Tosho Hirasawa

平澤 寅庄

Research Engineer in the R&D Group at LayerX. I work on large language models — building agent systems on top of them, and studying how they can be applied and evaluated. Before that I was a researcher at OMRON SINIC X, and my Ph.D. was on multimodal machine translation.

Interests

Career

Education

Awards

Publications

All venues are listed together by year. Author lists follow the published order.

2026

  1. conference Ritsu Sakabe, Hwichan Kim, Tosho Hirasawa, Mamoru Komachi. Assessing the Capabilities of LLMs in Humor: A Multi-dimensional Analysis of Oogiri Generation and Evaluation. Proceedings of the 40th AAAI Conference on Artificial Intelligence (AAAI-26). paperarXiv
  2. conference Huaying Zhang, Atsushi Hashimoto, Tosho Hirasawa. Evaluating the Capability of Video Question Generation for Expert Knowledge Elicitation. IEEE/CVF Winter Conference on Applications of Computer Vision (WACV 2026). paperarXiv
  3. conference 佐藤祥太, 木山朔, 平澤寅庄, 小町守. CLIP のモダリティギャップは本当に埋めるべきか？ 言語処理学会第32回年次大会 (NLP2026), 宇都宮. 論文
  4. conference 森清忠, 田中翔平, 平澤寅庄, 小津野将, 吉野幸一郎, 牛久祥孝. 主張と根拠の繋がりを評価する科学論文レビュー評価指標. 言語処理学会第32回年次大会 (NLP2026), 宇都宮. 論文
  5. workshop 佐藤祥太, 木山朔, 平澤寅庄, 小町守. 予測分布の偏りに着目した CLIP のモダリティギャップ分析. 情報処理学会 自然言語処理研究会 第268回研究発表会 (NL268), 一橋大学 — 優秀研究賞. 発表会
  6. workshop 坂部立, 金輝燦, 平澤寅庄, 小町守. LLM のユーモア評価能力に関する多次元分析：大喜利評価における人間との乖離とバイアス. 情報処理学会 自然言語処理研究会 第267回研究発表会 (NL267) — 優秀研究賞.
  7. preprint Kuniaki Saito, Risa Shinoda, Shohei Tanaka, Tosho Hirasawa, Fumio Okura, Yoshitaka Ushiku. HalDec-Bench: Benchmarking Hallucination Detector in Image Captioning. arXiv:2603.15253. arXiv
  8. preprint Yuzheng Xu, Tosho Hirasawa, Tadashi Kozuno, Yoshitaka Ushiku. Am I More Pointwise or Pairwise? Revealing Position Bias in Rubric-Based LLM-as-a-Judge. arXiv:2602.02219. arXiv
  9. preprint Kiyotada Mori, Shohei Tanaka, Tosho Hirasawa, Tadashi Kozuno, Koichiro Yoshino, Yoshitaka Ushiku. WarrantScore: Modeling Warrants between Claims and Evidence for Substantiation Evaluation in Peer Reviews. arXiv:2601.17377. arXiv

2025

  1. conference 中島京太郎, 金輝燦, 平澤寅庄, 榎本大晟, 陳宙斯, 小町守. 大規模言語モデルに対するチューニング手法の調査：内部のアクセス性に基づく分類と比較. 言語処理学会第31回年次大会 (NLP2025), 長崎. 論文
  2. conference 増田大河, 田中翔平, 平澤寅庄, 牛久祥孝. Paper2Poster: LLM を用いた学術ポスターの生成. 言語処理学会第31回年次大会 (NLP2025), 長崎. 論文
  3. conference 田中翔平, 平澤寅庄, 牛久祥孝. 根拠に基づいたレビュー生成のための LLM を用いた自動アノテーションの検討. 言語処理学会第31回年次大会 (NLP2025), 長崎. 論文
  4. workshop Yuzheng Xu, Tosho Hirasawa, Seiya Kawano, Shota Kato, Tadashi Kozuno. MK2 at PBIG Competition: A Prompt Generation Solution. Proceedings of the 2nd Workshop on Agent AI for Scenario Planning (AgentScen 2025), pp. 58–66. paperarXiv
  5. workshop Risa Shinoda, Kuniaki Saito, Shohei Tanaka, Tosho Hirasawa, Yoshitaka Ushiku. SBS Figures: Pre-training Figure QA from Stage-by-Stage Synthesized Images. The AAAI-25 Workshop on Document Understanding and Intelligence. projectarXiv
  6. preprint Shun Inadumi, Shohei Tanaka, Tosho Hirasawa, Atsushi Hashimoto, Koichiro Yoshino, Yoshitaka Ushiku. SciPostGen: Bridging the Gap between Scientific Papers and Poster Layouts. arXiv:2511.22490. arXiv

2024

  1. thesis Tosho Hirasawa. Multimodal Machine Translation as a Resource Scarcity Problem. Ph.D. thesis, Tokyo Metropolitan University.
  2. conference Hwichan Kim, Jun Suzuki, Tosho Hirasawa, Mamoru Komachi. Pruning Multilingual Large Language Models for Multilingual Inference. Findings of the Association for Computational Linguistics: EMNLP 2024. paperarXiv
  3. conference Koki Maeda, Tosho Hirasawa, Atsushi Hashimoto, Jun Harashima, Leszek Rybicki, Yusuke Fukasawa, Yoshitaka Ushiku. COM Kitchens: An Unedited Overhead-view Video Dataset as a Vision-Language Benchmark. European Conference on Computer Vision (ECCV 2024), pp. 123–140. paperarXivcode
  4. conference Ayako Sato, Tosho Hirasawa, Hwichan Kim, Zhousi Chen, Teruaki Oka, Masato Mita, Mamoru Komachi. DejaVu: Disambiguation evaluation dataset for English-JApanese machine translation on VisUal information. Proceedings of the 38th Pacific Asia Conference on Language, Information and Computation (PACLIC 38). paper
  5. conference Kyotaro Nakajima, Hwichan Kim, Tosho Hirasawa, Taisei Enomoto, Zhousi Chen, Mamoru Komachi. A Survey for LLM Tuning Methods: Classifying Approaches Based on Model Internal Accessibility. Proceedings of the 38th Pacific Asia Conference on Language, Information and Computation (PACLIC 38). paper
  6. conference Zhishen Yang, Tosho Hirasawa, Edison Marrese-Taylor, Naoaki Okazaki. Large Language Models as Manga Translators: A Case Study. 言語処理学会第30回年次大会 (NLP2024), 神戸. paper
  7. conference 佐藤郁子, 平澤寅庄, 金輝燦, 岡照晃, 小町守. 語義曖昧性解消に着目した英日マルチモーダル機械翻訳の評価セット構築と分析. 言語処理学会第30回年次大会 (NLP2024), 神戸. 論文
  8. workshop Kazuya Nishimura, Kuniaki Saito, Tosho Hirasawa, Yoshitaka Ushiku. Toward Structured Related Work Generation with Novelty Statements. Proceedings of the Fourth Workshop on Scholarly Document Processing (SDP 2024). paper
  9. workshop Tosho Hirasawa. OSX at Context24: How Well Can GPT Tackle Contexualizing Scientific Figures and Tables. Proceedings of the Fourth Workshop on Scholarly Document Processing (SDP 2024). paper
  10. workshop Taisei Enomoto, Hwichan Kim, Tosho Hirasawa, Yoshinari Nagai, Ayako Sato, Kyotaro Nakajima, Mamoru Komachi. TMU-HIT at MLSP 2024: How Well Can GPT-4 Tackle Multilingual Lexical Simplification? Proceedings of the 19th Workshop on Innovative Use of NLP for Building Educational Applications (BEA 2024). paper

2023

  1. journal Hwichan Kim, Tosho Hirasawa, Sangwhan Moon, Naoaki Okazaki, Mamoru Komachi. North Korean Neural Machine Translation through South Korean Resources. ACM Transactions on Asian and Low-Resource Language Information Processing, Volume 22, Issue 9. paper
  2. conference Tosho Hirasawa, Emanuele Bugliarello, Desmond Elliott, Mamoru Komachi. Visual Prediction Improves Zero-Shot Cross-Modal Machine Translation. Proceedings of the Eighth Conference on Machine Translation (WMT 2023), Singapore. paper
  3. conference Taisei Enomoto, Tosho Hirasawa, Hwichan Kim, Teruaki Oka, Mamoru Komachi. Simultaneous Domain Adaptation of Tokenization and Machine Translation. Proceedings of the 37th Pacific Asia Conference on Language, Information and Computation (PACLIC 37), Hong Kong. paper
  4. conference Hiroto Tamura, Tosho Hirasawa, Hwichan Kim, Mamoru Komachi. Does Masked Language Model Pre-training with Artificial Data Improve Low-resource Neural Machine Translation? Findings of the Association for Computational Linguistics: EACL 2023, Dubrovnik. paper
  5. conference 平澤寅庄, 小町守. 視覚翻訳言語モデルを用いた英日マルチモーダル機械翻訳. 言語処理学会第29回年次大会 (NLP2023), 宜野湾. 論文コード
  6. conference 佐藤郁子, 平澤寅庄, 金輝燦, 岡照晃, 小町守. マルチモーダル機械翻訳における画像・入力文間類似度と翻訳品質の相関の分析. 言語処理学会第29回年次大会 (NLP2023), 宜野湾. 論文
  7. conference 榎本大晟, 平澤寅庄, 金輝燦, 岡照晃, 小町守. ニューラル機械翻訳における単語分割器のドメイン適応. 言語処理学会第29回年次大会 (NLP2023), 宜野湾. 論文
  8. conference 木山朔, 金輝燦, 平澤寅庄, 岡照晃, 小町守. Decoder のみを用いた機械翻訳モデルの分析. 言語処理学会第29回年次大会 (NLP2023), 宜野湾. 論文
  9. workshop 中島京太郎, 金輝燦, 平澤寅庄, 岡照晃, 小町守. 語彙内トークンを媒介とした大規模言語モデルへのソフトプロンプトの転移. 情報処理学会 自然言語処理研究会 第257回研究発表会 (NL257), 東京. 発表会

2022

  1. journal Tosho Hirasawa, Masahiro Kaneko, Aizhan Imankulova, Mamoru Komachi. Pre-Trained Word Embedding and Language Model Improve Multimodal Machine Translation: A Case Study in Multi30K. IEEE Access, Volume 10, pp. 67653–67668. paper
  2. journal Zhishen Yang, Tosho Hirasawa, Naoaki Okazaki, Mamoru Komachi. Why Videos do not Guide Translations in Video-guided Machine Translation? An Empirical Evaluation of Video-guided Machine Translation Dataset. 情報処理学会論文誌 データベース (TOD), 15巻 2号. paper
  3. journal 喜友名朝視顕, 平澤寅庄, 小町守, 小木曽智信. 事前学習モデルを用いた近代文語文のニューラル機械翻訳. 情報処理学会論文誌, 63巻 2号, pp. 269–282. 論文
  4. conference Daisuke Suzuki, Yujin Takahashi, Ikumi Yamashita, Taichi Aida, Tosho Hirasawa, Michitaka Nakatsuji, Masato Mita, Mamoru Komachi. Construction of a Quality Estimation Dataset for Automatic Evaluation of Japanese Grammatical Error Correction. Proceedings of the Thirteenth Language Resources and Evaluation Conference (LREC 2022), Marseille. paperarXiv
  5. conference 田村弘人, 平澤寅庄, 金輝燦, 岡照晃, 小町守. 人工データでの事前学習によるニューラル機械翻訳の性能向上. 言語処理学会第28回年次大会 (NLP2022), オンライン. 論文

2021

  1. workshop Seiichiro Kondo, Aomi Koyama, Tomoshige Kiyuna, Tosho Hirasawa, Mamoru Komachi. Machine Translation with Pre-specified Target-side Words Using a Semi-autoregressive Model. Proceedings of the 8th Workshop on Asian Translation (WAT 2021), pp. 68–73. paper
  2. workshop Seiichiro Kondo, Kengo Hotate, Tosho Hirasawa, Masahiro Kaneko, Mamoru Komachi. Sentence Concatenation Approach to Data Augmentation for Neural Machine Translation. Proceedings of the 2021 Conference of the NAACL: Student Research Workshop (NAACL SRW 2021). paper
  3. workshop Zhishen Yang, Tosho Hirasawa, Naoaki Okazaki, Mamoru Komachi. Do Videos Guide Translations? Evaluation of a Video-Guided Machine Translation dataset. Proceedings of the Workshop on Visually Grounded Interaction and Language (ViGIL 2021). paper
  4. conference 蘆田真奈, 平澤寅庄, 金子正弘, 小町守. 日本語 BERT による否定表現認識についての分析. 言語処理学会第27回年次大会 (NLP2021), オンライン. 論文
  5. conference 喜友名朝視顕, 平澤寅庄, 小町守, 小木曽智信. 事前学習モデルを用いた近代文語文の現代語機械翻訳. 言語処理学会第27回年次大会 (NLP2021), オンライン. 論文
  6. conference 今藤誠一郎, 甫立健悟, 平澤寅庄, 金子正弘, 小町守. 2 文の連結を用いた機械翻訳におけるデータ拡張. 言語処理学会第27回年次大会 (NLP2021), オンライン. 論文

2020

  1. conference Masashi Takaku, Tosho Hirasawa, Mamoru Komachi, Kanako Komiya. Neural Machine Translation from Historical Japanese to Contemporary Japanese Using Diachronically Domain-Adapted Word Embeddings. Proceedings of the 34th Pacific Asia Conference on Language, Information and Computation (PACLIC 34), Hanoi. paper
  2. conference Aizhan Imankulova, Masahiro Kaneko, Tosho Hirasawa, Mamoru Komachi. Towards Multimodal Simultaneous Neural Machine Translation. Proceedings of the Fifth Conference on Machine Translation (WMT 2020). paperarXiv
  3. workshop Hwichan Kim, Tosho Hirasawa, Mamoru Komachi. Zero-shot North Korean to English Neural Machine Translation by Character Tokenization and Phoneme Decomposition. Proceedings of the 58th Annual Meeting of the ACL: Student Research Workshop (ACL SRW 2020), pp. 72–78. paper
  4. workshop Zizheng Zhang, Tosho Hirasawa, Wei Houjing, Masahiro Kaneko, Mamoru Komachi. Translation of New Named Entities from English to Chinese. Proceedings of the 7th Workshop on Asian Translation (WAT 2020), pp. 58–63. paper
  5. workshop Hiroto Tamura, Tosho Hirasawa, Masahiro Kaneko, Mamoru Komachi. TMU Japanese-English Multimodal Machine Translation System for WAT 2020. Proceedings of the 7th Workshop on Asian Translation (WAT 2020), pp. 80–91. paper
  6. workshop Hwichan Kim, Tosho Hirasawa, Mamoru Komachi. Korean-to-Japanese Neural Machine Translation System using Hanja Information. Proceedings of the 7th Workshop on Asian Translation (WAT 2020), pp. 127–134. paper
  7. workshop Masahiro Kaneko, Aizhan Imankulova, Tosho Hirasawa, Mamoru Komachi. English-to-Japanese Diverse Translation by Combining Forward and Backward Outputs. Proceedings of the Fourth Workshop on Neural Generation and Translation (WNGT 2020), pp. 134–138. paper
  8. workshop Tosho Hirasawa, Zhishen Yang, Mamoru Komachi, Naoaki Okazaki. Keyframe Segmentation and Positional Encoding for Video-guided Machine Translation Challenge 2020. The First Workshop on Advances in Language and Vision Research (ALVR 2020). arXivaward
  9. conference 金輝燦, 平澤寅庄, 小町守. 韓国語対訳データを利用した文字分割と音素分解による朝鮮語ニューラル機械翻訳. 言語処理学会第26回年次大会 (NLP2020), オンライン. 論文
  10. conference 高久雅史, 平澤寅庄, 小町守, 古宮嘉那子. 通時的な領域適応を行った単語分散表現を利用した古文から現代文へのニューラル機械翻訳. 言語処理学会第26回年次大会 (NLP2020), オンライン. 論文

2019

  1. conference Tosho Hirasawa, Mamoru Komachi. Debiasing Word Embeddings Improves Multimodal Machine Translation. Proceedings of the 17th Machine Translation Summit (MT Summit 2019), Dublin. paperarXivposterboaster
  2. workshop Tosho Hirasawa, Hayahide Yamagishi, Yukio Matsumura, Mamoru Komachi. Multimodal Machine Translation with Embedding Prediction. Proceedings of the 2019 Conference of the NAACL: Student Research Workshop (NAACL SRW 2019), Minneapolis. paperarXivcodeslides
  3. conference 平澤寅庄, 山岸駿秀, 松村雪桜, 小町守. 事前学習した単語分散表現を利用したマルチモーダル機械翻訳. 言語処理学会第25回年次大会 (NLP2019), 名古屋. 論文ポスター