Elamur Research⌕
☰
home / papers / 2026 / Memory Dial: A Training Framework for Controllable Memorization in Language Models

Memory Dial: A Training Framework for Controllable Memorization in Language Models

Paper · 2026 · Xiangbo Zhang and Ali Emami

Summary

Research on memorization in language models is mostly diagnostic. It inspects models that are already trained, so memorization effects cannot be separated from differences in architecture, data or optimization. The authors propose Memory Dial, a training framework that makes memorization pressure a variable the experimenter sets. The loss is a convex mix of ordinary cross-entropy and a temperature-sharpened term that rewards confident predictions on training sequences. A single coefficient α in [0, 1] controls the weighting. Sweeping α while keeping all other settings fixed yields matched model families that differ only in this pressure. Clean seen/unseen labels come from injecting a small set of benchmark examples into training.

Across six architectures and five benchmarks, accuracy on seen items grows monotonically with α while held-out accuracy stays roughly flat. Larger models respond more strongly, and frequent sequences are recalled more easily than rare ones. The framework offers a controlled setting for studying how memorization and generalization interact. It may also matter for tasks needing exact recall, such as preserving rare patterns in low-resource languages.

Key points

  • Models and benchmarks: the models are GPT-2, DistilGPT2, TinyLLaMA-1B and OPT models (reported as 250M to 27B parameters). The benchmarks are ARC, BoolQ, PIQA, COPA and OpenBookQA. All 30 model–benchmark pairs show positive seen-accuracy slopes, ranging from 0.03 to 0.38.
  • Scale effect: averaged over benchmarks, OPT-27B has a mean slope of 0.206, compared with 0.097 for DistilGPT2.
  • Frequency effect: at α = 0, frequent and rare sequences differ by about 4.4 in suffix NLL. At α = 0.8, rare sequences are still harder to recall (29.7 vs. 27.2). Naturally occurring single-occurrence sequences also show falling NLL as α increases.
  • Temperature and diversity: a sharpening temperature of τ = 0.1 gives the strongest signal without hurting unseen performance. Plain single-temperature cross-entropy fails to reproduce the same trade-off. Output self-similarity (Jaccard) rises from 0.372 to 0.528 as α goes from 0 to 1, meaning outputs become less diverse.
  • Extensions and resources:
    • Multilingual tests on XCOPA (Turkish, Chinese) show the effect transfers.
    • On TruthfulQA, truthfulness improves while ROUGE-L stays stable.
    • Calibration error does not worsen.
    • Code is available at github.com/xiangbo05/MemoryDial_Public.

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

🎓Georgia Institute of TechnologyAtlanta · United States152 works · 27,763 citations🎓Emory UniversityAtlanta · United States42 works · 39,899 citations

 Fields, topics and concepts

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 18,614 worksArtificial Intelligence · 17,965 worksArts and Humanities · 17,450 worksLanguage and Linguistics · 17,296 works

🧭 Topics 1Research subjects and languages the work is about.Language models & pretraining · 3,894 works

Cite

Cite as Xiangbo Zhang and Ali Emami. 2026. Memory Dial: A Training Framework for Controllable Memorization in Language Models. Findings of the Association for Computational Linguistics: ACL 2026.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.findings-acl.179

@inproceedings{zhang2026memory,
  title = {{Memory Dial: A Training Framework for Controllable Memorization in Language Models}},
  author = {Xiangbo Zhang and Ali Emami},
  booktitle = {Findings of the Association for Computational Linguistics: ACL 2026},
  year = {2026},
  doi = {10.18653/v1/2026.findings-acl.179},
  url = {https://doi.org/10.18653/v1/2026.findings-acl.179}
}

References

39 references, 19 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Yonatan Bisk, Rowan Zellers, Ronan Le bras, Jianfeng Gao, and Yejin Choi. 2020. PIQA: Reasoning about Physical Commonsense in Natural Language. AAAI, 34(05):7432–7439.
  2. Nicholas Carlini, Daphne Ippolito, Matthew Jagielski, Katherine Lee, Florian Tramer, and Chiyuan Zhang. 2022. Quantifying Memorization Across Neural Language Models. In The Eleventh International Conference on Learning Representations. find
  3. Nicholas Carlini, Daphne Ippolito, Matthew Jagielski, Katherine Lee, Florian Tramer, and Chiyuan Zhang. 2023. Quantifying memorization across neural language models. Preprint, arXiv:2202.07646. find
  4. Nicholas Carlini, Chang Liu, Úlfar Erlingsson, Jernej Kos, and Dawn Song. 2019. The Secret Sharer: Evaluating and Testing Unintended Memorization in Neural Networks. In 28th USENIX Security Symposium (USENIX Security 19), pages 267–284.
  5. Nicholas Carlini, Florian Tramèr, Eric Wallace, Matthew Jagielski, Ariel Herbert-Voss, Katherine Lee, Adam Roberts, Tom Brown, Dawn Song, Úlfar Erlingsson, Alina Oprea, and Colin Raffel. 2021. Extracting Training Data from Large Language Models. In 30th USENIX Security Symposium (USENIX Security 21)…
  6. Christopher Clark, Kenton Lee, Ming-Wei Chang, Tom Kwiatkowski, Michael Collins, and Kristina Toutanova. 2019. BoolQ: Exploring the Surprising Difficulty of Natural Yes/No Questions. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics:…
  7. Peter Clark, Isaac Cowhey, Oren Etzioni, Tushar Khot, Ashish Sabharwal, Carissa Schoenick, and Oyvind Tafjord. 2018. Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge. Preprint, arXiv:1803.05457.
  8. Yihong Dong, Xue Jiang, Huanyu Liu, Zhi Jin, Bin Gu, Mengfei Yang, and Ge Li. 2024. Generalization or Memorization: Data Contamination and Trustworthy Evaluation for Large Language Models. In Findings of the Association for Computational Linguistics: ACL 2024, pages 12039–12050, Bangkok, Thailand. A… find
  9. Ali Emami, Kaheer Suleman, Adam Trischler, and Jackie Chi Kit Cheung. 2020. An Analysis of Dataset Overlap on Winograd-Style Tasks. In Proceedings of the 28th International Conference on Computational Linguistics, pages 5855–5865, Barcelona, Spain (Online). International Committee on Computational L… find
  10. Mor Geva, Jasmijn Bastings, Katja Filippova, and Amir Globerson. 2023. Dissecting Recall of Factual Associations in Auto-Regressive Language Models. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 12216–12235, Singapore. Association for Computational… find
  11. Andrew Gordon, Zornitsa Kozareva, and Melissa Roemmele. 2012. SemEval-2012 Task 7: Choice of Plausible Alternatives: An Evaluation of Commonsense Causal Reasoning. In *SEM 2012: The First Joint Conference on Lexical and Computational Semantics – Volume 1: Proceedings of the Main Conference and the S…
  12. Valentin Hartmann, Anshuman Suri, Vincent Bindschaedler, David Evans, Shruti Tople, and Robert West. 2023. SoK: Memorization in General-Purpose Large Language Models. Preprint, arXiv:2310.18362. — [arXiv:2310.18362](https://arxiv.org/abs/2310.18362) find
  13. Geoffrey Hinton, Oriol Vinyals, and Jeff Dean. 2015. Distilling the Knowledge in a Neural Network. Preprint, arXiv:1503.02531.
  14. Ari Holtzman, Jan Buys, Li Du, Maxwell Forbes, and Yejin Choi. 2019. The Curious Case of Neural Text Degeneration. In International Conference on Learning Representations.
  15. Yihuai Hong, Meng Cao, Dian Zhou, Lei Yu, and Zhijing Jin. 2025. The Reasoning-Memorization Interplay in Language Models Is Mediated by a Single Direction. In Findings of the Association for Computational Linguistics: ACL 2025, pages 21565–21585, Vienna, Austria. Association for Computational Lingui… find
  16. Ko-Wei Huang, Yi-Fu Fu, Ching-Yu Tsai, Yu-Chieh Tu, Tzu-ling Cheng, Cheng-Yu Lin, Yi-Ting Yang, Heng-Yi Liu, Keng-Te Liao, Da-Cheng Juan, and Shou-De Lin. 2025. Neuron-Level Differentiation of Memorization and Generalization in Large Language Models. In Proceedings of the 2025 Conference on Empirica… find
  17. Katherine Lee, Daphne Ippolito, Andrew Nystrom, Chiyuan Zhang, Douglas Eck, Chris Callison-Burch, and Nicholas Carlini. 2022. Deduplicating Training Data Makes Language Models Better. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers),… find
  18. Kenneth Li, Oam Patel, Fernanda Viégas, Hanspeter Pfister, and Martin Wattenberg. 2023. Inference-Time Intervention: Eliciting Truthful Answers from a Language Model. In Thirty-Seventh Conference on Neural Information Processing Systems.
  19. Ziming Liu, Ouail Kitouni, Niklas Nolte, Eric J. Michaud, Max Tegmark, and Mike Williams. 2022. Towards Understanding Grokking: An Effective Theory of Representation Learning. In Advances in Neural Information Processing Systems. find
  20. Todor Mihaylov, Peter Clark, Tushar Khot, and Ashish Sabharwal. 2018. Can a Suit of Armor Conduct Electricity? A New Dataset for Open Book Question Answering. In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, pages 2381–2391, Brussels, Belgium. Association fo… find
  21. Felix B Mueller, Rebekka Görge, Anna K Bernzen, Janna C Pirk, and Maximilian Poretschkin. 2025. LLMs and Memorization: On Quality and Specificity of Copyright Compliance. In Proceedings of the 2024 AAAI/ACM Conference on AI, Ethics, and Society, pages 984–996. AAAI Press. find
  22. Neel Nanda, Lawrence Chan, Tom Lieberum, Jess Smith, and Jacob Steinhardt. 2022. Progress measures for grokking via mechanistic interpretability. In The Eleventh International Conference on Learning Representations.
  23. Yonatan Oren, Nicole Meister, Niladri S. Chatterji, Faisal Ladhak, and Tatsunori Hashimoto. 2023. Proving Test Set Contamination in Black-Box Language Models. In The Twelfth International Conference on Learning Representations.
  24. Gabriel Pereyra, George Tucker, Jan Chorowski, Łukasz Kaiser, and Geoffrey Hinton. 2017. Regularizing Neural Networks by Penalizing Confident Output Distributions. Preprint, arXiv:1701.06548.
  25. Fabio Petroni, Tim Rocktäschel, Sebastian Riedel, Patrick Lewis, Anton Bakhtin, Yuxiang Wu, and Alexander Miller. 2019. Language Models as Knowledge Bases? In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural…
  26. Le Scao, Sylvain Gugger, and 3 others. 2020. Transformers: State-of-the-Art Natural Language Processing. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing: System Demonstrations, pages 38–45, Online. Association for Computational Linguistics.
  27. Rowan Zellers, Yonatan Bisk, Roy Schwartz, and Yejin Choi. 2018. SWAG: A Large-Scale Adversarial Dataset for Grounded Commonsense Inference. In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, pages 93– 104, Brussels, Belgium. Association for Computational Ling… find
  28. Alethea Power, Yuri Burda, Harri Edwards, Igor Babuschkin, and Vedant Misra. 2022. Grokking: Generalization Beyond Overfitting on Small Algorithmic Datasets. Preprint, arXiv:2201.02177.
  29. Chiyuan Zhang, Daphne Ippolito, Katherine Lee, Matthew Jagielski, Florian Tramèr, and Nicholas Carlini. 2023. Counterfactual Memorization in Neural Language Models. In Thirty-Seventh Conference on Neural Information Processing Systems. find
  30. Alec Radford, Jeff Wu, R. Child, D. Luan, Dario Amodei, and I. Sutskever. 2019. Language Models are Unsupervised Multitask Learners. Technical report, OpenAI. find
  31. Peiyuan Zhang, Guangtao Zeng, Tianduo Wang, and Wei Lu. 2024. TinyLlama: An Open-Source Small Language Model. Preprint, arXiv:2401.02385.
  32. 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, Bangkok, Thaila… find
  33. Susan Zhang, Stephen Roller, Naman Goyal, Mikel Artetxe, Moya Chen, Shuohui Chen, Christopher Dewan, Mona Diab, Xian Li, Xi Victoria Lin, Todor Mihaylov, Myle Ott, Sam Shleifer, Kurt Shuster, Daniel Simig, Punit Singh Koura, Anjali Sridhar, Tianlu Wang, and Luke Zettlemoyer. 2022. OPT: Open Pre-trai…
  34. Weijia Shi, Anirudh Ajith, Mengzhou Xia, Yangsibo Huang, Daogao Liu, Terra Blevins, Danqi Chen, and Luke Zettlemoyer. 2024. Detecting Pretraining Data from Large Language Models. Preprint, arXiv:2310.16789. find
  35. Reza Shokri, Marco Stronati, Congzheng Song, and Vitaly Shmatikov. 2017. find
  36. Membership Inference Attacks Against Machine Learning Models. In 2017 IEEE Symposium on Security and Privacy (SP), pages 3–18.
  37. Christian Szegedy, Vincent Vanhoucke, Sergey Ioffe, Jon Shlens, and Zbigniew Wojna. 2016. Rethinking the Inception Architecture for Computer Vision. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR). find
  38. Kushal Tirumala, Aram H. Markosyan, Luke Zettlemoyer, and Armen Aghajanyan. 2022. Memorization without overfitting: Analyzing the training dynamics of large language models. In Proceedings of the 36th International Conference on Neural Information Processing Systems, NIPS ’22, pages 38274–38290, Red…
  39. Thomas Wolf, Lysandre Debut, Victor Sanh, Julien Chaumond, Clement Delangue, Anthony Moi, Pier-ric Cistac, Tim Rault, Remi Louf, Morgan Funtowicz, Joe Davison, Sam Shleifer, Patrick von Platen, Clara Ma, Yacine Jernite, Julien Plu, Canwen Xu, Teven find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/zhang-2026-memory-dial-training-framework-controllable/