Ukyo Honda

I am a research scientist at CyberAgent AI Lab. My research interests include text generation and its evaluation. Currently, I am working on text decoding algorithms, reinforcement learning for text generation, and out-of-distribution robustness. My dissertation was about enhancing scene coverage and informativeness of image captioning models.

Contact: honda_ukyo at cyberagent.co.jp
[Google Scholar] [Semantic Scholar] [GitHub] [Speaker Deck]

Publications

Refereed

• Not Eliminate but Aggregate: Post-Hoc Control over Mixture-of-Experts to Address Shortcut Shifts in Natural Language Understanding
  Ukyo Honda, Tatsushi Oka, Peinan Zhang, and Masato Mita. TACL 2024.
  [arxiv][code]

• Generating Diverse and High-Quality Texts by Minimum Bayes Risk Decoding
  Yuu Jinnai, Ukyo Honda, Tetsuro Morimura, and Peinan Zhang. Findings of ACL 2024.
  [arxiv][code]

• Model-Based Minimum Bayes Risk Decoding for Text Generation
  Yuu Jinnai, Tetsuro Morimura, Ukyo Honda, Kaito Ariu, and Kenshi Abe. ICML 2024.
  [arxiv][code]

• Reinforcement Learning for Edit-Based Non-Autoregressive Neural Machine Translation
  Hao Wang, Tetsuro Morimura, Ukyo Honda, and Daisuke Kawahara. NAACL-SRW 2024.
  [arxiv]

• On the True Distribution Approximation of Minimum Bayes-Risk Decoding
  Atsumoto Ohashi, Ukyo Honda, Tetsuro Morimura, and Yuu Jinnai. NAACL 2024.
  [arxiv][code]

• CAMERA³: An Evaluation Dataset for Controllable Ad Text Generation in Japanese
  Go Inoue, Akihiko Kato, Masato Mita, Ukyo Honda, and Peinan Zhang. LREC-COLING 2024.
  [code]

• A Single Linear Layer Yields Task-Adapted Low-Rank Matrices
  Hwichan Kim, Shota Sasaki, Sho Hoshino, and Ukyo Honda. LREC-COLING 2024.
  [arxiv][code]

• Switching to Discriminative Image Captioning by Relieving a Bottleneck of Reinforcement Learning
  Ukyo Honda, Taro Watanabe, and Yuji Matsumoto. WACV 2023.
  [arxiv][code]

• Explainable Natural Language Inference in the Legal Domain via Text Generation
  Jungmin Choi, Ukyo Honda, Taro Watanabe, and Kentaro Inui. Transactions of the Japanese Society for Artificial Intelligence 2023.

• Removing Partial Mismatches in Unsupervised Image Captioning
  Ukyo Honda, Atsushi Hashimoto, Taro Watanabe, and Yuji Matsumoto. Transactions of the Japanese Society for Artificial Intelligence 2022.

• Law Retrieval with Supervised Contrastive Learning using the Hierarchical Structure of Law
  Jungmin Choi, Ukyo Honda, Taro Watanabe, Kentaro Inui, and Hiroki Ouchi. PACLIC 2022.

• Removing Word-Level Spurious Alignment between Images and Pseudo-Captions in Unsupervised Image Captioning
  Ukyo Honda, Yoshitaka Ushiku, Atsushi Hashimoto, Taro Watanabe, and Yuji Matsumoto. EACL 2021.
  [arxiv][code]

• Characterization of Pulmonary Nodules in Computed Tomography Images Based on Pseudo-Labeling Using Radiology Reports
  Yohei Momoki, Akimichi Ichinose, Yutaro Shigeto, Ukyo Honda, Keigo Nakamura, and Yuji Matsumoto. IEEE Transactions on Circuits and Systems for Video Technology 2021.

• Pruning Basic Elements for Better Automatic Evaluation of Summaries
  Ukyo Honda, Tsutomu Hirao, and Masaaki Nagata. NAACL 2018.
  [code]

Preprints

• FaithCAMERA: Construction of a Faithful Dataset for Ad Text Generation
  Akihiko Kato, Masato Mita, Soichiro Murakami, Ukyo Honda, Sho Hoshino, and Peinan Zhang. arxiv 2024.
  [code]

• Annotation-Efficient Preference Optimization for Language Model Alignment
  Yuu Jinnai and Ukyo Honda. arxiv 2024.
  [code]

• On the Depth between Beam Search and Exhaustive Search for Text Generation
  Yuu Jinnai, Tetsuro Morimura, and Ukyo Honda. arxiv 2023.

Awards

• Committee Special Award / 委員特別賞
  Annual Meeting of the Association for Natural Language Processing / 言語処理学会年次大会 2025
  Mixture-of-Expertsの悲観的な統合による頑健な自然言語理解
  Ukyo Honda, Tatsushi Oka, Peinan Zhang, and Masato Mita

• Committee Special Award / 委員特別賞
  Annual Meeting of the Association for Natural Language Processing / 言語処理学会年次大会 2025
  FaithCAMERA: 広告文生成タスクのための忠実性を担保した評価データセットの構築
  Akihiko Kato, Masato Mita, Soichiro Murakami, Ukyo Honda, Sho Hoshino, and Peinan Zhang

• Best Paper Award / 最優秀賞
  Annual Meeting of the Association for Natural Language Processing / 言語処理学会年次大会 2022
  強化学習における画像キャプションの低識別性問題とLong-Tail分類手法を用いた対処
  Ukyo Honda, Taro Watanabe, and Yuji Matsumoto

• Outstanding Research Award / 優秀研究賞 (インタラクティブセッション発表部門)
  Annual Conference of the Japanese Society for Artificial Intelligence / 人工知能学会全国大会 2022
  法律の階層構造を利用した教師あり対照学習による法律検索
  Jungmin Choi, Ukyo Honda, Taro Watanabe, Kentaro Inui, and Hiroki Ouchi

• Young Researcher Award / 若手奨励賞
  Annual Meeting of the Association for Natural Language Processing / 言語処理学会年次大会 2021
  画像と単語の不一致を考慮した疑似教師ありキャプション生成
  Ukyo Honda, Yoshitaka Ushiku, Atsushi Hashimoto, Taro Watanabe, and Yuji Matsumoto

Education

• Mar. 2023: Ph.D. in Engineering at Nara Institute of Science and Technology

• Mar. 2019: M.S. in Engineering at Nara Institute of Science and Technology

• Sep. 2016: B.A. in Law at Keio University

Experience

• Sep. - Dec. 2019: Research Internship at OMRON SINIC X Corp.

• Aug. - Oct. 2018: Research Internship at OMRON SINIC X Corp.

• Sep. 2017: Research Internship at NTT Comunication Science Laboratories

• Sep. 2016 - Mar. 2017: Translation Checker at Honyaku Center Inc.

• Sep. 2015 - Jun. 2016: Exchange Student at University of California, Davis