---
title: "Qwen-Scope — paper by Deng et al. (2026) — summary and ideas"
description: "Sparse autoencoders (SAEs) break LLM hidden states into a small number of interpretable directions. 77 references, cited by 3 papers here."
authors: ["Boyi Deng", "Xu Wang", "Yaoning Wang", "Yu Wan", "Yubo Ma", "Baosong Yang", "Haoran Wei", "Jialong Tang", "Huan Lin", "Ruize Gao", "Tianhao Li", "Qian Cao", "Xuancheng Ren", "Xiaodong Deng", "An Yang", "Fei Huang", "Dayiheng Liu", "Jingren Zhou"]
year: 2026
venue: "arXiv:2605.11887"
source: "https://arxiv.org/abs/2605.11887"
doi: "10.48550/arxiv.2605.11887"
arxiv: "2605.11887"
licence: "arXiv — non-exclusive licence to distribute"
licence_url: "http://arxiv.org/licenses/nonexclusive-distrib/1.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/deng-2026-qwen-scope-turning-sparse/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/deng-2026-qwen-scope-turning-sparse/)

[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/6/) / Qwen-Scope: Turning Sparse Features into Development Tools for Large Language Models

# Qwen-Scope: Turning Sparse Features into Development Tools for Large Language Models

**Authors:** [Boyi Deng](https://md.research.elamur.ai/free/en/authors/boyi-deng/), [Xu Wang](https://md.research.elamur.ai/free/en/authors/xu-wang/), Yaoning Wang, [Yu Wan](https://md.research.elamur.ai/free/en/authors/yu-wan/), [Yubo Ma](https://md.research.elamur.ai/free/en/authors/yubo-ma/), [Baosong Yang](https://md.research.elamur.ai/free/en/authors/baosong-yang/), [Haoran Wei](https://md.research.elamur.ai/free/en/authors/haoran-wei/), [Jialong Tang](https://md.research.elamur.ai/free/en/authors/jialong-tang/), [Huan Lin](https://md.research.elamur.ai/free/en/authors/huan-lin/), Ruize Gao, [Tianhao Li](https://md.research.elamur.ai/free/en/authors/tianhao-li/), Qian Cao, [Xuancheng Ren](https://md.research.elamur.ai/free/en/authors/xuancheng-ren/), Xiaodong Deng, [An Yang](https://md.research.elamur.ai/free/en/authors/an-yang/), [Fei Huang](https://md.research.elamur.ai/free/en/authors/fei-huang/), [Dayiheng Liu](https://md.research.elamur.ai/free/en/authors/dayiheng-liu/) and [Jingren Zhou](https://md.research.elamur.ai/free/en/authors/jingren-zhou/) · **Year:** 2026 · **Venue:** arXiv:2605.11887 · **Licence:** [arXiv — non-exclusive licence to distribute](http://arxiv.org/licenses/nonexclusive-distrib/1.0/)

**Links:** [Publisher's page](https://arxiv.org/abs/2605.11887) · [PDF](https://arxiv.org/pdf/2605.11887) · [DOI 10.48550/arxiv.2605.11887](https://doi.org/10.48550/arxiv.2605.11887) · [arXiv 2605.11887](https://arxiv.org/abs/2605.11887) · [OpenAlex](https://openalex.org/W7161063287)

**Citations:** 0 citations (OpenAlex, 2026-10-14) · [3](https://md.research.elamur.ai/free/en/papers/deng-2026-qwen-scope-turning-sparse/cited-by/) in our library, of the 149,439 papers whose bibliographies we read

## Summary
Sparse autoencoders (SAEs) break LLM hidden states into a small number of interpretable directions. In practice, though, they have mostly been used to inspect and label features after the fact, not to build or fix models. The Qwen Team addresses this with Qwen-Scope, an openly released collection of SAEs trained on every layer of seven Qwen3 and Qwen3.5 backbones. The backbones include both dense and mixture-of-experts designs, and all SAEs come from a single shared training pipeline.

The authors then use these SAEs as working tools in four areas:
- **Steering:** adjusting output language, concepts, and preferences at inference time without changing weights.
- **Evaluation:** using feature coverage to measure redundancy within and overlap between benchmarks.
- **Data workflows:** classifying multilingual toxicity and generating safety training data guided by features.
- **Post-training:** adding SAE-based signals to SFT and RL.

The main takeaway is that one shared feature vocabulary can support diagnosis, control, evaluation, and training. For example, it helps reduce code-switching and endless repetition. This makes interpretability a reusable part of the development pipeline rather than only a research lens.

## Key points
- **Release scope:** 14 SAE groups covering 7 backbones, including Qwen3-1.7B/8B, Qwen3.5-2B/9B/27B, and the MoE models Qwen3-30B-A3B and Qwen3.5-35B-A3B. The SAEs are published on Hugging Face and ModelScope.
- **SAE configuration:** All SAEs are trained on the residual stream with a top-k sparsity of L0 = 50 or 100. Widths range from 32K to 128K, with expansion factors of 16, or 64 for the wider MoE variants. Only Qwen3.5-27B uses an instruct model; the others use base models.
- **Toxicity classification:** A rule-based classifier built on selected toxic features was tested across languages. Features found on English data were shown to transfer to other languages.
- **SFT against code-switching:** An auxiliary loss suppresses activations of language-specific features, which reduces unwanted switching between languages.
- **RL against repetition:** Steering repetition-related features produces rare negative rollouts. These cut the repeat ratio much more than vanilla RL, while effects on downstream benchmarks were mixed and depended on the task.

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

📚 Shelves (4): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 18,614 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 17,965 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 17,450 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 17,296 works

🧭 Topics (1): [🧠 Language models & pretraining](https://md.research.elamur.ai/free/en/topics/language-models/) · 3,894 works

## Cite

Cite as: Boyi Deng, Xu Wang, Yaoning Wang, Yu Wan, Yubo Ma, Baosong Yang, Haoran Wei, Jialong Tang, Huan Lin, Ruize Gao, Tianhao Li, Qian Cao, Xuancheng Ren, Xiaodong Deng, An Yang, Fei Huang, Dayiheng Liu and Jingren Zhou. 2026. Qwen-Scope: Turning Sparse Features into Development Tools for Large Language Models. arXiv:2605.11887.

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

```bibtex
@misc{deng2026qwen,
  title = {{Qwen-Scope: Turning Sparse Features into Development Tools for Large Language Models}},
  author = {Boyi Deng, Xu Wang, Yaoning Wang, Yu Wan, Yubo Ma, Baosong Yang, Haoran Wei, Jialong Tang, Huan Lin, Ruize Gao, Tianhao Li, Qian Cao, Xuancheng Ren, Xiaodong Deng, An Yang, Fei Huang, Dayiheng Liu and Jingren Zhou},
  year = {2026},
  doi = {10.48550/arxiv.2605.11887},
  eprint = {2605.11887},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2605.11887}
}
```

- [← previous QIAS 2026: Overview of the Shared Task on Islamic Inheritance Reasoning](https://md.research.elamur.ai/free/en/papers/bouchekif-2026-qias-2026-overview-shared-task/)
- [next → RAG over Tables: Hierarchical Memory Index, Multi-Stage Retrieval, and Benchmarking](https://md.research.elamur.ai/free/en/papers/zou-2026-rag-over-tables-hierarchical-memory/)
- [↑ back to papers of 2026](https://md.research.elamur.ai/free/en/papers/2026/page/6/)

## References

77 references, 39 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Anthropic. Introducing claude sonnet 4.6. Technical report, Anthropic, AI, 2026. URL [https://www.anth](https://www.anth) ropic.com/news/claude-sonnet-4-6. [↗](https://www.anth)
2. Dana Arad, Aaron Mueller, and Yonatan Belinkov. [Saes are good for steering – if you select the right features](https://md.research.elamur.ai/free/en/papers/arad-2025-saes-are-good-steering-if/). In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, pp. 10252–10270. Association for Computational Linguistics, 2025. doi: 10.18653/v1/2025.emnlp-main…
3. Andy Arditi, Oscar Obeso, Aaquib Syed, Daniel Paleka, Nina Panickssery, Wes Gurnee, and Neel Nanda. [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/). In Amir Globersons, Lester Mackey, Danielle Belgrave, Angela Fan, Ulrich Paquet, Jakub M. Tomczak, and Cheng Zhang (eds.), Advances in Neu…
4. Jacob Austin, Augustus Odena, Maxwell Nye, Maarten Bosma, Henryk Michalewski, David Dohan, Ellen Jiang, Carrie Cai, Michael Terry, Quoc Le, et al. [Program synthesis with large language models](https://md.research.elamur.ai/free/en/papers/austin-2021-program-synthesis-large-language-models/). arXiv preprint arXiv:2108.07732, 2021.
5. Reza Bayat, Ali Rahimi-Kalahroudi, Mohammad Pezeshki, Sarath Chandar, and Pascal Vincent. Steering large language model activations in sparse spaces. arXiv preprint arXiv:2503.00177, 2025. — [arXiv:2503.00177](https://arxiv.org/abs/2503.00177) [find](https://scholar.google.com/scholar?q=Reza+Bayat%2C+Ali+Rahimi%2DKalahroudi%2C+Mohammad+Pezeshki%2C+Sarath+Chandar%2C+and+Pascal+Vincent%2E+Steering+large+language+model+activations+in+sparse+spaces%2E+arXiv+preprint+arXiv%3A2503%2E00177%2C+2025%2E+%E2%80%94+%5BarXiv%3A25)
6. Leonard Bereska and Stratis Gavves. Mechanistic interpretability for AI safety - a review. Transactions on Machine Learning Research, 2024. [ISSN 2835-8856](https://md.research.elamur.ai/free/en/papers/bereska-2024-mechanistic-interpretability-ai-safety-review/). URL [https://openreview.net/forum?id=ePUVetPKu6](https://openreview.net/forum?id=ePUVetPKu6). Survey Certification, Expert Certification. [↗](https://openreview.net/forum?id=ePUVetPKu6)
7. Paul C. Bogdan, Uzay Macar, Neel Nanda, and Arthur Conmy. Thought anchors: Which llm reasoning steps matter?, 2025. URL https://arxiv.org/abs/2506.19143. [↗](https://arxiv.org/abs/2506.19143)
8. Trenton Bricken, Adly Templeton, Joshua Batson, Brian Chen, Adam Jermyn, Tom Conerly, Nick Turner, Cem Anil, Carson Denison, Amanda Askell, Robert Lasenby, Yifan Wu, Shauna Kravec, Nicholas Schiefer, Tim Maxwell, Nicholas Joseph, Zac Hatfield-Dodds, Alex Tamkin, Karina Nguyen, Brayden McLean, Josiah… [find](https://scholar.google.com/scholar?q=Trenton+Bricken%2C+Adly+Templeton%2C+Joshua+Batson%2C+Brian+Chen%2C+Adam+Jermyn%2C+Tom+Conerly%2C+Nick+Turner%2C+Cem+Anil%2C+Carson+Denison%2C+Amanda+Askell%2C+Robert+Lasenby%2C+Yifan+Wu%2C+Shauna+Kravec%2C+Nicholas+Schiefer%2C+)
9. Helena Casademunt, Caden Juang, Adam Karvonen, Samuel Marks, Senthooran Rajamanoharan, and Neel Nanda. Steering out-of-distribution generalization with concept ablation fine-tuning, 2025. URL https://arxiv.org/abs/2507.16795. — [arXiv:2507.16795](https://arxiv.org/abs/2507.16795) [↗](https://arxiv.org/abs/2507.16795)
10. Federico Cassano, John Gouwar, Daniel Nguyen, Sydney Nguyen, Luna Phipps-Costin, Donald Pinckney, Ming-Ho Yee, Yangtian Zi, Carolyn Jane Anderson, Molly Q Feldman, et al. [Multipl-e: A scalable and extensible approach to benchmarking neural code generation](https://md.research.elamur.ai/free/en/papers/cassano-2022-multipl-e-scalable-extensible-approach/). arXiv preprint arXiv:2208.08227, 2022.
11. Wenhu Chen, Ming Yin, Max Ku, Pan Lu, Yixin Wan, Xueguang Ma, Jianyu Xu, Xinyi Wang, and Tony Xia. [Theoremqa: A theorem-driven question answering dataset](https://md.research.elamur.ai/free/en/papers/chen-2023-theoremqa-theorem-driven-question-answering/). In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pp. 7889–7901, 2023.
12. Wentong Chen, Yankai Lin, ZhenHao Zhou, HongYun Huang, Yantao Jia, Zhao Cao, and Ji-Rong Wen. [Icleval: evaluating in-context learning ability of large language models](https://md.research.elamur.ai/free/en/papers/chen-2025-icleval-evaluating-context-learning-ability/). In Proceedings of the 31st International Conference on Computational Linguistics, pp. 10398–10422, 2025.
13. Zachary Coalson, Juhan Bae, Nicholas Carlini, and Sanghyun Hong. If-guide: Influence function-guided detoxification of llms, 2025. URL https://arxiv.org/abs/2506.01790.2 Boyi Deng, Xu Wang, and Yaoning Wang are in charge of experimental designs and empirical evidence. Yu Wan is the project leader. Y… — [arXiv:2506.01790](https://arxiv.org/abs/2506.01790) [↗](https://arxiv.org/abs/2506.01790.2)
14. Karl Cobbe, Vineet Kosaraju, Mohammad Bavarian, Mark Chen, Heewoo Jun, Lukasz Kaiser, Matthias Plappert, Jerry Tworek, Jacob Hilton, Reiichiro Nakano, et al. [Training verifiers to solve math word problems](https://md.research.elamur.ai/free/en/papers/cobbe-2021-training-verifiers-solve-math-word/). arXiv preprint arXiv:2110.14168, 2021.
15. Gheorghe Comanici, Eric Bieber, Mike Schaekermann, Ice Pasupat, Noveen Sachdeva, Inderjit Dhillon, Marcel Blistein, Ori Ram, Dan Zhang, Evan Rosen, Luke Marris, Sam Petulla, Colin Gaffney, Asaf Aharoni, Nathan Lintz, Tiago Cardal Pais, Henrik Jacobsson, Idan Szpektor, Nan-Jiang Jiang, Krishna Harida… — [Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities](https://md.research.elamur.ai/free/en/papers/comanici-2025-gemini-2-5-pushing-frontier/)
16. Hoagy Cunningham, Aidan Ewart, Logan Riggs, Robert Huben, and Lee Sharkey. Sparse autoencoders find highly interpretable features in language models. arXiv preprint arXiv:2309.08600, 2023. [find](https://scholar.google.com/scholar?q=Hoagy+Cunningham%2C+Aidan+Ewart%2C+Logan+Riggs%2C+Robert+Huben%2C+and+Lee+Sharkey%2E+Sparse+autoencoders+find+highly+interpretable+features+in+language+models%2E+arXiv+preprint+arXiv%3A2309%2E08600%2C+2023%2E)
17. Daryna Dementieva, Daniil Moskovskiy, Nikolay Babakov, Abinew Ali Ayele, Naquee Rizwan, Frolian Schneider, Xintog Wang, Seid Muhie Yimam, Dmitry Ustalov, Elisei Stakovskii, Alisa Smirnova, Ashraf Elnagar, Animesh Mukherjee, and Alexander Panchenko. Overview of the multilingual text detoxification ta… [find](https://scholar.google.com/scholar?q=Daryna+Dementieva%2C+Daniil+Moskovskiy%2C+Nikolay+Babakov%2C+Abinew+Ali+Ayele%2C+Naquee+Rizwan%2C+Frolian+Schneider%2C+Xintog+Wang%2C+Seid+Muhie+Yimam%2C+Dmitry+Ustalov%2C+Elisei+Stakovskii%2C+Alisa+Smirnova%2C+Ashraf+Elna)
18. Boyi Deng, Yu Wan, Baosong Yang, Yidan Zhang, and Fuli Feng. [Unveiling language-specific features in large language models via sparse autoencoders](https://md.research.elamur.ai/free/en/papers/deng-2025-unveiling-language-specific-features-large/). In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pp. 4563–4608, Vienna, Austria, July…
19. Boyi Deng, Yu Wan, Baosong Yang, Fei Huang, Wenjie Wang, and Fuli Feng. SASFT: Sparse autoencoder-guided supervised finetuning to mitigate unexpected code-switching in LLMs. In The Fourteenth International Conference on Learning Representations, 2026. URL [https://openreview.net/forum?id=BQ](https://openreview.net/forum?id=BQ) OFU9qO5j. [↗](https://openreview.net/forum?id=BQ)
20. Xinrun Du, Yifan Yao, Kaijing Ma, Bingli Wang, Tianyu Zheng, King Zhu, Minghao Liu, Yiming Liang, Xiaolong Jin, Zhenlin Wei, et al. [Supergpqa: Scaling llm evaluation across 285 graduate disciplines](https://md.research.elamur.ai/free/en/papers/team-2025-supergpqa-scaling-llm-evaluation-across/). arXiv preprint arXiv:2502.14739, 2025.
21. Jacob Dunefsky, Philippe Chlenski, and Neel Nanda. Transcoders find interpretable llm feature circuits. In A. Globerson, L. Mackey, D. Belgrave, A. Fan, U. Paquet, J. Tomczak, and C. Zhang (eds.), Advances in Neural Information Processing Systems, volume 37, pp. 24375–24410. Curran Associates, Inc.,… — [doi:10.52202/079017-0768](https://doi.org/10.52202/079017-0768) [find](https://scholar.google.com/scholar?q=Jacob+Dunefsky%2C+Philippe+Chlenski%2C+and+Neel+Nanda%2E+Transcoders+find+interpretable+llm+feature+circuits%2E+In+A%2E+Globerson%2C+L%2E+Mackey%2C+D%2E+Belgrave%2C+A%2E+Fan%2C+U%2E+Paquet%2C+J%2E+Tomczak%2C+and+C%2E+Zhang+%28eds%2E%29%2C+Adv)
22. Nelson Elhage, Tristan Hume, Catherine Olsson, Nicholas Schiefer, Tom Henighan, Shauna Kravec, Zac Hatfield-Dodds, Robert Lasenby, Dawn Drain, Carol Chen, Roger Grosse, Sam McCandlish, Jared Kaplan, Dario Amodei, Martin Wattenberg, and Christopher Olah. Toy models of superposition. Transformer Circu… [find](https://scholar.google.com/scholar?q=Nelson+Elhage%2C+Tristan+Hume%2C+Catherine+Olsson%2C+Nicholas+Schiefer%2C+Tom+Henighan%2C+Shauna+Kravec%2C+Zac+Hatfield%2DDodds%2C+Robert+Lasenby%2C+Dawn+Drain%2C+Carol+Chen%2C+Roger+Grosse%2C+Sam+McCandlish%2C+Jared+Kaplan%2C+D)
23. Yi Fang, Wenjie Wang, Mingfeng Xue, Boyi Deng, Fengli Xu, Dayiheng Liu, and Fuli Feng. [Controllable llm reasoning via sparse autoencoder-based steering](https://md.research.elamur.ai/free/en/papers/fang-2026-controllable-llm-reasoning-sparse-autoencoder/). arXiv preprint arXiv:2601.03595, 2026.
24. Eoin Farrell, Yeu-Tong Lau, and Arthur Conmy. Applying sparse autoencoders to unlearn knowledge in language models, 2024. URL https://arxiv.org/abs/2410.19278. — [arXiv:2410.19278](https://arxiv.org/abs/2410.19278) [↗](https://arxiv.org/abs/2410.19278)
25. Javier Ferrando, Oscar Balcells Obeso, Senthooran Rajamanoharan, and Neel Nanda. Do I know this entity? knowledge awareness and hallucinations in language models. In The Thirteenth International Conference on Learning Representations, ICLR 2025, Singapore, April 24-28, 2025. [find](https://scholar.google.com/scholar?q=Javier+Ferrando%2C+Oscar+Balcells+Obeso%2C+Senthooran+Rajamanoharan%2C+and+Neel+Nanda%2E+Do+I+know+this+entity%3F+knowledge+awareness+and+hallucinations+in+language+models%2E+In+The+Thirteenth+International+Confe)
26. OpenReview.net, 2025. URL [https://openreview.net/forum?id=WCRQFlji2q](https://openreview.net/forum?id=WCRQFlji2q). [↗](https://openreview.net/forum?id=WCRQFlji2q)
27. Leo Gao, Tom Dupré la Tour, Henk Tillman, Gabriel Goh, Rajan Troll, Alec Radford, Ilya Sutskever, Jan Leike, and Jeffrey Wu. Scaling and evaluating sparse autoencoders, 2024. URL [https://arxiv.org/ab](https://arxiv.org/ab) s/2406.04093. [↗](https://arxiv.org/ab)
28. Aryo Pradipta Gema, Joshua Ong Jun Leang, Giwon Hong, Alessio Devoto, Alberto Carlo Maria Mancino, Rohit Saxena, Xuanli He, Yu Zhao, Xiaotang Du, Mohammad Reza Ghasemi Madani, et al. Are we done with mmlu? In Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Associatio… [find](https://scholar.google.com/scholar?q=Aryo+Pradipta+Gema%2C+Joshua+Ong+Jun+Leang%2C+Giwon+Hong%2C+Alessio+Devoto%2C+Alberto+Carlo+Maria+Mancino%2C+Rohit+Saxena%2C+Xuanli+He%2C+Yu+Zhao%2C+Xiaotang+Du%2C+Mohammad+Reza+Ghasemi+Madani%2C+et+al%2E+Are+we+done+with+)
29. Gemini Team, Google. [Gemini: A family of highly capable multimodal models](https://md.research.elamur.ai/free/en/papers/gemini-team-2023-gemini-family-highly-capable-multimodal/). arXiv preprint arXiv:2312.11805, 2023. URL https://arxiv.org/abs/2312.11805. [↗](https://arxiv.org/abs/2312.11805)
30. Nicholas Goldowsky-Dill, Bilal Chughtai, Stefan Heimersheim, and Marius Hobbhahn. Detecting strategic deception using linear probes, 2025. URL https://arxiv.org/abs/2502.03407. — [arXiv:2502.03407](https://arxiv.org/abs/2502.03407) [↗](https://arxiv.org/abs/2502.03407)
31. Daya Guo, Dejian Yang, Haowei Zhang, Junxiao Song, Peiyi Wang, Qihao Zhu, Runxin Xu, Ruoyu Zhang, Shirong Ma, Xiao Bi, Xiaokang Zhang, Xingkai Yu, Yu Wu, Z. F. Wu, Zhibin Gou, Zhihong Shao, Zhuoshu Li, Ziyi Gao, Aixin Liu, Bing Xue, Bingxuan Wang, Bochao Wu, Bei Feng, Chengda Lu, Chenggang Zhao, Che… — [doi:10.1038/s41586-](https://doi.org/10.1038/s41586-) [find](https://scholar.google.com/scholar?q=Daya+Guo%2C+Dejian+Yang%2C+Haowei+Zhang%2C+Junxiao+Song%2C+Peiyi+Wang%2C+Qihao+Zhu%2C+Runxin+Xu%2C+Ruoyu+Zhang%2C+Shirong+Ma%2C+Xiao+Bi%2C+Xiaokang+Zhang%2C+Xingkai+Yu%2C+Yu+Wu%2C+Z%2E+F%2E+Wu%2C+Zhibin+Gou%2C+Zhihong+Shao%2C+Zhuoshu+Li)
32. Zhengfu He, Wentao Shu, Xuyang Ge, Lingjie Chen, Junxuan Wang, Yunhua Zhou, Frances Liu, Qipeng Guo, Xuanjing Huang, Zuxuan Wu, Yu-Gang Jiang, and Xipeng Qiu. [Llama scope: Extracting millions of features from llama-3.1-8b with sparse autoencoders](https://md.research.elamur.ai/free/en/papers/he-2024-llama-scope-extracting-millions-features/), 2024. URL https://arxiv.org/abs/2410.205 26. [↗](https://arxiv.org/abs/2410.205)
33. Zirui He, Haiyan Zhao, Yiran Qiao, Fan Yang, Ali Payani, Jing Ma, and Mengnan Du. Saif: A sparse autoencoder framework for interpreting and steering instruction following of language models. arXiv preprint arXiv:2502.11356, 2025. — [arXiv:2502.11356](https://arxiv.org/abs/2502.11356) [find](https://scholar.google.com/scholar?q=Zirui+He%2C+Haiyan+Zhao%2C+Yiran+Qiao%2C+Fan+Yang%2C+Ali+Payani%2C+Jing+Ma%2C+and+Mengnan+Du%2E+Saif%3A+A+sparse+autoencoder+framework+for+interpreting+and+steering+instruction+following+of+language+models%2E+arXiv+pre)
34. Dan Hendrycks, Collin Burns, Steven Basart, Andy Zou, Mantas Mazeika, Dawn Song, and Jacob Steinhardt. [Measuring massive multitask language understanding](https://md.research.elamur.ai/free/en/papers/hendrycks-2020-measuring-massive-multitask-language-understanding/). arXiv preprint arXiv:2009.03300, 2020.
35. Dan Hendrycks, Collin Burns, Saurav Kadavath, Akul Arora, Steven Basart, Eric Tang, Dawn Song, and Jacob Steinhardt. [Measuring mathematical problem solving with the math dataset](https://md.research.elamur.ai/free/en/papers/hendrycks-2021-measuring-mathematical-problem-solving-math/). NeurIPS, 2021.
36. Edward J. Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. [Lora: Low-rank adaptation of large language models](https://md.research.elamur.ai/free/en/papers/hu-2021-lora-low-rank-adaptation-large/), 2021. URL https://arxiv.org/abs/2106.09685. [↗](https://arxiv.org/abs/2106.09685)
37. Yuzhen Huang, Yuzhuo Bai, Zhihao Zhu, Junlei Zhang, Jinghan Zhang, Tangjun Su, Junteng Liu, Chuancheng Lv, Yikai Zhang, Jiayi Lei, Yao Fu, Maosong Sun, and Junxian He. [C-eval: A multi-level multi-discipline chinese evaluation suite for foundation models](https://md.research.elamur.ai/free/en/papers/huang-2023-c-eval-multi-level-multi/). In Advances in Neural Information Processing…
38. Liwei Jiang, Kavel Rao, Seungju Han, Allyson Ettinger, Faeze Brahman, Sachin Kumar, Niloofar Mireshghallah, Ximing Lu, Maarten Sap, Yejin Choi, et al. [Wildteaming at scale: From in-the-wild jailbreaks to (adversarially) safer language models](https://md.research.elamur.ai/free/en/papers/jiang-2024-wildteaming-scale-wild-jailbreaks-adversarially/). Advances in Neural Information Processing Systems, 37:470…
39. Haonan Li, Yixuan Zhang, Fajri Koto, Yifei Yang, Hai Zhao, Yeyun Gong, Nan Duan, and Timothy Baldwin. [Cmmlu: Measuring massive multitask language understanding in chinese](https://md.research.elamur.ai/free/en/papers/li-2023-cmmlu-measuring-massive-multitask-language/), 2023.
40. Zhe Li, Wei Zhao, Yige Li, and Jun Sun. Do influence functions work on large language models?, 2024. URL https://arxiv.org/abs/2409.19998. [↗](https://arxiv.org/abs/2409.19998)
41. Zhongzhi Li, Xuansheng Wu, Yijiang Li, Lijie Hu, and Ninghao Liu. Less is enough: Synthesizing diverse data in feature space of llms. arXiv preprint arXiv:2602.10388, 2026. — [arXiv:2602.10388](https://arxiv.org/abs/2602.10388) [find](https://scholar.google.com/scholar?q=Zhongzhi+Li%2C+Xuansheng+Wu%2C+Yijiang+Li%2C+Lijie+Hu%2C+and+Ninghao+Liu%2E+Less+is+enough%3A+Synthesizing+diverse+data+in+feature+space+of+llms%2E+arXiv+preprint+arXiv%3A2602%2E10388%2C+2026%2E+%E2%80%94+%5BarXiv%3A2602%2E10388%5D%28https%3A)
42. Zihao Li, Xu Wang, Yuzhe YANG, Ziyu Yao, Haoyi Xiong, and Mengnan Du. [Feature extraction and steering for enhanced chain-of-thought reasoning in language models](https://md.research.elamur.ai/free/en/papers/li-2025-feature-extraction-steering-enhanced-chain/). In The 2025 Conference on Empirical Methods in Natural Language Processing, 2025. URL [https://openreview.net/forum?id=u3n1](https://openreview.net/forum?id=u3n1) CDGrOA. [↗](https://openreview.net/forum?id=u3n1)
43. Tom Lieberum, Senthooran Rajamanoharan, Arthur Conmy, Lewis Smith, Nicolas Sonnerat, Vikrant Varma, Janos Kramar, Anca Dragan, Rohin Shah, and Neel Nanda. [Gemma scope: Open sparse autoencoders everywhere all at once on gemma 2](https://md.research.elamur.ai/free/en/papers/lieberum-2024-gemma-scope-open-sparse-autoencoders/). In Yonatan Belinkov, Najoung Kim, Jaap Jumelet, Hosein Mohebbi, Aaron Mu…
44. Stephanie Lin, Jacob Hilton, and Owain Evans. [Truthfulqa: Measuring how models mimic human falsehoods](https://md.research.elamur.ai/free/en/papers/lin-2022-truthfulqa-measuring-how-models-mimic/). In Proceedings of the 60th annual meeting of the association for computational linguistics (volume 1: long papers), pp. 3214–3252, 2022.
45. Jack Lindsey, Wes Gurnee, Emmanuel Ameisen, Brian Chen, Adam Pearce, Nicholas L. Turner, Craig Citro, David Abrahams, Shan Carter, Basil Hosmer, Jonathan Marcus, Michael Sklar, Adly Templeton, Trenton Bricken, Callum McDougall, Hoagy Cunningham, Thomas Henighan, Adam Jermyn, Andy Jones, Andrew Persi… [find](https://scholar.google.com/scholar?q=Jack+Lindsey%2C+Wes+Gurnee%2C+Emmanuel+Ameisen%2C+Brian+Chen%2C+Adam+Pearce%2C+Nicholas+L%2E+Turner%2C+Craig+Citro%2C+David+Abrahams%2C+Shan+Carter%2C+Basil+Hosmer%2C+Jonathan+Marcus%2C+Michael+Sklar%2C+Adly+Templeton%2C+Trenton)
46. Jiawei Liu, Chunqiu Steven Xia, Yuyao Wang, and Lingming Zhang. [Is your code generated by chatGPT really correct? rigorous evaluation of large language models for code generation](https://md.research.elamur.ai/free/en/papers/liu-2023-your-code-generated-chatgpt-really/). In Thirty-seventh Conference on Neural Information Processing Systems, 2023. URL [https://openreview.net/forum?id=1q](https://openreview.net/forum?id=1q) vx61… [↗](https://openreview.net/forum?id=1q)
47. George Ma, Zhongyuan Liang, Irene Y. Chen, and Somayeh Sojoudi. Falsifying sparse autoencoder reasoning features in language models, 2026. URL https://arxiv.org/abs/2601.05679. [↗](https://arxiv.org/abs/2601.05679)
48. Kaijing Ma, Xinrun Du, Yunran Wang, Haoran Zhang, Zhoufutu Wen, Xingwei Qu, Jian Yang, Jiaheng Liu, Minghao Liu, Xiang Yue, et al. Kor-bench: Benchmarking language models on knowledge-orthogonal reasoning tasks. arXiv preprint arXiv:2410.06526, 2024. — [arXiv:2410.06526](https://arxiv.org/abs/2410.06526) [find](https://scholar.google.com/scholar?q=Kaijing+Ma%2C+Xinrun+Du%2C+Yunran+Wang%2C+Haoran+Zhang%2C+Zhoufutu+Wen%2C+Xingwei+Qu%2C+Jian+Yang%2C+Jiaheng+Liu%2C+Minghao+Liu%2C+Xiang+Yue%2C+et+al%2E+Kor%2Dbench%3A+Benchmarking+language+models+on+knowledge%2Dorthogonal+reaso)
49. Uzay Macar, Paul C. Bogdan, Senthooran Rajamanoharan, and Neel Nanda. Thought branches: Interpreting llm reasoning requires resampling, 2026. URL https://arxiv.org/abs/2510.27484. — [arXiv:2510.27484](https://arxiv.org/abs/2510.27484) [↗](https://arxiv.org/abs/2510.27484)
50. Samuel Marks, Adam Karvonen, and Aaron Mueller. dictionary learning, 2024. URL [https://github.com/saprmarks/dictionary](https://github.com/saprmarks/dictionary) learning. [↗](https://github.com/saprmarks/dictionary)
51. Samuel Marks, Johannes Treutlein, Trenton Bricken, Jack Lindsey, Jonathan Marcus, Siddharth Mishra-Sharma, Daniel Ziegler, Emmanuel Ameisen, Joshua Batson, Tim Belonax, Samuel R. Bowman, Shan Carter, Brian Chen, Hoagy Cunningham, Carson Denison, Florian Dietz, Satvik Golechha, Akbir Khan, Jan Kirchn… [find](https://scholar.google.com/scholar?q=Samuel+Marks%2C+Johannes+Treutlein%2C+Trenton+Bricken%2C+Jack+Lindsey%2C+Jonathan+Marcus%2C+Siddharth+Mishra%2DSharma%2C+Daniel+Ziegler%2C+Emmanuel+Ameisen%2C+Joshua+Batson%2C+Tim+Belonax%2C+Samuel+R%2E+Bowman%2C+Shan+Carter%2C+)
52. Julian Minder, Clément Dumas, Stewart Slocum, Helena Casademunt, Cameron Holmes, Robert West, and Neel Nanda. Narrow finetuning leaves clearly readable traces in activation differences. In The Fourteenth International Conference on Learning Representations, 2026. URL [https://openreview.net/for](https://openreview.net/for) um?i… [↗](https://openreview.net/for)
53. Neel Nanda, Andrew Lee, and Martin Wattenberg. [Emergent linear representations in world models of self-supervised sequence models](https://md.research.elamur.ai/free/en/papers/nanda-2023-emergent-linear-representations-world-models/). In Yonatan Belinkov, Sophie Hao, Jaap Jumelet, Najoung Kim, Arya McCarthy, and Hosein Mohebbi (eds.), Proceedings of the 6th BlackboxNLP Workshop: Analyzing and Interpre…
54. Usman Naseem. Mechanistic interpretability for large language model alignment: Progress, challenges, and future directions. arXiv preprint arXiv:2602.11180, 2026. — [arXiv:2602.11180](https://arxiv.org/abs/2602.11180) [find](https://scholar.google.com/scholar?q=Usman+Naseem%2E+Mechanistic+interpretability+for+large+language+model+alignment%3A+Progress%2C+challenges%2C+and+future+directions%2E+arXiv+preprint+arXiv%3A2602%2E11180%2C+2026%2E+%E2%80%94+%5BarXiv%3A2602%2E11180%5D%28https%3A%2F%2Farxiv%2Eor)
55. OpenAI. GPT4 technical report. arXiv preprint arXiv:2303.[08774, 2023](https://md.research.elamur.ai/free/en/papers/openai-2023-gpt-4-technical-report/).
56. OpenAI. Multilingual massive multitask language understanding, 2024. URL [https://huggingface.co/datasets/openai/MMMLU](https://huggingface.co/datasets/openai/MMMLU). [↗](https://huggingface.co/datasets/openai/MMMLU)
57. Kiho Park, Yo Joong Choe, and Victor Veitch. [The linear representation hypothesis and the geometry of large language models](https://md.research.elamur.ai/free/en/papers/park-2023-linear-representation-hypothesis-geometry-large/), 2023.
58. Avi Parrack, Carlo Leonardo Attubato, and Stefan Heimersheim. Benchmarking deception probes via black-to-white performance boosts, 2026. URL https://arxiv.org/abs/2507.12691. — [arXiv:2507.12691](https://arxiv.org/abs/2507.12691) [↗](https://arxiv.org/abs/2507.12691)
59. Gonçalo Paulo, Alex Mallen, Caden Juang, and Nora Belrose. Automatically interpreting millions of features in large language models. In Aarti Singh, Maryam Fazel, Daniel Hsu, Simon Lacoste-Julien, Felix Berkenkamp, Tegan Maharaj, Kiri Wagstaff, and Jerry Zhu (eds.), Forty-second International Confe… [find](https://scholar.google.com/scholar?q=Gonc%CC%A7alo+Paulo%2C+Alex+Mallen%2C+Caden+Juang%2C+and+Nora+Belrose%2E+Automatically+interpreting+millions+of+features+in+large+language+models%2E+In+Aarti+Singh%2C+Maryam+Fazel%2C+Daniel+Hsu%2C+Simon+Lacoste%2DJulien%2C+Fe)
60. Gonçalo Santos Paulo, Alex Troy Mallen, Caden Juang, and Nora Belrose. Automatically interpreting millions of features in large language models. In Forty-second International Conference on Machine Learning, 2025b. URL [https://openreview.net/forum?id=EemtbhJOXc](https://openreview.net/forum?id=EemtbhJOXc). [↗](https://openreview.net/forum?id=EemtbhJOXc)
61. David Rein, Betty Li Hou, Asa Cooper Stickland, Jackson Petty, Richard Yuanzhe Pang, Julien Dirani, Julian Michael, and Samuel R Bowman. [Gpqa: A graduate-level google-proof q&a benchmark](https://md.research.elamur.ai/free/en/papers/rein-2023-gpqa-graduate-level-google-proof/). arXiv preprint arXiv:2311.12022, 2023.
62. Nina Rimsky, Nick Gabrieli, Julian Schulz, Meg Tong, Evan Hubinger, and Alexander Turner. Steering llama 2 via contrastive activation addition. In Lun-Wei Ku, Andre Martins, and Vivek Srikumar (eds.), Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1:… [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+Steering+llama+2+via+contrastive+activation+addition%2E+In+Lun%2DWei+Ku%2C+Andre+Martins%2C+and+Vivek+Srikumar+%28eds%2E%29%2C)
63. Angelika Romanou, Negar Foroutan, Anna Sotnikova, Zeming Chen, Sree Harsha Nelaturu, Shivalika Singh, Rishabh Maheshwary, Micol Altomare, Mohamed A Haggag, Alfonso Amayuelas, et al. [Include: Evaluating multilingual language understanding with regional knowledge](https://md.research.elamur.ai/free/en/papers/romanou-2024-include-evaluating-multilingual-language-understanding/). arXiv preprint arXiv:2411.19799, 2024…
64. Lee Sharkey, Bilal Chughtai, Joshua Batson, Jack Lindsey, Jeff Wu, Lucius Bushnaq, Nicholas Goldowsky-Dill, Stefan Heimersheim, Alejandro Ortega, Joseph Bloom, et al. Open problems in mechanistic interpretability. arXiv preprint arXiv:2501.16496, 2025. — [arXiv:2501.16496](https://arxiv.org/abs/2501.16496) [find](https://scholar.google.com/scholar?q=Lee+Sharkey%2C+Bilal+Chughtai%2C+Joshua+Batson%2C+Jack+Lindsey%2C+Jeff+Wu%2C+Lucius+Bushnaq%2C+Nicholas+Goldowsky%2DDill%2C+Stefan+Heimersheim%2C+Alejandro+Ortega%2C+Joseph+Bloom%2C+et+al%2E+Open+problems+in+mechanistic+inte)
65. Wei Shi, Sihang Li, Tao Liang, Mingyang Wan, Guojun Ma, Xiang Wang, and Xiangnan He. [Route sparse autoencoder to interpret large language models](https://md.research.elamur.ai/free/en/papers/shi-2025-route-sparse-autoencoder-interpret-large/). In Christos Christodoulopoulos, Tanmoy Chakraborty, Carolyn Rose, and Violet Peng (eds.), Proceedings of the 2025 Conference on Empirical Methods in Natur…
66. Dong Shu, Xuansheng Wu, Haiyan Zhao, Daking Rai, Ziyu Yao, Ninghao Liu, and Mengnan Du. A survey on sparse autoencoders: Interpreting the internal mechanisms of large language models. In Christos Christodoulopoulos, Tanmoy Chakraborty, Carolyn Rose, and Violet Peng (eds.), Findings of the Associatio… [find](https://scholar.google.com/scholar?q=Dong+Shu%2C+Xuansheng+Wu%2C+Haiyan+Zhao%2C+Daking+Rai%2C+Ziyu+Yao%2C+Ninghao+Liu%2C+and+Mengnan+Du%2E+A+survey+on+sparse+autoencoders%3A+Interpreting+the+internal+mechanisms+of+large+language+models%2E+In+Christos+Chri)
67. Aaditya Singh, Adam Fry, Adam Perelman, Adam Tart, Adi Ganesh, Ahmed El-Kishky, Aidan McLaughlin, Aiden Low, AJ Ostrow, Akhila Ananthram, et al. [Openai gpt-5 system card](https://md.research.elamur.ai/free/en/papers/singh-2025-openai-gpt-5-system-card/). arXiv preprint arXiv:2601.03267, 2025.
68. Chandan Singh, Jeevana Priya Inala, Michel Galley, Rich Caruana, and Jianfeng Gao. [Rethinking interpretability in the era of large language models](https://md.research.elamur.ai/free/en/papers/singh-2024-rethinking-interpretability-era-large-language/), 2024. URL https://arxiv.org/abs/2402.01761. [↗](https://arxiv.org/abs/2402.01761)
69. Mirac Suzgun, Nathan Scales, Nathanael Schärli, Sebastian Gehrmann, Yi Tay, Hyung Won Chung, Aakanksha Chowdhery, Quoc Le, Ed Chi, Denny Zhou, et al. [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: ACL 2023, pp.…
70. Rohan Taori, Ishaan Gulrajani, Tianyi Zhang, Yann Dubois, Xuechen Li, Carlos Guestrin, Percy Liang, and Tatsunori B. Hashimoto. Stanford alpaca: An instruction-following llama model. [https://github.com/tatsu-lab/stanford](https://github.com/tatsu-lab/stanford) alpaca, 2023. [↗](https://github.com/tatsu-lab/stanford)
71. Xu Wang, Zihao Li, Benyou Wang, Yan Hu, and Difan Zou. [Model unlearning via sparse autoencoder subspace guided projections](https://md.research.elamur.ai/free/en/papers/wang-2025-model-unlearning-sparse-autoencoder-subspace/). In The 2025 Conference on Empirical Methods in Natural Language Processing, 2025. URL [https://openreview.net/forum?id=neZbAv5JrY](https://openreview.net/forum?id=neZbAv5JrY). [↗](https://openreview.net/forum?id=neZbAv5JrY)
72. Xu Wang, Yan Hu, Benyou Wang, and Difan Zou. Does higher interpretability imply better utility? a pairwise analysis on sparse autoencoders. In The Fourteenth International Conference on Learning Representations, 2026. URL [https://openreview.net/forum?id=Q4ooLNOFeR](https://openreview.net/forum?id=Q4ooLNOFeR). [↗](https://openreview.net/forum?id=Q4ooLNOFeR)
73. Yubo Wang, Xueguang Ma, Ge Zhang, Yuansheng Ni, Abhranil Chandra, Shiguang Guo, Weiming Ren, Aaran Arulraj, Xuan He, Ziyan Jiang, et al. [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/). Advances in Neural Information Processing Systems, 37:95266–95290, 2024.
74. An Yang, Anfeng Li, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chang Gao, Chengen Huang, Chenxu Lv, Chujie Zheng, Dayiheng Liu, Fan Zhou, Fei Huang, Feng Hu, Hao Ge, Haoran Wei, Huan Lin, Jialong Tang, Jian Yang, Jianhong Tu, Jianwei Zhang, Jianxin Yang, Jiaxi Yang, Jing Zhou, Jin… — [Qwen3 Technical Report](https://md.research.elamur.ai/free/en/papers/yang-2025-qwen3-technical-report/)
75. Qiying Yu, Zheng Zhang, Ruofei Zhu, Yufeng Yuan, Xiaochen Zuo, Yu Yue, Weinan Dai, Tiantian Fan, Gaohong Liu, Lingjun Liu, et al. [Dapo: An open-source llm reinforcement learning system at scale](https://md.research.elamur.ai/free/en/papers/yu-2025-dapo-open-source-llm-reinforcement/). arXiv preprint arXiv:2503.14476, 2025.
76. Hengyuan Zhang, Zhihao Zhang, Mingyang Wang, Zunhai Su, Yiwei Wang, Qianli Wang, Shuzhou Yuan, Ercong Nie, Xufeng Duan, Qibo Xue, et al. [Locate, steer, and improve: A practical survey of actionable mechanistic interpretability in large language models](https://md.research.elamur.ai/free/en/papers/zhang-2026-locate-steer-improve-practical-survey/). arXiv preprint arXiv:2601.14004, 2026.
77. Jeffrey Zhou, Tianjian Lu, Swaroop Mishra, Siddhartha Brahma, Sujoy Basu, Yi Luan, Denny Zhou, and Le Hou. [Instruction-following evaluation for large language models](https://md.research.elamur.ai/free/en/papers/zhou-2023-instruction-following-evaluation-large-language/). arXiv preprint arXiv:2311.07911, 2023.

*Updated 2 November 2026*