{"type":"ScholarlyArticle","title":"Interpreting Pretrained Language Models via Concept Bottlenecks (Extended Abstract)","authors":[{"name":"Zhen Tan","slug":"zhen-tan","given":"Zhen","family":"Tan","affiliation":"Arizona State University , Arizona , USA"},{"name":"Lu Cheng","slug":"lu-cheng","given":"Lu","family":"Cheng","affiliation":"University of Illinois Chicago , Illinois , USA"},{"name":"Song Wang","slug":"song-wang","given":"Song","family":"Wang","affiliation":"University of Virginia , Virginia , USA"},{"name":"Yuan Bo","slug":"yuan-bo","given":"Yuan","family":"Bo","affiliation":"Independent Researcher"},{"name":"Jundong Li","slug":"jundong-li","given":"Jundong","family":"Li","affiliation":"University of Virginia , Virginia , USA"},{"name":"Huan Liu","slug":"huan-liu","given":"Huan","family":"Liu","affiliation":"Arizona State University , Arizona , USA"}],"abstract":"Pretrained language models (PLMs) achieve state-of-the-art results but often function as ``black boxes'', hindering interpretability and responsible deployment. While methods like attention analysis exist, they often lack clarity and intuitiveness. We propose interpreting PLMs through high-level, human-understandable concepts using Concept Bottleneck Models (CBMs). This extended abstract introduces C3M (ChatGPT-guided Concept augmentation with Concept-level Mixup), a novel framework for training Concept-Bottleneck-Enabled PLMs (CBE-PLMs). C3M leverages Large Language Models (LLMs) like ChatGPT to augment concept sets and generate noisy concept labels, combined with a concept-level MixUp mechanism to enhance robustness and effectively learn from both human-annotated and machine-generated concepts. Empirical results show our approach provides intuitive explanations, aids model diagnosis via test-time intervention, and improves the interpretability-utility trade-off, even with limited or noisy concept annotations. This is an concise version of [Tan et al., 2024b], recipient of the Best Paper Award at PAKDD 2024. Code and data are released at https://github.com/Zhen-Tan-dmml/CBM_NLP.git.","doi":"10.24963/ijcai.2025/1221","url":"https://ijcai.org/proceedings/2025/1221","pdf":"https://ijcai.org/proceedings/2025/1221.pdf","markdown":"https://ijcai.org/proceedings/2025/1221.md","bibtex":"https://ijcai.org/proceedings/2025/bibtex/1221","volume":{"slug":"2025","title":"Proceedings of the Thirty-Fourth International Joint Conference on Artificial Intelligence","short_title":"IJCAI 2025","url":"https://ijcai.org/proceedings/2025/","year":2025,"location":"Montreal, Canada","dates":"16-22 August 2025","editors":"James Kwok","isbn":"978-1-956792-06-5","publisher":"International Joint Conferences on Artificial Intelligence Organization","doi_prefix":"10.24963/ijcai.2025"},"section":"Sister Conferences Best Papers","track":"Sister Conferences Best Papers","number":1221,"pages":{"first":10942,"last":10946},"keywords":["Sister Conferences Best Papers: AI Ethics, Trust, Fairness","Sister Conferences Best Papers: Natural Language Processing","Sister Conferences Best Papers: Machine Learning","Sister Conferences Best Papers: Humans and AI"],"videos":[],"references":[{"position":1,"raw":"[Berthelot et al., 2019] David Berthelot, Nicholas Carlini, Ian Goodfellow, Nicolas Papernot, Avital Oliver, and Colin A Raffel. Mixmatch: A holistic approach to semi- supervised learning. Advances in neural information pro- cessing systems, 32, 2019.","title":"Mixmatch: A holistic approach to semisupervised learning","authors":"David Berthelot, Nicholas Carlini, Ian Goodfellow, Nicolas Papernot, Avital Oliver, Colin A Raffel","year":"2019","venue":"Advances in neural information processing systems","doi":null,"url":null,"internal_url":null},{"position":2,"raw":"[Bills et al., 2023] Steven Bills, Nick Cammarata, Dan Mossing, Henk Tillman, Leo Gao, Gabriel Goh, Ilya Sutskever, Jan Leike, Jeff Wu, and William Saun- ders. Language models can explain neurons in lan- guage models. https://openaipublic.blob.core.windows. net/neuron-explainer/paper/index.html, 2023.","title":"Language models can explain neurons in language models","authors":"Steven Bills, Nick Cammarata, Dan Mossing, Henk Tillman, Leo Gao, Gabriel Goh, Ilya Sutskever, Jan Leike, Jeff Wu, William Saunders","year":"2023","venue":null,"doi":null,"url":"https://openaipublic.blob.core.windows.net/neuron-explainer/paper/index.html","internal_url":null},{"position":3,"raw":"[Devlin et al., 2018] Jacob Devlin, Ming-Wei Chang, Ken- ton Lee, and Kristina Toutanova. Bert: Pre-training of deep bidirectional transformers for language understand- ing. arXiv preprint arXiv:1810.04805, 2018.","title":"Bert: Pre-training of deep bidirectional transformers for language understanding","authors":"Jacob Devlin, Ming-Wei Chang, Kenton Lee, Kristina Toutanova","year":"2018","venue":null,"doi":null,"url":null,"internal_url":null},{"position":4,"raw":"[Diao et al., 2022] Shizhe Diao, Zhichao Huang, Ruijia Xu, Xuechun Li, Yong Lin, Xiao Zhou, and Tong Zhang. Black-box prompt learning for pre-trained language mod- els. arXiv preprint arXiv:2201.08531, 2022.","title":"Black-box prompt learning for pre-trained language models","authors":"Diao","year":"2022","venue":null,"doi":null,"url":null,"internal_url":null},{"position":5,"raw":"[Englesson and Azizpour, 2021] Erik Englesson and Hos- sein Azizpour. Generalized jensen-shannon divergence loss for learning with noisy labels. Advances in Neural Information Processing Systems, 34:30284-30297, 2021.","title":"Erik Englesson and Hossein Azizpour. Generalized jensen-shannon divergence loss for learning with noisy labels","authors":"Azizpour Englesson","year":"2021","venue":"Advances in Neural Information Processing Systems","doi":null,"url":null,"internal_url":null},{"position":6,"raw":"[Galassi et al., 2020] Andrea Galassi, Marco Lippi, and Paolo Torroni. Attention in natural language processing. IEEE transactions on neural networks and learning sys- tems, 32(10):4291-4308, 2020.","title":"Attention in natural language processing","authors":"Andrea Galassi, Marco Lippi, Paolo Torroni","year":"2020","venue":"IEEE transactions on neural networks and learning systems","doi":null,"url":null,"internal_url":null},{"position":7,"raw":"[He et al., 2016] Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recog- nition. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 770-778, 2016.","title":"Deep residual learning for image recognition","authors":"Kaiming He, Xiangyu Zhang, Shaoqing Ren, Jian Sun","year":"2016","venue":"Proceedings of the IEEE conference on computer vision and pattern recognition","doi":null,"url":null,"internal_url":null},{"position":8,"raw":"[Kim et al., 2018] Been Kim, Martin Wattenberg, Justin Gilmer, Carrie Cai, James Wexler, Fernanda Viegas, et al. Interpretability beyond feature attribution: Quantitative testing with concept activation vectors (tcav). In Interna- tional conference on machine learning, pages 2668-2677.","title":"Interpretability beyond feature attribution: Quantitative testing with concept activation vectors (tcav)","authors":"Kim","year":"2018","venue":"International conference on machine learning","doi":null,"url":null,"internal_url":null},{"position":9,"raw":"PMLR, 2018.","title":null,"authors":"Pmlr","year":"2018","venue":null,"doi":null,"url":null,"internal_url":null},{"position":10,"raw":"[Koh et al., 2020] Pang Wei Koh, Thao Nguyen, Yew Siang Tang, Stephen Mussmann, Emma Pierson, Been Kim, and Percy Liang. Concept bottleneck models. In Inter- national Conference on Machine Learning, pages 5338- 5348. PMLR, 2020.","title":"Concept bottleneck models","authors":"Pang Wei Koh, Thao Nguyen, Siang Yew, Stephen Tang, Emma Mussmann, Been Pierson, Percy Kim, Liang","year":"2020","venue":"International Conference on Machine Learning","doi":null,"url":null,"internal_url":null},{"position":11,"raw":"[Liu et al., 2022] Yang Liu, Hao Cheng, and Kun Zhang. Identifiability of label noise transition matrix. arXiv preprint arXiv:2202.02016, 2022.","title":"Identifiability of label noise transition matrix","authors":"Yang Liu, Hao Cheng, Kun Zhang","year":"2022","venue":null,"doi":null,"url":null,"internal_url":null},{"position":12,"raw":"[Losch et al., 2019] Max Losch, Mario Fritz, and Bernt Schiele. Interpretability beyond classification out- put: Semantic bottleneck networks. arXiv preprint arXiv:1907.10882, 2019.","title":"Interpretability beyond classification output: Semantic bottleneck networks","authors":"Losch","year":"2019","venue":null,"doi":null,"url":null,"internal_url":null},{"position":13,"raw":"[Madsen et al., 2022] Andreas Madsen, Siva Reddy, and Sarath Chandar. Post-hoc interpretability for neural nlp: A survey. ACM Computing Surveys, 55(8):1-42, 2022.","title":"Post-hoc interpretability for neural nlp: A survey","authors":"Andreas Madsen, Siva Reddy, Sarath Chandar","year":"2022","venue":"ACM Computing Surveys","doi":null,"url":null,"internal_url":null},{"position":14,"raw":"[Mikolov et al., 2013] Tomas Mikolov, Kai Chen, Greg Corrado, and Jeffrey Dean. Efficient estimation of word representations in vector space. arXiv preprint arXiv:1301.3781, 2013.","title":"Efficient estimation of word representations in vector space","authors":"Tomas Mikolov, Kai Chen, Greg Corrado, Jeffrey Dean","year":"2013","venue":null,"doi":null,"url":null,"internal_url":null},{"position":15,"raw":"[Németh et al., 2020] Renáta Németh, Domonkos Sik, and Fanni Máté. Machine learning of concepts hard even for humans: The case of online depression fo- rums. International Journal of Qualitative Methods, 19:1609406920949338, 2020. [","title":"Machine learning of concepts hard even for humans: The case of online depression forums","authors":"Renáta Németh, Domonkos Sik, Fanni Máté","year":"2020","venue":"International Journal of Qualitative Methods","doi":null,"url":null,"internal_url":null},{"position":16,"raw":"[OpenAI, 2023] OpenAI. Gpt-4 technical report, 2023.","title":null,"authors":"Openai","year":"2023","venue":null,"doi":null,"url":null,"internal_url":null},{"position":17,"raw":"[Ross et al., 2021] Alexis Ross, Ana Marasović, and Matthew E Peters. Explaining nlp models via minimal contrastive editing (mice). In Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021, pages 3840-3852, 2021.","title":"Explaining nlp models via minimal contrastive editing (mice)","authors":"Alexis Ross, Ana Marasović, Matthew E Peters","year":"2021","venue":"Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021","doi":null,"url":null,"internal_url":null},{"position":18,"raw":"[Sohn et al., 2020] Kihyuk Sohn, David Berthelot, Nicholas Carlini, Zizhao Zhang, Han Zhang, Colin A Raffel, Ekin Dogus Cubuk, Alexey Kurakin, and Chun-Liang Li. Fixmatch: Simplifying semi-supervised learning with con- sistency and confidence. Advances in neural information processing systems, 33:596-608, 2020.","title":"Fixmatch: Simplifying semi-supervised learning with consistency and confidence","authors":"Sohn","year":"2020","venue":"Advances in neural information processing systems","doi":null,"url":null,"internal_url":null},{"position":19,"raw":"[Tan and Liu, 2025] Zhen Tan and Huan Liu. Intrinsic barri- ers to explaining deep foundation models, 2025.","title":"Intrinsic barriers to explaining deep foundation models","authors":"Zhen Tan, Huan Liu","year":"2025","venue":null,"doi":null,"url":null,"internal_url":null},{"position":20,"raw":"[Tan et al., 2024a] Zhen Tan, Tianlong Chen, Zhenyu Zhang, and Huan Liu. Sparsity-guided holistic explanation for llms with interpretable inference-time intervention. In Proceedings of the AAAI conference on artificial intelli- gence, volume 38, pages 21619-21627, 2024.","title":"Sparsity-guided holistic explanation for llms with interpretable inference-time intervention","authors":"Tan","year":"2024","venue":"Proceedings of the AAAI conference on artificial intelligence","doi":null,"url":null,"internal_url":null},{"position":21,"raw":"[Tan et al., 2024b] Zhen Tan, Lu Cheng, Song Wang, Bo Yuan, Jundong Li, and Huan Liu. Interpreting pre- trained language models via concept bottlenecks. In Pacific-Asia Conference on Knowledge Discovery and Data Mining, pages 56-74. Springer, 2024.","title":"Interpreting pretrained language models via concept bottlenecks","authors":"Tan","year":"2024","venue":"Pacific-Asia Conference on Knowledge Discovery and Data Mining","doi":null,"url":null,"internal_url":null},{"position":22,"raw":"[Tan et al., 2025a] Zhen Tan, Jie Peng, Song Wang, Lijie Hu, Tianlong Chen, and Huan Liu. Tuning-free accountable in- tervention for llm deployment-a metacognitive approach. In Proceedings of the AAAI Conference on Artificial Intel- ligence, volume 39, pages 25237-25245, 2025.","title":"Tuning-free accountable intervention for llm deployment-a metacognitive approach","authors":"Tan","year":"2025","venue":"Proceedings of the AAAI Conference on Artificial Intelligence","doi":null,"url":null,"internal_url":null},{"position":23,"raw":"[Tan et al., 2025b] Zhen Tan, Song Wang, Yifan Li, Yu Kong, Jundong Li, Tianlong Chen, and Huan Liu. Are we merely justifying results ex post facto? quantifying ex- planatory inversion in post-hoc model explanations. arXiv preprint arXiv:2504.08919, 2025.","title":"Are we merely justifying results ex post facto? quantifying explanatory inversion in post-hoc model explanations","authors":"Tan","year":"2025","venue":null,"doi":null,"url":null,"internal_url":null},{"position":24,"raw":"[Vig et al., 2020] Jesse Vig, Sebastian Gehrmann, Yonatan Belinkov, Sharon Qian, Daniel Nevo, Yaron Singer, and Stuart Shieber. Investigating gender bias in language mod- els using causal mediation analysis. Advances in neural information processing systems, 33:12388-12401, 2020.","title":"Investigating gender bias in language models using causal mediation analysis","authors":"Jesse Vig, Sebastian Gehrmann, Yonatan Belinkov, Sharon Qian, Daniel Nevo, Yaron Singer, Stuart Shieber","year":"2020","venue":"Advances in neural information processing systems","doi":null,"url":null,"internal_url":null},{"position":25,"raw":"[Yin and Neubig, 2022] Kayo Yin and Graham Neubig. In- terpreting language models with contrastive explanations. arXiv preprint arXiv:2202.10419, 2022.","title":"Interpreting language models with contrastive explanations","authors":"Neubig Yin, Kayo Yin, Graham Neubig","year":"2022","venue":null,"doi":null,"url":null,"internal_url":null},{"position":26,"raw":"[Zarlenga et al., 2022] Mateo Espinosa Zarlenga, Pietro Bar- biero, Gabriele Ciravegna, Giuseppe Marra, Francesco Gi- annini, Michelangelo Diligenti, Frederic Precioso, Stefano Melacci, Adrian Weller, Pietro Lio, et al. Concept embed- ding models. In NeurIPS 2022-36th Conference on Neural Information Processing Systems, 2022.","title":"Concept embedding models","authors":"Mateo Espinosa Zarlenga, Pietro Barbiero, Gabriele Ciravegna, Giuseppe Marra, Francesco Giannini, Michelangelo Diligenti, Frederic Precioso, Stefano Melacci, Adrian Weller, Pietro Lio","year":"2022","venue":"NeurIPS 2022-36th Conference on Neural Information Processing Systems","doi":null,"url":null,"internal_url":null},{"position":27,"raw":"[Zhang et al., 2017] Hongyi Zhang, Moustapha Cisse, Yann N Dauphin, and David Lopez-Paz. mixup: Be- yond empirical risk minimization. arXiv preprint arXiv:1710.09412, 2017.","title":"mixup: Beyond empirical risk minimization","authors":"Hongyi Zhang, Moustapha Cisse, David Yann N Dauphin, Lopez-Paz","year":"2017","venue":null,"doi":null,"url":null,"internal_url":null},{"position":28,"raw":"[Zhang et al., 2022] Wenxuan Zhang, Xin Li, Yang Deng, Lidong Bing, and Wai Lam. A survey on aspect-based sentiment analysis: tasks, methods, and challenges. IEEE Transactions on Knowledge and Data Engineering, 2022.","title":"A survey on aspect-based sentiment analysis: tasks, methods, and challenges","authors":"Wenxuan Zhang, Xin Li, Yang Deng, Lidong Bing, Wai Lam","year":"2022","venue":"IEEE Transactions on Knowledge and Data Engineering","doi":null,"url":null,"internal_url":null},{"position":29,"raw":"[Zhu et al., 2020] Jinhua Zhu, Yingce Xia, Lijun Wu, Di He, Tao Qin, Wengang Zhou, Houqiang Li, and Tieyan Liu. In- corporating bert into neural machine translation. In Inter- national Conference on Learning Representations, 2020.","title":"Incorporating bert into neural machine translation","authors":"Jinhua Zhu, Yingce Xia, Lijun Wu, Di He, Tao Qin, Wengang Zhou, Houqiang Li, Tieyan Liu","year":"2020","venue":"International Conference on Learning Representations","doi":null,"url":null,"internal_url":null}],"license_note":"Copyright International Joint Conferences on Artificial Intelligence."}