.acl-long.709},
  url = {https://doi.org/10.18653/v1/2026.acl-long.709}
}
```

- [← previous Language Reconstruction with Brain Predictive Coding from fMRI Data](https://md.research.elamur.ai/free/en/papers/yin-2026-language-reconstruction-brain-predictive-coding/)
- [next → Law in Silico: Simulating Legal Society with LLM-Based Agents](https://md.research.elamur.ai/free/en/papers/wang-2026-law-silico-simulating-legal-society/)
- [↑ back to papers of 2026](https://md.research.elamur.ai/free/en/papers/2026/page/8/)

## References

41 references, 23 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Andy Arditi, Oscar Obeso, Aaquib Syed, Daniel Paleka, Nina Panickssery, Wes Gurnee, and Neel Nanda. 2024. [Refusal in language models is mediated by a single direction](https://md.research.elamur.ai/free/en/papers/arditi-2024-refusal-language-models-mediated-single/). Advances in Neural Information Processing Systems, 37:136037–136083.
2. Tomer Ashuach, Dana Arad, Aaron Mueller, Martin Tutek, and Yonatan Belinkov. 2025. Crisp: Persistent concept unlearning via sparse autoencoders. arXiv preprint arXiv:2508.13650. [find](https://scholar.google.com/scholar?q=Tomer+Ashuach%2C+Dana+Arad%2C+Aaron+Mueller%2C+Martin+Tutek%2C+and+Yonatan+Belinkov%2E+2025%2E+Crisp%3A+Persistent+concept+unlearning+via+sparse+autoencoders%2E+arXiv+preprint+arXiv%3A2508%2E13650%2E)
3. Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D. Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, and 1 others. 2020. [Language models are few-shot learners](https://md.research.elamur.ai/free/en/papers/brown-2020-language-models-few-shot-learners/). In Advances in Neural Information Processing Systems, volume 33, pages 1877–1901. Curr…
4. Runjin Chen, Andy Arditi, Henry Sleight, Owain Evans, and Jack Lindsey. 2025. Persona vectors: Monitoring and controlling character traits in language models. arXiv preprint arXiv:2507.21509. — [arXiv:2507.21509](https://arxiv.org/abs/2507.21509) [find](https://scholar.google.com/scholar?q=Runjin+Chen%2C+Andy+Arditi%2C+Henry+Sleight%2C+Owain+Evans%2C+and+Jack+Lindsey%2E+2025%2E+Persona+vectors%3A+Monitoring+and+controlling+character+traits+in+language+models%2E+arXiv+preprint+arXiv%3A2507%2E21509%2E+%E2%80%94+%5BarXiv)
5. Wenhu Chen, Xueguang Ma, Xinyi Wang, and William W Cohen. 2022. Program of thoughts prompting: Disentangling computation from reasoning for numerical reasoning tasks. arXiv preprint arXiv:2211.12588. [find](https://scholar.google.com/scholar?q=Wenhu+Chen%2C+Xueguang+Ma%2C+Xinyi+Wang%2C+and+William+W+Cohen%2E+2022%2E+Program+of+thoughts+prompting%3A+Disentangling+computation+from+reasoning+for+numerical+reasoning+tasks%2E+arXiv+preprint+arXiv%3A2211%2E12588%2E)
6. Karl Cobbe, Vineet Kosaraju, Mohammad Bavarian, Mark Chen, Heewoo Jun, Lukasz Kaiser, Matthias Plappert, Jerry Tworek, Jacob Hilton, Reiichiro Nakano, and 1 others. 2021. [Training verifiers to solve math word problems](https://md.research.elamur.ai/free/en/papers/cobbe-2021-training-verifiers-solve-math-word/). Preprint, arXiv:2110.14168.
7. Yuntian Deng, Yejin Choi, and Stuart Shieber. 2024. From explicit cot to implicit cot: Learning to internalize cot step by step. arXiv preprint arXiv:2405.14838. [find](https://scholar.google.com/scholar?q=Yuntian+Deng%2C+Yejin+Choi%2C+and+Stuart+Shieber%2E+2024%2E+From+explicit+cot+to+implicit+cot%3A+Learning+to+internalize+cot+step+by+step%2E+arXiv+preprint+arXiv%3A2405%2E14838%2E)
8. Yuntian Deng, Kiran Prasad, Roland Fernandez, Paul Smolensky, Vishrav Chaudhary, and Stuart Shieber. 2023. Implicit chain of thought reasoning via knowledge distillation. arXiv preprint arXiv:2311.01460. [find](https://scholar.google.com/scholar?q=Yuntian+Deng%2C+Kiran+Prasad%2C+Roland+Fernandez%2C+Paul+Smolensky%2C+Vishrav+Chaudhary%2C+and+Stuart+Shieber%2E+2023%2E+Implicit+chain+of+thought+reasoning+via+knowledge+distillation%2E+arXiv+preprint+arXiv%3A2311%2E014)
9. Yilun Du, Shuang Li, Antonio Torralba, Joshua B. Tenenbaum, and Igor Mordatch. 2023. [Improving factuality and reasoning in language models through multi-agent debate](https://md.research.elamur.ai/free/en/papers/du-2023-improving-factuality-reasoning-language-models/). Preprint, arXiv:2305.14325.
10. Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Amy Yang, Angela Fan, and 1 others. 2024. [The llama 3 herd of models](https://md.research.elamur.ai/free/en/papers/grattafiori-2024-llama-3-herd-models/). arXiv e-prints, pages arXiv–2407.
11. Daya Guo, Dejian Yang, Haowei Zhang, Junxiao Song, Ruoyu Zhang, Runxin Xu, Qihao Zhu, Shirong Ma, Peiyi Wang, Xiao Bi, and 1 others. 2025. [Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning](https://md.research.elamur.ai/free/en/papers/deepseek-ai-2025-deepseek-r1-incentivizing-reasoning-capability/). arXiv preprint arXiv:2501.12948.
12. Shibo Hao, Sainbayar Sukhbaatar, DiJia Su, Xian Li, Zhiting Hu, Jason Weston, and Yuandong Tian. 2024. Training large language models to reason in a continuous latent space. arXiv preprint arXiv:2412.06769. [find](https://scholar.google.com/scholar?q=Shibo+Hao%2C+Sainbayar+Sukhbaatar%2C+DiJia+Su%2C+Xian+Li%2C+Zhiting+Hu%2C+Jason+Weston%2C+and+Yuandong+Tian%2E+2024%2E+Training+large+language+models+to+reason+in+a+continuous+latent+space%2E+arXiv+preprint+arXiv%3A2412%2E)
13. Tian Liang, Zhiwei He, Wenxiang Jiao, Xing Wang, Yan Wang, Rui Wang, Yujiu Yang, Shuming Shi, and Zhaopeng Tu. 2024. [Encouraging divergent thinking in large language models through multi-agent debate](https://md.research.elamur.ai/free/en/papers/liang-2023-encouraging-divergent-thinking-large-language/). In Proceedings of the 2024 conference on empirical methods in natural language processing, pages 17…
14. Chin-Yew Lin. 2004. [Rouge: A package for automatic evaluation of summaries](https://md.research.elamur.ai/free/en/papers/lin-2004-rouge-package-automatic-evaluation-summaries/). In Text summarization branches out, pages 74–81.
15. Yong Lin, Hangyu Lin, Wei Xiong, Shizhe Diao, Jianmeng Liu, Jipeng Zhang, Rui Pan, Haoxiang Wang, Wenbin Hu, Hanning Zhang, and 1 others. 2024. [Mitigating the alignment tax of rlhf](https://md.research.elamur.ai/free/en/papers/lin-2024-mitigating-alignment-tax-rlhf/). In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 580–606.
16. Yinyi Luo, Yiqiao Jin, Weichen Yu, Mengqi Zhang, Srijan Kumar, Xiaoxiao Li, Weijie Xu, Xin Chen, and Jindong Wang. 2026. Agentark: Distilling multi-agent intelligence into a single llm agent. arXiv preprint arXiv:2602.03955. — [arXiv:2602.03955](https://arxiv.org/abs/2602.03955) [find](https://scholar.google.com/scholar?q=Yinyi+Luo%2C+Yiqiao+Jin%2C+Weichen+Yu%2C+Mengqi+Zhang%2C+Srijan+Kumar%2C+Xiaoxiao+Li%2C+Weijie+Xu%2C+Xin+Chen%2C+and+Jindong+Wang%2E+2026%2E+Agentark%3A+Distilling+multi%2Dagent+intelligence+into+a+single+llm+agent%2E+arXiv+pr)
17. Samuel Marks and Max Tegmark. 2023. [The geometry of truth: Emergent linear structure in large language model representations of true/false datasets](https://md.research.elamur.ai/free/en/papers/marks-2023-geometry-truth-emergent-linear-structure/). arXiv preprint arXiv:2310.06824.
18. Yisong Miao and Min-Yen Kan. 2025. Discursive circuits: How do language models understand discourse relations? In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, pages 32558–32577. [find](https://scholar.google.com/scholar?q=Yisong+Miao+and+Min%2DYen+Kan%2E+2025%2E+Discursive+circuits%3A+How+do+language+models+understand+discourse+relations%3F+In+Proceedings+of+the+2025+Conference+on+Empirical+Methods+in+Natural+Language+Processing)
19. Mistral AI. 2024. Mistral NeMo: Our new best small model. url [https://mistral.ai/news/mistral-nemo/](https://mistral.ai/news/mistral-nemo/) . [↗](https://mistral.ai/news/mistral-nemo/)
20. OpenAI. 2024. Gpt-4o mini: advancing cost-efficient intelligence. [find](https://scholar.google.com/scholar?q=OpenAI%2E+2024%2E+Gpt%2D4o+mini%3A+advancing+cost%2Defficient+intelligence%2E)
21. Karl Moritz Hermann, Tomas Kocisky, Edward Grefenstette, Lasse Espeholt, Will Kay, Mustafa Suleyman, and Phil Blunsom. 2015. [Teaching machines to read and comprehend](https://md.research.elamur.ai/free/en/papers/hermann-2015-teaching-machines-read-comprehend/). In Advances in Neural Information Processing Systems, volume 28. Curran Associates, Inc.
22. Hadas Orgad, Michael Toker, Zorik Gekhman, Roi Reichart, Idan Szpektor, Hadas Kotek, and Yonatan Belinkov. 2024. Llms know more than they show: On the intrinsic representation of llm hallucinations. arXiv preprint arXiv:2410.02707. [find](https://scholar.google.com/scholar?q=Hadas+Orgad%2C+Michael+Toker%2C+Zorik+Gekhman%2C+Roi+Reichart%2C+Idan+Szpektor%2C+Hadas+Kotek%2C+and+Yonatan+Belinkov%2E+2024%2E+Llms+know+more+than+they+show%3A+On+the+intrinsic+representation+of+llm+hallucinations%2E+a)
23. Bairu Hou, Yang Zhang, Jiabao Ji, Yujian Liu, Kaizhi Qian, Jacob Andreas, and Shiyu Chang. 2025. [Thinkprune: Pruning long chain-of-thought of llms via reinforcement learning](https://md.research.elamur.ai/free/en/papers/hou-2025-thinkprune-pruning-long-chain-thought/). arXiv preprint arXiv:2504.01296.
24. Long Ouyang, Jeffrey Wu, Xu Jiang, Diogo Almeida, Carroll Wainwright, Pamela Mishkin, Chong Zhang, Sandhini Agarwal, Katarina Slama, Alex Ray, and 1 others. 2022. [Training language models to follow instructions with human feedback](https://md.research.elamur.ai/free/en/papers/ouyang-2022-training-language-models-follow-instructions/). Advances in neural information processing systems, 35:27730–27744.
25. Edward J Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, Weizhu Chen, and 1 others. 2022. [Lora: Low-rank adaptation of large language models](https://md.research.elamur.ai/free/en/papers/hu-2021-lora-low-rank-adaptation-large/). ICLR, 1(2):3.
26. Weizhen Li, Jianbo Lin, Zhuosong Jiang, Jingyi Cao, Xinpeng Liu, Jiayu Zhang, Zhenqiang Huang, Qianben Chen, Weichen Sun, Qiexiang Wang, and 1 others. 2025. Chain-of-agents: End-to-end agent foundation models via multi-agent distillation and agentic rl. arXiv preprint arXiv:2508.13167. — [arXiv:2508.13167](https://arxiv.org/abs/2508.13167) [find](https://scholar.google.com/scholar?q=Weizhen+Li%2C+Jianbo+Lin%2C+Zhuosong+Jiang%2C+Jingyi+Cao%2C+Xinpeng+Liu%2C+Jiayu+Zhang%2C+Zhenqiang+Huang%2C+Qianben+Chen%2C+Weichen+Sun%2C+Qiexiang+Wang%2C+and+1+others%2E+2025%2E+Chain%2Dof%2Dagents%3A+End%2Dto%2Dend+agent+foundatio)
27. Melissa Z Pan, Mert Cemri, Lakshya A Agrawal, Shuyi Yang, Bhavya Chopra, Rishabh Tiwari, Kurt Keutzer, Aditya Parameswaran, Kannan Ramchandran, Dan Klein, and 1 others. 2025. Why do multiagent systems fail? In ICLR 2025 Workshop on Building Trust in Language Models and Applications. [find](https://scholar.google.com/scholar?q=Melissa+Z+Pan%2C+Mert+Cemri%2C+Lakshya+A+Agrawal%2C+Shuyi+Yang%2C+Bhavya+Chopra%2C+Rishabh+Tiwari%2C+Kurt+Keutzer%2C+Aditya+Parameswaran%2C+Kannan+Ramchandran%2C+Dan+Klein%2C+and+1+others%2E+2025%2E+Why+do+multiagent+systems)
28. Qwen, :, An Yang, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chengyuan Li, Dayiheng Liu, Fei Huang, Haoran Wei, Huan Lin, Jian Yang, Jianhong Tu, Jianwei Zhang, Jianxin Yang, Jiaxi Yang, Jingren Zhou, and 25 others. 2025. Qwen2.5 technical report. Preprint, arXiv:2412.15115. [find](https://scholar.google.com/scholar?q=Qwen%2C+%3A%2C+An+Yang%2C+Baosong+Yang%2C+Beichen+Zhang%2C+Binyuan+Hui%2C+Bo+Zheng%2C+Bowen+Yu%2C+Chengyuan+Li%2C+Dayiheng+Liu%2C+Fei+Huang%2C+Haoran+Wei%2C+Huan+Lin%2C+Jian+Yang%2C+Jianhong+Tu%2C+Jianwei+Zhang%2C+Jianxin+Yang%2C+Jiaxi+)
29. Shauli Ravfogel, Yanai Elazar, Hila Gonen, Michael Twiton, and Yoav Goldberg. 2020. [Null it out: Guarding protected attributes by iterative nullspace projection](https://md.research.elamur.ai/free/en/papers/ravfogel-2020-null-it-out-guarding-protected/). In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 7237–7256, Online. Association for Compu…
30. Nina Rimsky, Nick Gabrieli, Julian Schulz, Meg Tong, Evan Hubinger, and Alexander Turner. 2024. Steering llama 2 via contrastive activation addition. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 15504–15522. [find](https://scholar.google.com/scholar?q=Nina+Rimsky%2C+Nick+Gabrieli%2C+Julian+Schulz%2C+Meg+Tong%2C+Evan+Hubinger%2C+and+Alexander+Turner%2E+2024%2E+Steering+llama+2+via+contrastive+activation+addition%2E+In+Proceedings+of+the+62nd+Annual+Meeting+of+the+A)
31. Melanie Sclar, Hieu Le, Xinyi Zhang, Yifei Yuan, Al-Ghosn Al-Ghosn, Li-Ping Iv, Yao Fu, Wenda Wang, Yixin Chen, and Kyle Chang. 2024. [MMLU-Pro: A more robust and challenging multi-task language understanding benchmark](https://md.research.elamur.ai/free/en/papers/wang-2024-mmlu-pro-more-robust-challenging/). Preprint, arXiv:2405.15284.
32. Zhihong Shao, Peiyi Wang, Qihao Zhu, Runxin Xu, Junxiao Song, Xiao Bi, Haowei Zhang, Mingchuan Zhang, Y. K. Li, Y. Wu, and Daya Guo. 2024. [Deepseekmath: Pushing the limits of mathematical reasoning in open language models](https://md.research.elamur.ai/free/en/papers/shao-2024-deepseekmath-pushing-limits-mathematical-reasoning/). Preprint, arXiv:2402.03300.
33. Gaurav Srivastava, Zhenyu Bi, Meng Lu, and Xuan Wang. 2025. [Debate, train, evolve: Self evolution of language model reasoning](https://md.research.elamur.ai/free/en/papers/srivastava-2025-debate-train-evolve-self-evolution/). arXiv preprint arXiv:2505.15734.
34. Nishant Subramani, Nivedita Suresh, and Matthew E Peters. 2022. [Extracting latent steering vectors from pretrained language models](https://md.research.elamur.ai/free/en/papers/subramani-2022-extracting-latent-steering-vectors-pretrained/). arXiv preprint arXiv:2205.05124.
35. Vighnesh Subramaniam, Yilun Du, Joshua B Tenenbaum, Antonio Torralba, Shuang Li, and Igor Mordatch. 2025. Multiagent finetuning: Self improvement with diverse reasoning chains. arXiv preprint arXiv:2501.05707. — [arXiv:2501.05707](https://arxiv.org/abs/2501.05707) [find](https://scholar.google.com/scholar?q=Vighnesh+Subramaniam%2C+Yilun+Du%2C+Joshua+B+Tenenbaum%2C+Antonio+Torralba%2C+Shuang+Li%2C+and+Igor+Mordatch%2E+2025%2E+Multiagent+finetuning%3A+Self+improvement+with+diverse+reasoning+chains%2E+arXiv+preprint+arXiv%3A25)
36. Vighnesh Subramaniam, Antonio Torralba, and Shuang Li. 2024. Debategpt: Fine-tuning large language models with multi-agent debate supervision. [find](https://scholar.google.com/scholar?q=Vighnesh+Subramaniam%2C+Antonio+Torralba%2C+and+Shuang+Li%2E+2024%2E+Debategpt%3A+Fine%2Dtuning+large+language+models+with+multi%2Dagent+debate+supervision%2E)
37. Mirac Suzgun, Nathan Scales, Nathanael Schärli, Sebastian Gehrmann, Yi Tay, Hyung Won Chung, Aakanksha Chowdhery, Quoc Le, Ed Chi, Denny Zhou, and 1 others. 2023. [Challenging big-bench tasks and whether chain-of-thought can solve them](https://md.research.elamur.ai/free/en/papers/suzgun-2023-challenging-big-bench-tasks-whether/). In Findings of the Association for Computational Linguistics: AC…
38. Alexander Matt Turner, Lisa Thiergart, Gavin Leech, David Udell, Juan J Vazquez, Ulisse Mini, and Monte MacDiarmid. 2023. [Steering language models with activation engineering](https://md.research.elamur.ai/free/en/papers/turner-2023-steering-language-models-activation-engineering/). arXiv preprint arXiv:2308.10248.
39. Jason Wei, Maarten Bosma, Vincent Y Zhao, Kelvin Guu, Adams Wei Yu, Brian Lester, Nan Du, Andrew M Dai, and Quoc V Le. 2021. [Finetuned language models are zero-shot learners](https://md.research.elamur.ai/free/en/papers/wei-2021-finetuned-language-models-zero-shot/). arXiv preprint arXiv:2109.01652.
40. Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Ed Chi, Quoc Le, and Denny Zhou. 2022. [Chain-of-thought prompting elicits reasoning in large language models](https://md.research.elamur.ai/free/en/papers/wei-2022-chain-thought-prompting-elicits-reasoning/). In Advances in Neural Information Processing Systems, volume 35, pages 24824–24837. Curran Associates, Inc.
41. Zhengxuan Wu, Aryaman Arora, Atticus Geiger, Zheng Wang, Jing Huang, Dan Jurafsky, Christopher D Manning, and Christopher Potts. 2025. Axbench: Steering LLMs? even simple baselines outperform sparse autoencoders. In Forty-second International Conference on Machine Learning. [find](https://scholar.google.com/scholar?q=Zhengxuan+Wu%2C+Aryaman+Arora%2C+Atticus+Geiger%2C+Zheng+Wang%2C+Jing+Huang%2C+Dan+Jurafsky%2C+Christopher+D+Manning%2C+and+Christopher+Potts%2E+2025%2E+Axbench%3A+Steering+LLMs%3F+even+simple+baselines+outperform+sparse+a)

*Updated 31 October 2026*