AI Chat Paper
Note: Please note that the following content is generated by AMiner AI. SciOpen does not take any responsibility related to this content.
{{lang === 'zh_CN' ? '文章概述' : 'Summary'}}
{{lang === 'en_US' ? '中' : 'Eng'}}
Chat more with AI
Powered by AMiner. Clicking this button will take you away from SciOpen.
Home Journal of Social Computing Article
PDF (3.3 MB)
Cite
EndNote(RIS) BibTeX
Collect
Submit Manuscript AI Chat Paper
Show Outline
Outline
Show full outline
Hide outline
Outline
Show full outline
Hide outline
Open Access

Generative Models as a Complex Systems Science: How Can We Make Sense of Large Language Model Behavior?

University of Chicago, Chicago, IL 60605, USA
Show Author Information

Abstract

Coaxing out desired behavior from pretrained models, while avoiding undesirable ones, has redefined Natural Language Processing (NLP) and is reshaping how we interact with computers. What was once a scientific engineering discipline—in which building blocks are stacked one on top of the other—is arguably already a complex systems science—in which emergent behaviors are sought out to support previously unimagined use cases. Despite the ever increasing number of benchmarks that measure task performance, we lack explanations of what behaviors language models exhibit that allow them to complete these tasks in the first place. We argue for a systematic effort to decompose language model behavior into categories that explain cross-task performance, to guide mechanistic explanations and help future-proof analytic research.

Keywords

language model behavioremergent properties in ai systeminterpretability and analysis method

References

[1]
A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, Ł. Kaiser, and I. Polosukhin, Attention is all you need, in Proc. 31st International Conference on Neural Information Processing Systems, Long Beach, CA, USA, 2017, pp. 6000–6060.
[2]
R. Al-Rfou, D. Choe, N. Constant, M. Guo, and L. Jones, Character-level language modeling with deeper self-attention, arXiv preprint arXiv: 1808.04444, 2018.
[3]
N. Elhage, N. Nanda, C. Olsson, T. Henighan, N. Joseph, B. Mann, A. Askell, Y. Bai, A. Chen, T. Conerly, et al., A mathematical framework for transformer circuits, https://transformer-circuits.pub/2021/framework/index.html, 2021.
[4]
C. Olsson, N. Elhage, N. Nanda, N. Joseph, N. DasSarma, T. Henighan, B. Mann, A. Askell, Y. Bai, A. Chen, et al., In-context learning and induction heads, https://transformer-circuits.pub/2022/in-context-learning-and-induction-heads/index.html#argument-architectural-requirements, 2022.
[5]
L. Wang, J. Huang, K. Huang, Z. Hu, G. Wang, and Q. Gu, Improving neural language generation with spectrum control, presented at the International Conference on Learning Representations, Addis Ababa, Ethiopia, 2020.
[6]
K. Ethayarajh, How contextual are contextualized word representations? Comparing the geometry of BERT, ELMo, and GPT-2 embeddings, in Proc. 2019 Conf. Empirical Methods in Natural Language Processing and the 9th Int. Joint Conf. Natural Language Processing (EMNLP-IJCNLP), Hong Kong, China, 2019, pp. 55–65.
Crossref
[7]
J. Gao, D. He, X. Tan, T. Qin, L. Wang, and T. Liu, Representation degeneration problem in training natural language generation models, presented at the International Conference on Learning Representations 2019, New Orleans, LA, USA, 2019.
[8]
D. Biś, M. Podkorytov, and X. Liu, Too much in common: Shifting of embeddings in transformer language models and its implications, in Proc. 2021 Conf. North American Chapter of the Association for Computational Linguistics: Human Language Technologies. Online, 2021, pp. 5117–5130.
Crossref
[9]
N. Godey, E. de la Clergerie, and B. Sagot, Is anisotropy inherent to transformers? arXiv preprint arXiv: 2306.07656, 2023.
[10]
W. Rudman and C. Eickhoff, Stable anisotropic regularization, arXiv preprint arXiv: 2305.19358, 2023.
[11]
Y. Lakretz, G. Kruszewski, T. Desbordes, D. Hupkes, S. Dehaene, and M. Baroni, The emergence of number and syntax units in LSTM language models, arXiv preprint arXiv: 1903.07435, 2019.
Crossref
[12]
C. Olah, Understanding LSTM networks, https://colah.github.io/posts/2015-08-Understanding-LSTMs/, 2015.
[13]

S. Hochreiter and J. Schmidhuber, Long short-term memory, Neural Comput., vol. 9, no. 8, pp. 1735–1780, 1997.
Crossref Google Scholar
[14]
M. Peters, M. Neumann, M. Iyyer, M. Gardner, C. Clark, K. Lee, and L. Zettlemoyer, Deep contextualized word representations, in Proc. 2018 Conf. North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long Papers), New Orleans, LA, US, 2018, pp. 2227–2237.
Crossref
[15]
J. Devlin, M. W. Chang, K. Lee, and K. Toutanova, BERT: Pre-training of deep bidirectional transformers for language understanding, in Proc. 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers), Minneapolis, Minnesota, 2019, pp. 4171–4186.
[16]
A. Karpathy, J. Johnson, and L. Fei-Fei, Visualizing and understanding recurrent networks, arXiv preprint arXiv: 1506.02078, 2015.
[17]
G. Weiss, Y. Goldberg, and E. Yahav, Thinking like transformers, in Proc. 38th International Conference on Machine Learning, Virtual Event, 2021, pp. 11080–11090.
[18]
A. Rogers, O. Kovaleva, and A. Rumshisky, A primer in BERTology: What we know about how BERT works, Trans. Assoc. Comput. Linguist., vol. 8, pp. 842–866, 2020.
Crossref
[19]
J. Bastings, Y. Belinkov, Y. Elazar, D. Hupkes, N. Saphra, and S. Wiegreffe, BlackboxNLP analyzing and interpreting neural networks for NLP, presented at the Microsoft at EMNLP 2022, Hybrid, United Arab Emirates, 2022.
[20]
M. T. Ribeiro, T. Wu, C. Guestrin, and S. Singh, Beyond accuracy: Behavioral testing of NLP models with CheckList, in Proc. 58th Annual Meeting of the Association for Computational Linguistics, Online, 2020, pp. 4902–4912.
Crossref
[21]
T. Linzen, G. Chrupała, Y. Belinkov, and D. Hupkes, Proceedings of the 2019 ACL Workshop BlackboxNLP: Analyzing and Interpreting Neural Networks for NLP, https://aclanthology.org/W19-4800/, 2019.
[22]
Y. Belinkov and Y. Bisk, Synthetic and natural noise both break neural machine translation, arXiv preprint arXiv: 1711.02173, 2017.
[23]
I. Provilkov, D. Emelianenko, and E. Voita, BPE-dropout: Simple and effective subword regularization, in Proc. 58th Annual Meeting of the Association for Computational Linguistics, Online, 2020, pp. 1882–1892.
Crossref
[24]

D. H. Wolpert and W. G. Macready, No free lunch theorems for optimization, IEEE Trans. Evol. Comput., vol. 1, no. 1, pp. 67–82, 1997.
Crossref Google Scholar
[25]
T. B. Brown, B. Mann, N. Ryder, M. Subbiah, J. Kaplan, P. Dhariwal, A. Neelakantan, P. Shyam, G. Sastry, A. Askell, et al., arXiv preprint arXiv: 2005.14165v1, 2020.
[26]
A. Venigalla and L. Li Mosaic LLMs: GPT-3 quality for <$500k, https://www.mosaicml.com/blog/gpt-3-quality-for-500k, 2023.
[27]
J. Schulman, B. Zoph, C. Kim, J. Hilton, J. Menick, J. Weng, J. F. C. Uribe, L. Fedus, L. Metz, M. Pokorny, et al., Introducing ChatGPT, https://openai.com/index/chatgpt/, 2022.
[28]
OpenAI, GPT-4 API general availability and deprecation of older models in the completions API, https://openai.com/blog/gpt-4-api-general-availability, 2023.
[29]
J. Yang, H. Jin, R. Tang, X. Han, Q. Feng, H. Jiang, B. Yin, and X. Hu, Harnessing the power of LLMs in practice: A survey on ChatGPT and beyond, arXiv preprint arXiv: 2304.13712, 2023.
[30]
C. Zhang, C. Zhang, C. Li, Y. Qiao, S. Zheng, S. K. Dam, M. Zhang, J. U. Kim, S. T. Kim, J. Choi, et al., One small step for generative AI, one giant leap for AGI: A complete survey on ChatGPT in AIGC era, arXiv preprint arXiv: 2304.06488, 2023.
[31]
T. Eloundou, S. Manning, P. Mishkin, and D. Rock, GPTs are GPTs: An early look at the labor market impact potential of large language models, arXiv preprint arXiv: 2303.10130, 2023.
[32]
A. S. George and A. S. H. George, A review of ChatGPT AI’s impact on several business sectors, Partners Universal International Innovation Journal, vol. 1, no. 1, pp. 9–23, 2023.
[33]

P. P. Ray, ChatGPT: A comprehensive review on background, applications, key challenges, bias, ethics, limitations and future scope, Internet Things Cyber Phys. Syst., vol. 3, pp. 121–154, 2023.
Crossref Google Scholar
[34]
Y. Liu, D. Iter, Y. Xu, S. Wang, R. Xu, and C. Zhu, G-eval: NLG evaluation using GPT-4 with better human alignment, arXiv preprint arXiv: 2303.16634, 2023.
Crossref
[35]
L. Zheng, W. L. Chiang, Y. Sheng, S. Zhuang, Z. Wu, Y. Zhuang, Z. Lin, Z. Li, D. Li, E. P. Xing, et al., Judging LLM-as-a-judge with MT-Bench and chatbot arena, arXiv preprint arXiv: 2306.05685v4, 2023.
[36]
A. Perry, OpenAI updates GPT-4 with new features, https://mashable.com/article/openai-chatgpt-gpt-4-function-calling-update, 2023.
[37]
M. G. Southern, OpenAI’s ChatGPT update brings improved accuracy, https://www.searchenginejournal.com/openai-chatgpt-update/476116/, 2023.
[38]
Y. Deng, OpenAI watch, https://openaiwatch.com/, 2023.
[39]
M. Alizadeh, M. Kubli, Z. Samei, S. Dehghani, J. D. Bermeo, M. Korobeynikova, and F. Gilardi, Open-source large language models outperform crowd workers and approach ChatGPT in text-annotation tasks, arXiv preprint arXiv: 2307.02179, 2023.
[40]
S. Mukherjee, A. Mitra, G. Jawahar, S. Agarwal, H. Palangi, and A. Awadallah, Orca: Progressive learning from complex explanation traces of GPT-4, https://www.microsoft.com/en-us/research/publication/orca-progressive-learning-from-complex-explanation-traces-of-gpt-4/?locale=zh-cn, 2023.
[41]
S. Gunasekar, Y. Zhang, J. Aneja, C. C. T. Mendes, A. Del Giorno, S. Gopi, M. Javaheripi, P. Kauffmann, G. de Rosa, O. Saarikivi, et al., Textbooks are all you need, arXiv preprint arXiv: 2306.11644, 2023.
[42]
H. Touvron, T. Lavril, G. Izacard, X. Martinet, M. A. Lachaux, T. Lacroix, B. Roziere, N. Goyal, E. Hambro, F. Azhar, et al., LLaMA: Open and efficient foundation language models, arXiv preprint arXiv: 2302.13971v1, 2023.
[43]
E. Almazrouei, H. Alobeidli, A. Alshamsi, A. Cappelli, R. Cojocaru, M. Debbah, E. Goffinet, D. Heslow, J. Launay, Q. Malartic, et al., Falcon-40B: An open large language model with state-of-the-art performance, 2023.
[44]
Stability AI, StableLM: StableLM: Stability AI language models, 2023.
[45]
R. Taori, I. Gulrajani, T. Zhang, Y. Dubois, X. Li, C. Guestrin, P. Liang, and T. B. Hashimoto, Stanford alpaca: An instruction-following LLaMA model, https://github.com/tatsu-lab/stanford_alpaca, 2023.
[46]
The Vicuna Team, Vicuna: An open-source chatbot impressing GPT-4 with 90%* ChatGPT quality, https://lmsys.org/blog/2023-03-30-vicuna/, 2023.
[47]
T. Dettmers, A. Pagnoni, A. Holtzman, and L. Zettlemoyer, QLoRA: Efficient finetuning of quantized LLMs, arXiv preprint arXiv: 2305.14314, 2023.
[48]
R. C. Fong and A. Vedaldi, Interpretable explanations of black boxes by meaningful perturbation, in Proc. IEEE Int. Conf. Computer Vision (ICCV), Venice, Italy, 2017, pp. 3449–3457.
Crossref
[49]
A. Holtzman, J. Buys, L. Du, M. Forbes, and Y. Choi, The curious case of neural text degeneration, arXiv preprint arXiv: 1904.09751, 2019.
[50]
C. Olah, Mechanistic interpretability, variables, and the importance of interpretable bases, https://transformer-circuits.pub/2022/mech-interp-essay/index.html, 2022.
[51]

Z. C. Lipton, The mythos of model interpretability: In machine learning, the concept of interpretability is both important and slippery, Queue, vol. 16, no. 3, pp. 31–57, 2018.
Crossref Google Scholar
[52]
A. Jacovi and Y. Goldberg, Towards faithfully interpretable NLP systems: How should we define and evaluate faithfulness? in Proc. 58th Annual Meeting of the Association for Computational Linguistics, Online, 2020, pp. 4198–4205.
Crossref
[53]
C. Chen, S. Feng, A. Sharma, and C. Tan, Machine explanations and human understanding, in Proc. 2023 ACM Conf. Fairness, Accountability, and Transparency, Chicago, IL, USA, 2023, p. 1.
Crossref
[54]
T. M. Cover, Elements of Information Theory. New York, NY, USA: John Wiley & Sons, 1999.
[55]
P. D. Grünwald, The Minimum Description Length Principle. Cambridge, MA, USA: The MIT Press, 2007.
Crossref
[56]
C. M. Barry, Who sharpened occam’s razor? https://www.irishphilosophy.com/2014/05/27/who-sharpened-occams-razor/, 2014.
[57]

R. R. Barton and M. Meckesheimer, Metamodel-based simulation optimization, Handbooks in Operations Research and Management Science, vol. 13, pp. 535–57, 2006.
Crossref Google Scholar
[58]

A. Ćwiek, S. Fuchs, C. Draxler, E. L. Asu, D. Dan, K. Hiovain, S. Kawahara, S. Koutalidis, M. Krifka, P. Lippus, et al., The bouba/kiki effect is robust across cultures and writing systems, Philos. Trans. R. Soc. Lond. B Biol. Sci., vol. 377, no. 1841, p. 20200390, 2022.
Crossref Google Scholar
[59]
A. Radford, J. Wu, R. Child, D. Luan, D. Amodei, and I. Sutskever, Language models are unsupervised multitask learners, https://cdn.openai.com/better-language-models/language_models_are_unsupervised_multitask_learners.pdf, 2019.
[60]
P. Liu, W. Yuan, J. Fu, Z. Jiang, H. Hayashi, and G. Neubig, Pre-train, prompt, and predict: A systematic survey of prompting methods in natural language processing, arXiv preprint arXiv: 2107.13586, 2021.
[61]
A. Radford, J. Wu, D. Amodei, D. Amodei, J. Clark, M. Brundage, and I. Sutskever, Better language models and their implications, https://openai.com/index/better-language-models/, 2019.
[62]
B. Z. Li, J. Yu, M. Khabsa, L. Zettlemoyer, A. Halevy, and J. Andreas, Quantifying adaptability in pre-trained language models with 500 tasks, arXiv preprint arXiv: 2112.03204, 2021.
[63]
A. Rohrbach, L. A. Hendricks, K. Burns, T. Darrell, and K. Saenko, Object hallucination in image captioning, in Proc. 2018 Conf. Empirical Methods in Natural Language Processing, Brussels, Belgiu, 2018, pp. 4035–4045.
Crossref
[64]
T. Liao, R. Taori, I. D. Raji, and L. Schmidt, Are we learning yet? A meta review of evaluation failures across machine learning, in 35th Conference on Neural Information Processing Systems Datasets and Benchmarks Track (Round 2), 2021.
[65]
National Association for College Admission Counseling, Report of the commission on the use of standardized tests in undergraduate admission, https://files.eric.ed.gov/fulltext/ED502721.pdf, 2008.
[66]
Baidu Baike, Nationwide Unified Examination for Admissions to General Universities and Colleges, (in Chinese), https://baike.baidu.com/item/普通高等学校招生全国统一考试/2567351, 2022.
[67]
C. Wang and R. Sennrich, On exposure bias, hallucination and domain shift in neural machine translation, in Proc. 58th Annual Meeting of the Association for Computational Linguistics, Online, 2020, pp. 3544–3552.
Crossref
[68]
M. Müller, A. Rios, and R. Sennrich, Domain robustness in neural machine translation, arXiv preprint arXiv: 1911.03109, 2020.
[69]
A. See, P. J. Liu, and C. D. Manning, Get to the point: Summarization with pointer-generator networks, arXiv preprint arXiv: 1704.04368, 2017.
Crossref
[70]
A. Poliak, J. Naradowsky, A. Haldar, R. Rudinger, and B. Van Durme, Hypothesis only baselines in natural language inference, in Proc. Seventh Joint Conf. Lexical and Computational Semantics, New Orleans, LA, USA, 2018, pp. 180–191.
Crossref
[71]
S. Gururangan, S. Swayamdipta, O. Levy, R. Schwartz, S. Bowman, and N. A. Smith, Annotation artifacts in natural language inference data, in Proc. 2018 Conf. North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 2 (Short Papers), New Orleans, LA, USA, 2018, pp. 107–112.
Crossref
[72]
T. A. Chang and B. K. Bergen, Language model behavior: A comprehensive survey, arXiv preprint arXiv: 2303.11504, 2023.
[73]
N. Jain, K. Saifullah, Y. Wen, J. Kirchenbauer, M. Shu, A. Saha, M. Goldblum, J. Geiping, and T. Goldstein, Bring your own data! Self-supervised evaluation for large language models, arXiv preprint arXiv: 2306.13651, 2023.
[74]
S. Pichai, An important next step on our AI journey, https://blog.google/technology/ai/bard-google-ai-search-updates/, 2023.
[75]
BigScience, Bigscience model training launched. BigScience Blog, 2022.
[76]

M. E. J. Newman, Resource letter CS–1: Complex systems, Am. J. Phys., vol. 79, no. 8, pp. 800–810, 2011.
Crossref Google Scholar
[77]
S. Bubeck, V. Chandrasekaran, R. Eldan, J. Gehrke, E. Horvitz, E. Kamar, P. Lee, Y. T. Lee, Y. Li, S. Lundberg, et al., Sparks of artificial general intelligence: Early experiments with GPT-4, arXiv preprint arXiv: 2303.12712, 2023.
[78]
J. Wei, Y. Tay, R. Bommasani, C. Raffel, B. Zoph, S. Borgeaud, D. Yogatama, M. Bosma, D. Zhou, D. Metzler, et al., Emergent abilities of large language models, arXiv preprint arXiv: 2206.07682, 2022.
[79]
R. Teehan, M. Clinciu, O. Serikov, E. Szczechla, N. Seelam, S. Mirkin, and A. Gokaslan, Emergent structures and training dynamics in large language models, in Proc. BigScience Episode #5—Workshop on Challenges & Perspectives in Creating Large Language Models, Virtual Event, 2022, pp. 146–159.
Crossref
[80]

C. D. Manning, K. Clark, J. Hewitt, U. Khandelwal, and O. Levy, Emergent linguistic structure in artificial neural networks trained by self-supervision, Proc. Natl. Acad. Sci. USA, vol. 117, no. 48, pp. 30046–30054, 2020.
Crossref Google Scholar
[81]
J. H. Holland, Complexity: A Very Short Introduction. Oxford, UK: Oxford University Press, 2014.
Crossref
[82]
U. Khandelwal, K. Clark, D. Jurafsky, and L. Kaiser, Sample efficient text summarization using a single pre-trained transformer, arXiv preprint arXiv: 1905.08836, 2019.
[83]
U. Khandelwal, H. He, P. Qi, and D. Jurafsky, Sharp nearby, fuzzy far away: How neural language models use context, in Proc. 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), Melbourne, Australia, 2018, pp. 284–294.
Crossref
[84]
L. Yu, D. Simig, C. Flaherty, A. Aghajanyan, L. Zettlemoyer, and M. Lewis, MEGABYTE: Predicting million-byte sequences with multiscale transformers, arXiv preprint arXiv: 2305.07185, 2023.
[85]
I. Beltagy, M. E. Peters, and A. Cohan, Longformer: The long-document transformer, arXiv preprint arXiv: 2004.05150, 2020.
[86]
R. Child, S. Gray, A. Radford, and I. Sutskever, Generating long sequences with sparse transformers, arXiv preprint arXiv: 1904.10509, 2019.
[87]

N. F. Liu, K. Lin, J. Hewitt, A. Paranjape, M. Bevilacqua, F. Petroni, and P. Liang, Lost in the middle: How language models use long contexts, Trans. Assoc. Comput. Linguist., vol. 12, pp. 157–173, 2024.
Crossref Google Scholar
[88]
S. Sun, K. Krishna, A. Mattarella-Micke, and M. Iyyer, Do long-range language models actually use long-range context? in Proc. 2021 Conf. Empirical Methods in Natural Language Processing, Online, 2021, pp. 807–822.
Crossref
[89]
O. Press, N. A. Smith, and M. Lewis, Shortformer: Better language modeling using shorter inputs, arXiv preprint arXiv: 2012.15832v2, 2021.
[90]
F. Petroni, T. Rocktäschel, S. Riedel, P. Lewis, A. Bakhtin, Y. Wu, and A. Miller, Language models as knowledge bases? in Proc. 2019 Conf. Empirical Methods in Natural Language Processing and the 9th Int. Joint Conf. Natural Language Processing (EMNLP-IJCNLP), Hong Kong, China, 2019, pp. 2463–2473.
Crossref
[91]
K. Cho, B. van Merrienboer, D. Bahdanau, and Y. Bengio, On the properties of neural machine translation: Encoder–decoder approaches, in Proc. SSST-8, Eighth Workshop on Syntax, Semantics and Structure in Statistical Translation, Doha, Qatar, 2014, pp. 103–111.
Crossref
[92]
N. S. Keskar, B. McCann, L. R. Varshney, C. Xiong, and R. Socher, CTRL: A conditional transformer language model for controllable generation, arXiv preprint arXiv: 1909.05858, 2019.
[93]
R. Zellers, A. Holtzman, H. Rashkin, Y. Bisk, A. Farhadi, F. Roesner, and Y. Choi, Defending against neural fake news, arXiv preprint arXiv: 1905.12616, 2019.
[94]
A. Aghajanyan, D. Okhonko, M. Lewis, M. Joshi, H. Xu, G. Ghosh, and L. Zettlemoyer, HTLM: Hyper-text pre-training and prompting of language models, arXiv preprint arXiv: 2107.06955, 2021.
[95]
S. Mishra, D. Khashabi, C. Baral, and H. Hajishirzi, Cross-task generalization via natural language crowdsourcing instructions, in Proc. 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), Dublin, Ireland, 2022, pp. 3470–3487.
Crossref
[96]
H. W. Chung, L. Hou, S. Longpre, B. Zoph, Y. Tay, W. Fedus, Y. Li, X. Wang, M. Dehghani, S. Brahma, et al., Scaling instruction-finetuned language models, arXiv preprint arXiv: 2210.11416, 2022.
[97]
L. Ouyang, J. Wu, X. Jiang, D. Almeida, C. Wainwright, P. Mishkin, C. Zhang, S. Agarwal, K. Slama, A. Ray, et al., Training language models to follow instructions with human feedback, arXiv preprint arXiv: 2203.02155, 2022.
[98]
C. Zhou, P. Liu, P. Xu, S. Iyer, J. Sun, Y. Mao, X. Ma, A. Efrat, P. Yu, L. Yu, et al., LIMA: Less is more for alignment, arXiv preprint arXiv: 2305.11206, 2023.
[99]
S. Mittal, S. Diallo, and A. Tolk, Emergent Behavior in Complex Systems Engineering: A Modeling and Simulation Approach. New York, NY, USA: John Wiley & Sons, 2018.
Crossref
[100]
G. Ilharco, R. Zellers, A. Farhadi, and H. Hajishirzi, Probing contextual language models for common ground with visual representations, in Proc. 2021 Conf. North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Online, 2021, pp. 5367–5377.
Crossref
[101]
L. Parcalabescu, A. Gatt, A. Frank, and I. Calixto, Seeing past words: Testing the cross-modal capabilities of pretrained V&L models on counting tasks, arXiv preprint arXiv: 2012.12352, 2020.
[102]
I. Tenney, D. Das, and E. Pavlick, BERT rediscovers the classical NLP pipeline, in Proc. 57th Annual Meeting of the Association for Computational Linguistics, Florence, Italy, 2019, pp. 4593–4601.
Crossref
[103]
G. Daras and A. G. Dimakis, Discovering the hidden vocabulary of DALLE-2, arXiv preprint arXiv: 2206.00169, 2022.
[104]
B. Hilton, No, DALL-E doesn’t have a secret language. (or at least, we haven’t found one yet) this viral DALL-E thread has some pretty astounding claims, but maybe the reason they’re so astounding is that, for the most part, they’re not true. thread (1/15), https://t.co/8F2WDp7lTK. https://twitter.com/benjamin_hilton/status/1531780892972175361?lang=en, 2022.
[105]

R. T. McCoy, P. Smolensky, T. Linzen, J. Gao, and A. Celikyilmaz, How much do language models copy from their training data? Evaluating linguistic novelty in text generation using RAVEN, Trans. Assoc. Comput. Linguist., vol. 11, pp. 652–670, 2023.
Crossref Google Scholar
[106]
N. Carlini, D. Ippolito, M. Jagielski, K. Lee, F. Tramer, and C. Zhang, Quantifying memorization across neural language models, arXiv preprint arXiv: 2202.07646, 2022.
[107]
K. Lee, D. Ippolito, A. Nystrom, C. Zhang, D. Eck, C. Callison-Burch, and N. Carlini, Deduplicating training data makes language models better, in Proc. 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), Dublin, Ireland, 2022, pp. 8424–8445.
Crossref
[108]
K. Tirumala, A. H. Markosyan, L. Zettlemoyer, and A. Aghajanyan, Memorization without overfitting: Analyzing the training dynamics of large language models, arXiv preprint arXiv: 2205.10770v2, 2022.
[109]
T. Blevins and L. Zettlemoyer, Language contamination helps explain the cross-lingual capabilities of English pretrained models, arXiv preprint arXiv: 2204.08110, 2022.
Crossref
[110]
X. V. Lin, T. Mihaylov, M. Artetxe, T. Wang, S. Chen, D. Simig, M. Ott, N. Goyal, S. Bhosale, J. Du, et al., Few-shot learning with multilingual language models, arXiv preprint arXiv: 2112.10668, 2021.
Crossref
[111]
N. Kandpal, E. Wallace, and C. Raffel, Deduplicating training data mitigates privacy risks in language models, arXiv preprint arXiv: 2202.06539, 2022.
[112]
L. Kiho, ChatGPT_DAN: ChatGPT DAN, jailbreaks prompt.
[113]
F. Stahlberg, I. Kulikov, and S. Kumar, Uncertainty determines the adequacy of the mode and the tractability of decoding in sequence-to-sequence models, in Proc. 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), Dublin, Ireland, 2022, pp. 8634–8645.
Crossref
[114]
X. Shi, Y. Xiao, and K. Knight, Why neural machine translation prefers empty outputs, arXiv preprint arXiv: 2012.13454, 2020.
[115]
F. Stahlberg and B. Byrne, On NMT search errors and model errors: Cat got your tongue? in Proc. 2019 Conf. Empirical Methods in Natural Language Processing and the 9th Int. Joint Conf. Natural Language Processing (EMNLP-IJCNLP), Hong Kong, China, 2019, pp. 3356–3362.
Crossref
[116]
Z. Xie, T. Cohn, and J. H. Lau, Can very large pretrained language models learn storytelling with a few examples? 2023.
[117]
A. See, A. Pappu, R. Saxena, A. Yerukola, and C. D. Manning, Do massively pretrained language models make better storytellers? in Proc. 23rd Conf. Computational Natural Language Learning (CoNLL), Hong Kong, China, 2019, pp. 843–861.
Crossref
[118]
A. Lazaridou and M. Baroni, Emergent multi-agent communication in the deep learning era, arXiv preprint arXiv: 2006.02419, 2020.
[119]
S. Steinert-Threlkeld, X. Zhou, Z. Liu, and C. M. Downey, Emergent communication fine-tuning (EC-FT) for pretrained language models, presented at the ICLR 2022 EmeCom Workshop, 2022.
[120]
A. Warstadt, L. Choshen, A. Mueller, A. Williams, E. Wilcox, and C. Zhuang, Call for papers: The BabyLM challenge: Sample-efficient pretraining on a developmentally plausible corpus, arXiv preprint arXiv: 2301.11796, 2023.
Crossref
[121]
D. Hafner, T. Lillicrap, I. S. Fischer, R. Villegas, D. R. Ha, H. Lee, and J. Davidson, Learning latent dynamics for planning from pixels, in Proc. 36th International Conference on Machine Learning: ICML 2019, Long Beach, CA, USA, 2019, pp. 2555–2565.
[122]
M. Morin and M. Willetts, Non-determinism in TensorFlow ResNets, arXiv preprint arXiv: 2001.11396, 2020.
[123]

W. Prinz, Messung kontra augenschein, Psychol. Rundsch., vol. 57, no. 2, pp. 106–111, 2006.
Crossref Google Scholar
[124]
R. Rombach, A. Blattmann, D. Lorenz, P. Esser, and B. Ommer, High-resolution image synthesis with latent diffusion models, in Proc. IEEE/CVF Conf. Computer Vision and Pattern Recognition (CVPR), New Orleans, LA, USA, 2022, pp. 10674–10685.
Crossref
[125]
T. McCoy, E. Pavlick, and T. Linzen, Right for the wrong reasons: Diagnosing syntactic heuristics in natural language inference, in Proc. 57th Annual Meeting of the Association for Computational Linguistics, Florence, Italy, 2019, pp. 3428–3448.
Crossref
[126]
M. Caro, Caro’s Book of Poker Tells. Washington, DC, USA: Cardoza Publishing, 2003.
[127]
M. Mori, The uncanny valley: The original essay by Masahiro Mori, https://spectrum.ieee.org/the-uncanny-valley, 2012.
Crossref
[128]
A. Wilson, How to jailbreak ChatGPT to unlock its full potential, https://approachableai.com/how-to-jailbreak-chatgpt/, 2023.
[129]
E. Eliaçık, Playing with fire: The leaked plugin DAN unchains ChatGPT from its moral and ethical restrictions, https://dataconomy.com/2023/03/31/chatgpt-dan-prompt-how-to-jailbreak-chatgpt/, 2023.
[130]
M. Le, A. Vyas, B. Shi, B. Karrer, L. Sari, R. Moritz, M. Williamson, V. Manohar, Y. Adi, J. Mahadeokar, et al., Voicebox: Text-guided multilingual universal speech generation at scale, arXiv preprint arXiv: 2306.15687, 2023.
[131]
Z. Luo, D. Chen, Y. Zhang, Y. Huang, L. Wang, Y. Shen, D. Zhao, J. Zhou, and T. Tan, VideoFusion: Decomposed diffusion models for high-quality video generation, arXiv preprint arXiv: 2303.08320, 2023.
Crossref
Journal of Social Computing
Volume 6 Issue 2,
June 2025
Pages 75-94
DOI: 10.23919/JSC.2025.0009
Cite this article:
Holtzman A, West P, Zettlemoyer L. Generative Models as a Complex Systems Science: How Can We Make Sense of Large Language Model Behavior?. Journal of Social Computing, 2025, 6(2): 75-94. https://doi.org/10.23919/JSC.2025.0009

102

Views

19

Downloads

0

Crossref

0

Scopus

Altmetrics
Received: 29 October 2024
Revised: 18 May 2025
Accepted: 21 May 2025
Published: 30 June 2025
© The author(s) 2025.

The articles published in this open access journal are distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/).
Return