1.
Yoshinari Wada; Rei Kusunose; Ryoma Shinto; Ayane Matsuzaki; Seiji Adachi; Kota Ando; Tetsuya Asai; Takao Marukame
Live Demonstration: Fingerspelling Recognition in Quantized Neural Networks with Emulated Memristors for 2D/3D Camera Applications Proceedings Article Forthcoming
In: Proceedings of 2025 IEEE Biomedical Circuits and Systems Conference, Abu Dhabi, UAE, Forthcoming.
@inproceedings{ShintoDemo2025,
title = {Live Demonstration: Fingerspelling Recognition in Quantized Neural Networks with Emulated Memristors for 2D/3D Camera Applications},
author = {Yoshinari Wada and Rei Kusunose and Ryoma Shinto and Ayane Matsuzaki and Seiji Adachi and Kota Ando and Tetsuya Asai and Takao Marukame},
year = {2025},
date = {2025-10-16},
booktitle = {Proceedings of 2025 IEEE Biomedical Circuits and Systems Conference, Abu Dhabi, UAE},
keywords = {},
pubstate = {forthcoming},
tppubtype = {inproceedings}
}
2.
佐藤 汰功斗; ジェプカ ラファウ
LLM の危険と安全の認識はどこで 立ち上がるか ーLogit Lens による 成分別可視化ー Technical Report
2025.
@techreport{SatoLAU2025,
title = {LLM の危険と安全の認識はどこで 立ち上がるか ーLogit Lens による 成分別可視化ー},
author = {佐藤 汰功斗 and ジェプカ ラファウ},
year = {2025},
date = {2025-09-28},
urldate = {2025-09-27},
issue = {Summer 2025},
howpublished = {言語獲得と理解研究会 予稿集, pp. 75-82, Summer 2025},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
3.
細野 仁; ジェプカ ラファウ
子供の絵画記述データセットの構築とLLM との比較分析 Technical Report
2025.
@techreport{HosonoLAU2025,
title = {子供の絵画記述データセットの構築とLLM との比較分析},
author = {細野 仁 and ジェプカ ラファウ},
year = {2025},
date = {2025-09-27},
urldate = {2025-09-27},
issue = {Summer 2025},
howpublished = {言語獲得と理解研究会 予稿集, pp. 38-45, Summer 2025},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
4.
Makoto Misaizu; Koshi Watanabe; Keisuke Maeda; Takahiro Ogawa; Rafal Rzepka; Toshihiko Itoh; Miki Haseyama
Rethinking Continual Learning with Pre-Trained Models: Knowledge-Preserving Approach Best Paper Conference
2025 IEEE 14th Global Conference on Consumer Electronics (GCCE 2025), 2025.
@conference{misaizu2025,
title = {Rethinking Continual Learning with Pre-Trained Models: Knowledge-Preserving Approach},
author = {Makoto Misaizu and Koshi Watanabe and Keisuke Maeda and Takahiro Ogawa and Rafal Rzepka and Toshihiko Itoh and Miki Haseyama},
year = {2025},
date = {2025-09-25},
urldate = {2025-09-25},
booktitle = {2025 IEEE 14th Global Conference on Consumer Electronics (GCCE 2025)},
pages = {899--901},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
5.
Makoto Misaizu; Koshi Watanabe; Keisuke Maeda; Takahiro Ogawa; Rafal Rzepka; Toshihiko Itoh; Miki Haseyama
Rethinking Continual Learning with Pre-Trained Models: Knowledge-Preserving Approach Proceedings Article
In: Proceedings of 2025 IEEE 14th Global Conference on Consumer Electronics (GCCE), 2025.
@inproceedings{nokey,
title = {Rethinking Continual Learning with Pre-Trained Models: Knowledge-Preserving Approach},
author = {Makoto Misaizu and Koshi Watanabe and Keisuke Maeda and Takahiro Ogawa and Rafal Rzepka and Toshihiko Itoh and Miki Haseyama},
year = {2025},
date = {2025-09-23},
urldate = {2025-09-23},
booktitle = {Proceedings of 2025 IEEE 14th Global Conference on Consumer Electronics (GCCE)},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
6.
Don Divin Anemeta; Rafal Rzepka
Cross-Cultural Safety Judgments in Child Environments: A Semantic Comparison of Vision-Language Models and Humans Journal Article Forthcoming
In: Algorithms, Forthcoming.
@article{Anemeta:2025,
title = {Cross-Cultural Safety Judgments in Child Environments: A Semantic Comparison of Vision-Language Models and Humans},
author = {Don Divin Anemeta and Rafal Rzepka},
year = {2025},
date = {2025-08-04},
urldate = {2025-08-04},
journal = {Algorithms},
keywords = {},
pubstate = {forthcoming},
tppubtype = {article}
}
7.
山川 宏; 市瀬 龍太郎; 遠藤 太一郎; 加藤 洋平; ジェプカ ラファウ; 林 祐輔
創発機械倫理に向けて Technical Report
no. AGI-030-05, 2025, (第30回 汎用人工知能研究会(SIG-AGI)).
@techreport{025,
title = {創発機械倫理に向けて},
author = {山川 宏 and 市瀬 龍太郎 and 遠藤 太一郎 and 加藤 洋平 and ジェプカ ラファウ and 林 祐輔},
doi = {10.11517/jsaisigtwo.2025.AGI-030_05},
year = {2025},
date = {2025-08-01},
urldate = {2025-08-01},
journal = {人工知能学会第二種研究会資料},
volume = {2025},
number = {AGI-030-05},
organization = {第30回 汎用人工知能研究会(SIG-AGI)},
howpublished = {第30回 汎用人工知能研究会(SIG-AGI)},
note = {第30回 汎用人工知能研究会(SIG-AGI)},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
8.
山川 宏; 市瀬 龍太郎; 遠藤 太一郎; 加藤 洋平; ジェプカ ラファウ; 林 祐輔
SIG-AGI-30th パネル討論「いま創発機械倫理を研究する必要性」 Technical Report
第30回 汎用人工知能研究会(SIG-AGI) no. AGI-030-07, 2025.
@techreport{025b,
title = {SIG-AGI-30th パネル討論「いま創発機械倫理を研究する必要性」},
author = {山川 宏 and 市瀬 龍太郎 and 遠藤 太一郎 and 加藤 洋平 and ジェプカ ラファウ and 林 祐輔},
doi = {10.11517/jsaisigtwo.2025.AGI-030_07},
year = {2025},
date = {2025-08-01},
urldate = {2025-08-01},
booktitle = {人工知能学会第二種研究会資料},
journal = {第30回 汎用人工知能研究会(SIG-AGI)},
volume = {2025},
number = {AGI-030-07},
institution = {第30回 汎用人工知能研究会(SIG-AGI)},
howpublished = {人工知能学会第二種研究会資料},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
9.
Masashi Takeshita; Rafal Rzepka
JETHICS: Japanese Ethics Understanding Evaluation Dataset Miscellaneous
2025.
@misc{takeshita2025jethicsjapaneseethicsunderstanding,
title = {JETHICS: Japanese Ethics Understanding Evaluation Dataset},
author = {Masashi Takeshita and Rafal Rzepka},
url = {https://arxiv.org/abs/2506.16187},
year = {2025},
date = {2025-06-19},
urldate = {2025-06-23},
keywords = {},
pubstate = {published},
tppubtype = {misc}
}
10.
Mariusz Ziółko; Karol Kamiński; Napoleon Waszkiewicz; Wojciech Datka; Karolina Kozłowska; Michał Kucharski; Rafal Rzepka; Bartosz Ziółko
Vocabulary variation and lemma probabilities in speech analysis for diagnosing dementia and depression Journal Article
In: Computers in Biology and Medicine, vol. 194, pp. 110409, 2025, ISSN: 0010-4825.
@article{ZIOLKO2025110409,
title = {Vocabulary variation and lemma probabilities in speech analysis for diagnosing dementia and depression},
author = {Mariusz Ziółko and Karol Kamiński and Napoleon Waszkiewicz and Wojciech Datka and Karolina Kozłowska and Michał Kucharski and Rafal Rzepka and Bartosz Ziółko},
url = {https://www.sciencedirect.com/science/article/pii/S0010482525007607},
doi = {https://doi.org/10.1016/j.compbiomed.2025.110409},
issn = {0010-4825},
year = {2025},
date = {2025-06-08},
urldate = {2025-06-08},
journal = {Computers in Biology and Medicine},
volume = {194},
pages = {110409},
abstract = {Neurodegenerative and mental disorders significantly affect the manner of speaking, syntax, semantics and specific habits of word choice. The aim of our research was to develop linguistic speech analysis methods to provide screening tests in neurology and psychiatry as subjective techniques supporting medical diagnostics. Using features of speech samples recorded by the subjects and the control group, we created classifiers which distinguish one group of recordings from the other. We used two methods to diagnose dementia. The first method is based on the observation that people with dementia have lower vocabulary variations. The second diagnostic method is based on probabilities of lemmas. This method was also used in depression screening tests. For neurodegenerative and mental disorders, linguistic analysis appeared to be satisfactorily effective. Linguistic changes were easily detectable in dementia and less noticeable in depression. We achieved a 95% precision of diagnosis in the control group for dementia and 100% for the three stages of depression. This is higher than in groups of patients (dementia 93% and average 70% for the three stages of depression).},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Neurodegenerative and mental disorders significantly affect the manner of speaking, syntax, semantics and specific habits of word choice. The aim of our research was to develop linguistic speech analysis methods to provide screening tests in neurology and psychiatry as subjective techniques supporting medical diagnostics. Using features of speech samples recorded by the subjects and the control group, we created classifiers which distinguish one group of recordings from the other. We used two methods to diagnose dementia. The first method is based on the observation that people with dementia have lower vocabulary variations. The second diagnostic method is based on probabilities of lemmas. This method was also used in depression screening tests. For neurodegenerative and mental disorders, linguistic analysis appeared to be satisfactorily effective. Linguistic changes were easily detectable in dementia and less noticeable in depression. We achieved a 95% precision of diagnosis in the control group for dementia and 100% for the three stages of depression. This is higher than in groups of patients (dementia 93% and average 70% for the three stages of depression).
11.
倉知 祥太朗; 伊藤 敏彦
実在話者の口調特性を反映したベクトル生成と識別精度の検証 Technical Report
2025, (2025年度 人工知能学会全国大会（第39回）3Win5-29).
@techreport{Kurachi_JSAI2025,
title = {実在話者の口調特性を反映したベクトル生成と識別精度の検証},
author = {倉知 祥太朗 and 伊藤 敏彦},
year = {2025},
date = {2025-05-29},
issue = {3Win5-29]},
abstract = {対話システムにおいて実在の個人性を再現する研究が注目されている．個人性を再現するには，一般的に対象人物の大量な発話データが必要とされるが，その収集は簡単ではない．しかし，信念的な個人性に比べ，口調的な個人性にはある程度パターンが存在すると考えられる． そこで本研究では，少量の発話データから口調のパターンを発見・再現するために，口調ベクトルの獲得と利用を提案する．多数の人物の発話データを基に機械学習を行い，口調の類似性を評価できる固定長の口調ベクトルを獲得する．この口調ベクトルを用いてクラスタリングを行うことで，類似した人物の口調をパターン化した．新規の個人性の再現が必要な場合は，どのパターンに類似するかを推定することでその個人性の再現が可能となる．この方式の有効性を検証するために，クラスタリング結果に対して，生成したベクトルでクラスタ識別を行った．結果，少量のデータでも高いスコアを得ることができた．このことから口調ベクトルを用いて発話者がどの口調パターンに類似しているか推定可能であり，少数発話から口調を再現する可能性を確認できた．},
howpublished = {2025年度 人工知能学会全国大会（第39回）[2Win5-48]},
note = {2025年度 人工知能学会全国大会（第39回）3Win5-29},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
対話システムにおいて実在の個人性を再現する研究が注目されている．個人性を再現するには，一般的に対象人物の大量な発話データが必要とされるが，その収集は簡単ではない．しかし，信念的な個人性に比べ，口調的な個人性にはある程度パターンが存在すると考えられる． そこで本研究では，少量の発話データから口調のパターンを発見・再現するために，口調ベクトルの獲得と利用を提案する．多数の人物の発話データを基に機械学習を行い，口調の類似性を評価できる固定長の口調ベクトルを獲得する．この口調ベクトルを用いてクラスタリングを行うことで，類似した人物の口調をパターン化した．新規の個人性の再現が必要な場合は，どのパターンに類似するかを推定することでその個人性の再現が可能となる．この方式の有効性を検証するために，クラスタリング結果に対して，生成したベクトルでクラスタ識別を行った．結果，少量のデータでも高いスコアを得ることができた．このことから口調ベクトルを用いて発話者がどの口調パターンに類似しているか推定可能であり，少数発話から口調を再現する可能性を確認できた．
12.
Rafal Rzepka
Toward Safer World and Agent Modeling with Natural Semantic Metalanguage-based Controlled Cognition Technical Report
2025, (Proceedings of the 39th Annual Conference of The Japanese Society for Artificial Intelligence, 2E4-OS-4a-04).
@techreport{Rzepka_JSAI2025,
title = {Toward Safer World and Agent Modeling with Natural Semantic Metalanguage-based Controlled Cognition},
author = {Rafal Rzepka},
year = {2025},
date = {2025-05-28},
urldate = {2025-05-28},
issue = {2E4-OS-4a-04},
abstract = {The development of foundation models has shown how large datasets and powerful computing can create useful tools for daily life. However, these models lack agency and fail to ground language in sensory experience. Current methods for building intelligent entities rely on pre-training with massive datasets and using human annotators to teach models safe behavior. This approach is inefficient, unsustainable, and its safety remains uncertain. In this paper, I argue for a shift toward knowledge acquisition methods based on readable cognitive concepts rather than opaque weight-based representations. As the autonomy of physical agents grows with technological advancements, higher-level processing is needed to enable transparent tracking of how an agent’s behavior is designed. I hypothesize that a fixed set of semantic building blocks for perception and learning could improve the explainability of artificial entities. The Natural Semantic Metalanguage framework offers a promising example of how such a set of basic perceptual concepts might be defined.},
howpublished = {Proceedings of the 39th Annual Conference of The Japanese Society for Artificial Intelligence, 2E4-OS-4a-04},
note = {Proceedings of the 39th Annual Conference of The Japanese Society for Artificial Intelligence, 2E4-OS-4a-04},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
The development of foundation models has shown how large datasets and powerful computing can create useful tools for daily life. However, these models lack agency and fail to ground language in sensory experience. Current methods for building intelligent entities rely on pre-training with massive datasets and using human annotators to teach models safe behavior. This approach is inefficient, unsustainable, and its safety remains uncertain. In this paper, I argue for a shift toward knowledge acquisition methods based on readable cognitive concepts rather than opaque weight-based representations. As the autonomy of physical agents grows with technological advancements, higher-level processing is needed to enable transparent tracking of how an agent’s behavior is designed. I hypothesize that a fixed set of semantic building blocks for perception and learning could improve the explainability of artificial entities. The Natural Semantic Metalanguage framework offers a promising example of how such a set of basic perceptual concepts might be defined.
13.
連 慎治; 伊藤 敏彦
ゲームの表情パターンを用いたキャラクターなりきり応答生成の検証 Technical Report
2025, (2025年度 人工知能学会全国大会（第39回）[2Win5-48] ).
@techreport{Muraji_JSAI2025,
title = {ゲームの表情パターンを用いたキャラクターなりきり応答生成の検証},
author = {連 慎治 and 伊藤 敏彦 },
year = {2025},
date = {2025-05-28},
urldate = {2025-05-28},
issue = {[2Win5-48] },
abstract = {近年，創作上の架空の人物（以下，キャラ）になりきって雑談を行うシステムの研究が盛んである．多くのなりきりシステムはテキスト発話のみでの会話を想定しており，キャラらしさの学習や評価に関してもテキストのみから行っている．一方で，人間がキャラのことを知り，キャラらしさを評価する際には，テキスト発話のみからではなく，キャラの言動や表情といった情報も用いており，学習や評価時の設定とギャップが生じている．そこで本研究では，テキスト以外のキャラらしさの情報としてビジュアルノベルで使われるような表情のパターンを用いることで，なりきり雑談システムがよりキャラらしく発話できるようになるかを検証する．検証の結果，表情予測がセリフ生成改良には大きく寄与しないものの表情が見えることでユーザの満足度が上昇することが示された．},
howpublished = {第39回 人工知能学会全国大会 [2Win5-48] },
note = {2025年度 人工知能学会全国大会（第39回）[2Win5-48] },
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
近年，創作上の架空の人物（以下，キャラ）になりきって雑談を行うシステムの研究が盛んである．多くのなりきりシステムはテキスト発話のみでの会話を想定しており，キャラらしさの学習や評価に関してもテキストのみから行っている．一方で，人間がキャラのことを知り，キャラらしさを評価する際には，テキスト発話のみからではなく，キャラの言動や表情といった情報も用いており，学習や評価時の設定とギャップが生じている．そこで本研究では，テキスト以外のキャラらしさの情報としてビジュアルノベルで使われるような表情のパターンを用いることで，なりきり雑談システムがよりキャラらしく発話できるようになるかを検証する．検証の結果，表情予測がセリフ生成改良には大きく寄与しないものの表情が見えることでユーザの満足度が上昇することが示された．
14.
連 慎治; 伊藤 敏彦
なりきり雑談システムを評価するためのキャライメージの評価者間一致に関する検証-４択クイズを用いた原作者とファンの比較- Technical Report
2025, (言語処理学会第31回年次大会(NLP2025)の予稿集 P10-9).
@techreport{MurajiNLP2025,
title = {なりきり雑談システムを評価するためのキャライメージの評価者間一致に関する検証-４択クイズを用いた原作者とファンの比較-},
author = {連 慎治 and 伊藤 敏彦},
url = {https://www.anlp.jp/proceedings/annual_meeting/2025/pdf_dir/P10-9.pdf},
year = {2025},
date = {2025-03-13},
urldate = {2025-03-11},
issue = {P10-9},
note = {言語処理学会第31回年次大会(NLP2025)の予稿集 P10-9},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
15.
進藤 稜真; 竹下 昌志; ジェプカ ラファウ; 伊藤 敏彦
算術タスクを用いた文脈内学習による外挿能力の分析 Technical Report
2025, (言語処理学会第31回年次大会(NLP2025)の予稿集 P2-21).
@techreport{ShintoNLP2025,
title = {算術タスクを用いた文脈内学習による外挿能力の分析},
author = {進藤 稜真 and 竹下 昌志 and ジェプカ ラファウ and 伊藤 敏彦},
url = {https://www.anlp.jp/proceedings/annual_meeting/2025/pdf_dir/P2-21.pdf},
year = {2025},
date = {2025-03-11},
urldate = {2025-03-11},
issue = {Q1-3},
note = {言語処理学会第31回年次大会(NLP2025)の予稿集 P2-21},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
16.
竹下 昌志; ジェプカ ラファウ
JETHICS: 日本語道徳理解度評価用データセット Best Paper Technical Report
2025, (言語処理学会第31回年次大会(NLP2025)の予稿集 Q1-3).
@techreport{JETHICS2025,
title = {JETHICS: 日本語道徳理解度評価用データセット},
author = {竹下 昌志 and ジェプカ ラファウ},
url = {https://www.anlp.jp/proceedings/annual_meeting/2025/pdf_dir/Q1-3.pdf},
year = {2025},
date = {2025-03-11},
urldate = {2025-03-11},
issue = {Q1-3},
note = {言語処理学会第31回年次大会(NLP2025)の予稿集 Q1-3},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
17.
片岸 祥帆; 伊藤 敏彦
雑談におけるユーモア応答に向けた単語置換型漫才利用の検討 Technical Report
2025, (言語理解とコミュニケーション研究会 (NLC-263)).
@techreport{Katagishi2025,
title = {雑談におけるユーモア応答に向けた単語置換型漫才利用の検討},
author = {片岸 祥帆 and 伊藤 敏彦},
year = {2025},
date = {2025-03-08},
urldate = {2025-03-08},
note = {言語理解とコミュニケーション研究会 (NLC-263)},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
18.
Rafal Rzepka; Akihiko Obayashi
Effectiveness of Security Export Control Ontology for Predicting Answer Type and Regulation Categories Proceedings Article
In: Proceedings of the 2024 8th International Conference on Advances in Artificial Intelligence, pp. 156–161, Association for Computing Machinery, New York, NY, USA, 2025, ISBN: 9798400718014.
@inproceedings{ICAAI2024,
title = {Effectiveness of Security Export Control Ontology for Predicting Answer Type and Regulation Categories},
author = {Rafal Rzepka and Akihiko Obayashi},
url = {https://doi.org/10.1145/3704137.3704180},
doi = {10.1145/3704137.3704180},
isbn = {9798400718014},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
booktitle = {Proceedings of the 2024 8th International Conference on Advances in Artificial Intelligence},
pages = {156–161},
publisher = {Association for Computing Machinery},
address = {New York, NY, USA},
series = {ICAAI '24},
abstract = {In this paper we present results of our experiments investigating if an expert knowledge graph can improve Large Language Models accuracy in predicting correct answer labels and regulations related to the topic of security export control. As the lack of related data prevents machine-learning or fine-tuning approaches, we implement prompt expansion by searching most relevant nodes of the graph and adding the expanded context to the prompt. Results of our experiments show that the addition improved answer type selection but clearly hamper the capability of finding a correct regulation category.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
In this paper we present results of our experiments investigating if an expert knowledge graph can improve Large Language Models accuracy in predicting correct answer labels and regulations related to the topic of security export control. As the lack of related data prevents machine-learning or fine-tuning approaches, we implement prompt expansion by searching most relevant nodes of the graph and adding the expanded context to the prompt. Results of our experiments show that the addition improved answer type selection but clearly hamper the capability of finding a correct regulation category.
19.
Rafal Rzepka; Kei Okada
Simulating Perception With LLMs as Underpinnings for More Controllable Knowledge Acquisition Technical Report
no. AGI-028, 2024, (28th Symposium of JSAI Special Interest Group for Artificial General Intelligence).
@techreport{RzepkaSIGAGI2024,
title = {Simulating Perception With LLMs as Underpinnings for More Controllable Knowledge Acquisition},
author = {Rafal Rzepka and Kei Okada},
doi = {10.11517/jsaisigtwo.2024.AGI-028_07},
year = {2024},
date = {2024-12-21},
urldate = {2024-12-21},
journal = {人工知能学会第二種研究会資料},
volume = {2024},
number = {AGI-028},
pages = {07},
note = {28th Symposium of JSAI Special Interest Group for Artificial General Intelligence},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
20.
橋本 亮; ジェプカ ラファウ
機械翻訳及び知識蒸留を用いた学習データの有効性 Technical Report
2024, (人工知能学会 第133回知識ベースシステム研究会).
@techreport{Hashimoto_24,
title = {機械翻訳及び知識蒸留を用いた学習データの有効性},
author = {橋本 亮 and ジェプカ ラファウ},
doi = {10.11517/jsaikbs.133.0_89},
year = {2024},
date = {2024-12-20},
urldate = {2024-12-20},
journal = {人工知能学会研究会資料 知識ベースシステム研究会},
volume = {133},
pages = {89-94},
note = {人工知能学会 第133回知識ベースシステム研究会},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
21.
Masashi Takeshita; Rafal Rzepka
Speciesism in natural language processing research Journal Article
In: AI and Ethics, 2024, ISBN: 2730-5961.
@article{cite-key,
title = {Speciesism in natural language processing research},
author = {Masashi Takeshita and Rafal Rzepka},
url = {https://doi.org/10.1007/s43681-024-00606-3},
doi = {10.1007/s43681-024-00606-3},
isbn = {2730-5961},
year = {2024},
date = {2024-11-27},
journal = {AI and Ethics},
abstract = {Natural Language Processing (NLP) research on AI Safety and social bias in AI has focused on safety for humans and social bias against human minorities. However, some AI ethicists have argued that the moral significance of nonhuman animals has been ignored in AI research. Therefore, the purpose of this study is to investigate whether there is speciesism, i.e., discrimination against nonhuman animals, in NLP research. First, we explain why nonhuman animals are relevant in NLP research. Next, we survey the findings of existing research on speciesism in NLP researchers, data, and models and further investigate this problem in this study. The findings of this study suggest that speciesism exists within researchers, data, and models, respectively. Specifically, our survey and experiments show that (a) among NLP researchers, even those who study social bias in AI, do not recognize speciesism or speciesist bias; (b) among NLP data, speciesist bias is inherent in the data annotated in the datasets used to evaluate NLP models; (c) OpenAI GPTs, recent NLP models, exhibit speciesist bias by default. Finally, we discuss how we can reduce speciesism in NLP research.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Natural Language Processing (NLP) research on AI Safety and social bias in AI has focused on safety for humans and social bias against human minorities. However, some AI ethicists have argued that the moral significance of nonhuman animals has been ignored in AI research. Therefore, the purpose of this study is to investigate whether there is speciesism, i.e., discrimination against nonhuman animals, in NLP research. First, we explain why nonhuman animals are relevant in NLP research. Next, we survey the findings of existing research on speciesism in NLP researchers, data, and models and further investigate this problem in this study. The findings of this study suggest that speciesism exists within researchers, data, and models, respectively. Specifically, our survey and experiments show that (a) among NLP researchers, even those who study social bias in AI, do not recognize speciesism or speciesist bias; (b) among NLP data, speciesist bias is inherent in the data annotated in the datasets used to evaluate NLP models; (c) OpenAI GPTs, recent NLP models, exhibit speciesist bias by default. Finally, we discuss how we can reduce speciesism in NLP research.
22.
Shinji Muraji; Rafal Rzepka; Toshihiko Itoh
Evaluation of active generation of interlocutor profiling sentences from utterances and their implicit context Proceedings Article
In: Rzepka, Rafal; Ptaszynski, Michal; Dybala, Pawel; Chung, Siaw-Fong; Vallverdu, Jordi (Ed.): Proceedings of the 9th Linguistic and Cognitive Approaches To Dialog Agents Workshop – LaCATODA 2024 co-located with the Pacific Rim International Conference on Artificial Intelligence (PRICAI 2024), pp. 45-55, WS-CEUR, 2024.
@inproceedings{Muraji_LACATODA2024,
title = {Evaluation of active generation of interlocutor profiling sentences from utterances and their implicit context},
author = {Shinji Muraji and Rafal Rzepka and Toshihiko Itoh},
editor = {Rafal Rzepka and Michal Ptaszynski and Pawel Dybala and Siaw-Fong Chung and Jordi Vallverdu},
year = {2024},
date = {2024-11-19},
urldate = {2024-11-19},
booktitle = {Proceedings of the 9th Linguistic and Cognitive Approaches To Dialog Agents Workshop - LaCATODA 2024 co-located with the Pacific Rim International Conference on Artificial Intelligence (PRICAI 2024)},
pages = {45-55},
publisher = {WS-CEUR},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
23.
Huizhong Ji; Rafal Rzepka
Fame Bias — Large Language Models Change Their Judgement Depending on Personal Name Proceedings Article
In: Patricia Anthony Rafik Hadfi, Alok Sharma (Ed.): PRICAI 2024: Trends in Artificial Intelligence, Lecture Notes LNAI 15283, 21st Pacific Rim International Conference on Artificial Intelligence, PRICAI 2024, Kyoto, Japan, November 18–24, 2024, Proceedings, Part III, pp. 15–20, 2024.
@inproceedings{Ji2024pricai,
title = {Fame Bias -- Large Language Models Change Their Judgement Depending on Personal Name},
author = {Huizhong Ji and Rafal Rzepka},
editor = {Rafik Hadfi, Patricia Anthony, Alok Sharma, Takayuki Ito, Quan Bai},
doi = {https://doi.org/10.1007/978-981-96-0122-6},
year = {2024},
date = {2024-11-18},
urldate = {2024-11-18},
booktitle = {PRICAI 2024: Trends in Artificial Intelligence, Lecture Notes LNAI 15283, 21st Pacific Rim International Conference on Artificial Intelligence, PRICAI 2024, Kyoto, Japan, November 18–24, 2024, Proceedings, Part III},
pages = {15--20},
series = {Lecture Notes in Computer Science},
abstract = {This paper focuses on bias caused by changing personal names in the input of proprietary LLMs, fine-tuned language models, and lexicon-based sentiment analysis tool. It extracts examples from NLI and story cloze task datasets with personal names, replaces them with names of celebrities, ordinary people or mixtures. Results show that altered names significantly influence model assessments, especially with more than two choices, indicating potential real-world application hazards. A mitigation method for “fame bias” is proposed but the problem requires further research.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
This paper focuses on bias caused by changing personal names in the input of proprietary LLMs, fine-tuned language models, and lexicon-based sentiment analysis tool. It extracts examples from NLI and story cloze task datasets with personal names, replaces them with names of celebrities, ordinary people or mixtures. Results show that altered names significantly influence model assessments, especially with more than two choices, indicating potential real-world application hazards. A mitigation method for “fame bias” is proposed but the problem requires further research.
24.
會田 尚平; 伊藤 敏彦
改稿機構を含む階層処理による物語あらすじ生成手法 Technical Report
2024, (令和6年度 電気・情報関係学会北海道支部連合大会).
@techreport{Aida2024,
title = {改稿機構を含む階層処理による物語あらすじ生成手法},
author = {會田 尚平 and 伊藤 敏彦},
editor = {令和6年度 電気・情報関係学会北海道支部連合大会},
year = {2024},
date = {2024-11-02},
urldate = {2024-11-02},
booktitle = {令和6年度 電気・情報関係学会北海道支部連合大会},
issuetitle = {令和6年度 電気・情報関係学会北海道支部連合大会},
journal = {令和6年度 電気・情報関係学会北海道支部連合大会},
howpublished = {令和6年度 電気・情報関係学会北海道支部連合大会},
note = {令和6年度 電気・情報関係学会北海道支部連合大会},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
25.
ジェプカ ラファウ
AIが言葉を理解する時代へ：大規模言語モデルの光と影 Presentation
eシルクロード大学 （招待講演）, 24.10.2024.
@misc{esilk2024Rzepka,
title = {AIが言葉を理解する時代へ：大規模言語モデルの光と影},
author = {ジェプカ ラファウ },
year = {2024},
date = {2024-10-24},
howpublished = {eシルクロード大学 （招待講演）},
keywords = {},
pubstate = {published},
tppubtype = {presentation}
}
26.
Don Divin Anemeta; Rafal Rzepka
Cross-Cultural Perception of Child Safety: A Dataset of Annotated Hazards in Everyday Environments Technical Report
2024, (Language Acquisition and Understanding Symposium 2024).
@techreport{Anemeta2024LAU,
title = {Cross-Cultural Perception of Child Safety: A Dataset of Annotated Hazards in Everyday Environments},
author = {Don Divin Anemeta and Rafal Rzepka},
year = {2024},
date = {2024-09-28},
note = {Language Acquisition and Understanding Symposium 2024},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
27.
Rafal Rzepka
人工知能とことばの理解〜AIはことばをどのように「理解」してきたのか〜 Presentation
北大道新アカデミー 招待講義, 14.09.2024.
@misc{Rzepka2024Doushin,
title = {人工知能とことばの理解〜AIはことばをどのように「理解」してきたのか〜},
author = {Rafal Rzepka },
year = {2024},
date = {2024-09-14},
urldate = {2024-09-14},
howpublished = {北大道新アカデミー 招待講義},
keywords = {},
pubstate = {published},
tppubtype = {presentation}
}
28.
Rafal Rzepka
NLPの多様な応用と可能性：現実的な視点からの考察 Presentation
第10回北大・部局横断シンポジウム（招待発表）, 06.09.2024.
@misc{RzepkaBOS10,
title = {NLPの多様な応用と可能性：現実的な視点からの考察},
author = {Rafal Rzepka},
year = {2024},
date = {2024-09-06},
urldate = {2024-09-06},
howpublished = {第10回北大・部局横断シンポジウム（招待発表）},
keywords = {},
pubstate = {published},
tppubtype = {presentation}
}
29.
進藤 稜真; 竹下 昌志; ジェプカ ラファウ; 伊藤 敏彦
LLMはなぜ算数が苦手なのか？ Transformerの外挿能力に関する分析 Best Paper Technical Report
2024, (第19回YANSシンポジウム [奨励賞] [株式会社オルツ賞] [ストックマーク株式会社賞]).
@techreport{Shinto2024YANS,
title = {LLMはなぜ算数が苦手なのか？ Transformerの外挿能力に関する分析},
author = {進藤 稜真 and 竹下 昌志 and ジェプカ ラファウ and 伊藤 敏彦},
year = {2024},
date = {2024-09-04},
urldate = {2024-09-04},
note = {第19回YANSシンポジウム [奨励賞] [株式会社オルツ賞] [ストックマーク株式会社賞]},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
30.
竹下 昌志; ジェプカ ラファウ
言語モデルの日本語道徳理解能力の評価データセットの構築 Best Paper Technical Report
2024, (第19回シンポジウム [奨励賞] ).
@techreport{TakeshitaYANS2024,
title = {言語モデルの日本語道徳理解能力の評価データセットの構築},
author = {竹下 昌志 and ジェプカ ラファウ},
year = {2024},
date = {2024-09-04},
urldate = {2024-09-04},
note = {第19回シンポジウム [奨励賞] },
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
31.
Rafal Rzepka
人工知能の入門 Presentation
中央地区福祉のまち推進センター おしゃべりサロン6年度第9回 招待講義, 19.08.2024.
@misc{RzepkaSalon2024,
title = {人工知能の入門},
author = {Rafal Rzepka},
year = {2024},
date = {2024-08-19},
howpublished = {中央地区福祉のまち推進センター おしゃべりサロン6年度第9回 招待講義},
keywords = {},
pubstate = {published},
tppubtype = {presentation}
}
32.
Rafal Rzepka; Ryoma Shinto; Kenji Araki
Semantic Primes-Inspired Tacit Knowledge Dataset for Simulating Basic Perception Capabilities of Cognitive Architectures Proceedings Article
In: Artificial General Intelligence: 17th International Conference, AGI 2024, Seattle, WA, USA, August 13–16, 2024, Proceedings, pp. 145–154, Springer-Verlag, SEATTLE, WA, USA, 2024, ISBN: 978-3-031-65571-5.
@inproceedings{Rzepka2024AGI,
title = {Semantic Primes-Inspired Tacit Knowledge Dataset for Simulating Basic Perception Capabilities of Cognitive Architectures},
author = {Rafal Rzepka and Ryoma Shinto and Kenji Araki},
url = {https://doi.org/10.1007/978-3-031-65572-2_16},
doi = {10.1007/978-3-031-65572-2_16},
isbn = {978-3-031-65571-5},
year = {2024},
date = {2024-08-15},
urldate = {2024-01-01},
booktitle = {Artificial General Intelligence: 17th International Conference, AGI 2024, Seattle, WA, USA, August 13–16, 2024, Proceedings},
pages = {145–154},
publisher = {Springer-Verlag},
address = {SEATTLE, WA, USA},
abstract = {In this paper we present a novel dataset of tacit knowledge represented in natural language (Japanese) inspired by semantic primes categories. The main goals of this data is to a) allow investigations regarding influence of perception data in various cognitive tasks, b) mimic signals for cognitive processes of an artificial agent to extend the understanding of the world and c) testing cognitive capabilities of intelligent instances like foundation models. We describe the dataset and share results of preliminary experiments showing that the tacit knowledge recognition is still hard for language models. We also discuss how such redirecting neural approaches to cognition only and then perform reasoning in a symbolic realms could become beneficial for new type of simulations before AGIs are equipped with more sophisticated sensory apparatus.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
In this paper we present a novel dataset of tacit knowledge represented in natural language (Japanese) inspired by semantic primes categories. The main goals of this data is to a) allow investigations regarding influence of perception data in various cognitive tasks, b) mimic signals for cognitive processes of an artificial agent to extend the understanding of the world and c) testing cognitive capabilities of intelligent instances like foundation models. We describe the dataset and share results of preliminary experiments showing that the tacit knowledge recognition is still hard for language models. We also discuss how such redirecting neural approaches to cognition only and then perform reasoning in a symbolic realms could become beneficial for new type of simulations before AGIs are equipped with more sophisticated sensory apparatus.
33.
竹下昌志
長期主義と人間以外――パンデミックの例を通してＡＩと非ヒト動物について考える Journal Article
In: 現代思想, vol. 52, no. 11, 2024, (Invited paper).
@article{takeshita2024longtermism,
title = {長期主義と人間以外――パンデミックの例を通してＡＩと非ヒト動物について考える},
author = {竹下昌志},
year = {2024},
date = {2024-07-15},
urldate = {2024-01-01},
journal = {現代思想},
volume = {52},
number = {11},
note = {Invited paper},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
34.
Rafal Rzepka
AIの能力、限界、社会への潜在的な影響 ～専門家でも意見が分かれる人工知能の今～ Presentation
北大祭2024公開講義, 09.06.2024.
@misc{HokudaisaiRzepka2024,
title = {AIの能力、限界、社会への潜在的な影響 ～専門家でも意見が分かれる人工知能の今～},
author = {Rafal Rzepka},
year = {2024},
date = {2024-06-09},
howpublished = {北大祭2024公開講義},
keywords = {},
pubstate = {published},
tppubtype = {presentation}
}
35.
竹下昌志
動物実験、功利主義、生産-消費ギャップ Technical Report
豊田工業大学人文科学研究室 no. 32, 2024, ISSN: 2432-7921.
@techreport{takeshita2024animal,
title = {動物実験、功利主義、生産-消費ギャップ},
author = {竹下昌志},
doi = {10.60327/ttidiscussionpaper.32.0_1},
issn = {2432-7921},
year = {2024},
date = {2024-03-11},
urldate = {2024-03-11},
number = {32},
pages = {1–17},
institution = {豊田工業大学人文科学研究室},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
36.
兪 東根; 伊藤 敏彦
雑談を用いた精神健康支援システムにおける情報獲得に自然な話題転換と情報獲得タイミングの検知に関する研究 Technical Report
兵庫, no. 416, 2024, (2024年3月10日(日)-3月11日(月) 三宮コンベンションセンター (NLC, IPSJ-NL)).
@techreport{yu2024nlc,
title = {雑談を用いた精神健康支援システムにおける情報獲得に自然な話題転換と情報獲得タイミングの検知に関する研究},
author = {兪 東根 and 伊藤 敏彦},
year = {2024},
date = {2024-03-10},
urldate = {2024-03-10},
booktitle = {信学技報 NLC2023-24},
volume = {123},
number = {416},
pages = {7-12},
address = {兵庫},
series = {NLC2023-24},
note = {2024年3月10日(日)-3月11日(月) 三宮コンベンションセンター (NLC, IPSJ-NL)},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
37.
Rafal Rzepka; Kacper Dudzic; Arisa Abe; Kenji Araki
DanSto — Japanese Dataset of Short Stories for Evaluating Context Understanding Technical Report
2024, (Technical Report of JSAI Special Interest Group for Artificial General Intelligence SIG-AGI-026-06).
@techreport{Dansto2024,
title = {DanSto -- Japanese Dataset of Short Stories for Evaluating Context Understanding},
author = {Rafal Rzepka and Kacper Dudzic and Arisa Abe and Kenji Araki},
url = {https://doi.org/10.11517/jsaisigtwo.2023.AGI-026_32},
doi = {10.11517/jsaisigtwo.2023.AGI-026_32},
year = {2024},
date = {2024-03-08},
urldate = {2024-03-08},
issue = {人工知能学会第二種研究会資料 汎用人工知能研究会 SIG-AGI-026-06},
abstract = {This paper introduces a dataset comprising more than 8,000 manually crafted short stories in the Japanese language. The primary objectives of this dataset encompass addressing the dearth of comparable data in Japanese. Additionally, the dataset provides alternative endings for the narratives through crowdsourcing, ensuring they remain both plausible and marginally less probable than the original ones. This approach contributes to the creation of a testing benchmark that poses heightened challenges for contemporary large language models, particularly when contrasted with analogous benchmarks in English where conclusions are typically dichotomized into correct and incorrect endings. The dataset is further expanded through automated manipulation of subjects and objects, and the study evaluates the performance of popular models across three key tasks: a) predicting story endings, b) substituting antonyms, and c) swapping nouns. Preliminary experiments show that zero-shot GPT-4 capabilities are relatively high, especially in case of recognizing sentences with swapped nouns (94% accuracy) while open-source Japanese LLMs struggle with processing proposed stories.},
howpublished = {Technical Report of JSAI Special Interest Group for Artificial General Intelligence SIG-AGI-026-06},
note = {Technical Report of JSAI Special Interest Group for Artificial General Intelligence SIG-AGI-026-06},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
This paper introduces a dataset comprising more than 8,000 manually crafted short stories in the Japanese language. The primary objectives of this dataset encompass addressing the dearth of comparable data in Japanese. Additionally, the dataset provides alternative endings for the narratives through crowdsourcing, ensuring they remain both plausible and marginally less probable than the original ones. This approach contributes to the creation of a testing benchmark that poses heightened challenges for contemporary large language models, particularly when contrasted with analogous benchmarks in English where conclusions are typically dichotomized into correct and incorrect endings. The dataset is further expanded through automated manipulation of subjects and objects, and the study evaluates the performance of popular models across three key tasks: a) predicting story endings, b) substituting antonyms, and c) swapping nouns. Preliminary experiments show that zero-shot GPT-4 capabilities are relatively high, especially in case of recognizing sentences with swapped nouns (94% accuracy) while open-source Japanese LLMs struggle with processing proposed stories.
38.
Rafal Rzepka; Kenji Araki
In: Wu, Peggy; Salpukas, Michael; Wu, Hsin-Fu; Ellsworth, Shannon (Ed.): Book Title, pp. 209-223, Academic Press, 2024, ISBN: 9780443159916.
@incollection{rzepka2024obtaining,
title = {Chapter Twelve - Obtaining hints to understand language model-based moral decision making by generating consequences of acts},
author = {Rafal Rzepka and Kenji Araki},
editor = {Peggy Wu and Michael Salpukas and Hsin-Fu Wu and Shannon Ellsworth},
url = {https://doi.org/10.1016/B978-0-44-315991-6.00018-2},
isbn = {9780443159916},
year = {2024},
date = {2024-02-09},
urldate = {2024-02-09},
booktitle = {Book Title},
pages = {209-223},
publisher = {Academic Press},
keywords = {},
pubstate = {published},
tppubtype = {incollection}
}
39.
荒木 健治
ChatGPTの仕組みと最新動向および教育現場での使用方法について Presentation
令和5年度北海道国⽴⼤学機構三⼤学FDSD研修会, 19.01.2024.
@misc{Araki2024ChatGPT,
title = {ChatGPTの仕組みと最新動向および教育現場での使用方法について},
author = {荒木 健治},
year = {2024},
date = {2024-01-19},
urldate = {2024-01-19},
howpublished = {令和5年度北海道国⽴⼤学機構三⼤学FDSD研修会},
keywords = {},
pubstate = {published},
tppubtype = {presentation}
}
40.
Akiko Aizawa; Eiji Aramaki; Bowen Chen; Fei Cheng; Hiroyuki Deguchi; Rintaro Enomoto; Kazuki Fujii; Kensuke Fukumoto; Takuya Fukushima; Namgi Han; Yuto Harada; Chikara Hashimoto; Tatsuya Hiraoka; Shohei Hisada; Sosuke Hosokawa; Lu Jie; Keisuke Kamata; Teruhito Kanazawa; Hiroki Kanezashi; Hiroshi Kataoka; Satoru Katsumata; Daisuke Kawahara; Seiya Kawano; Atsushi Keyaki; Keisuke Kiryu; Hirokazu Kiyomaru; Takashi Kodama; Takahiro Kubo; Yohei Kuga; Ryoma Kumon; Shuhei Kurita; Sadao Kurohashi; Conglong Li; Taiki Maekawa; Hiroshi Matsuda; Yusuke Miyao; Kentaro Mizuki; Sakae Mizuki; Yugo Murawaki; Ryo Nakamura; Taishi Nakamura; Kouta Nakayama; Tomoka Nakazato; Takuro Niitsuma; Jiro Nishitoba; Yusuke Oda; Hayato Ogawa; Takumi Okamoto; Naoaki Okazaki; Yohei Oseki; Shintaro Ozaki; Koki Ryu; Rafal Rzepka; Keisuke Sakaguchi; Shota Sasaki; Satoshi Sekine; Kohei Suda; Saku Sugawara; Issa Sugiura; Hiroaki Sugiyama; Hisami Suzuki; Jun Suzuki; Toyotaro Suzumura; Kensuke Tachibana; Yu Takagi; Kyosuke Takami; Koichi Takeda; Masashi Takeshita; Masahiro Tanaka; Kenjiro Taura; Arseny Tolmachev; Nobuhiro Ueda; Zhen Wan; Shuntaro Yada; Sakiko Yahata; Yuya Yamamoto; Yusuke Yamauchi; Hitomi Yanaka; Rio Yokota; Koichiro Yoshino
LLM-jp: A Cross-organizational Project for the Research and Development of Fully Open Japanese LLMs Miscellaneous
2024.
@misc{llmjp2024llmjpcrossorganizationalprojectresearch,
title = {LLM-jp: A Cross-organizational Project for the Research and Development of Fully Open Japanese LLMs},
author = {Akiko Aizawa and Eiji Aramaki and Bowen Chen and Fei Cheng and Hiroyuki Deguchi and Rintaro Enomoto and Kazuki Fujii and Kensuke Fukumoto and Takuya Fukushima and Namgi Han and Yuto Harada and Chikara Hashimoto and Tatsuya Hiraoka and Shohei Hisada and Sosuke Hosokawa and Lu Jie and Keisuke Kamata and Teruhito Kanazawa and Hiroki Kanezashi and Hiroshi Kataoka and Satoru Katsumata and Daisuke Kawahara and Seiya Kawano and Atsushi Keyaki and Keisuke Kiryu and Hirokazu Kiyomaru and Takashi Kodama and Takahiro Kubo and Yohei Kuga and Ryoma Kumon and Shuhei Kurita and Sadao Kurohashi and Conglong Li and Taiki Maekawa and Hiroshi Matsuda and Yusuke Miyao and Kentaro Mizuki and Sakae Mizuki and Yugo Murawaki and Ryo Nakamura and Taishi Nakamura and Kouta Nakayama and Tomoka Nakazato and Takuro Niitsuma and Jiro Nishitoba and Yusuke Oda and Hayato Ogawa and Takumi Okamoto and Naoaki Okazaki and Yohei Oseki and Shintaro Ozaki and Koki Ryu and Rafal Rzepka and Keisuke Sakaguchi and Shota Sasaki and Satoshi Sekine and Kohei Suda and Saku Sugawara and Issa Sugiura and Hiroaki Sugiyama and Hisami Suzuki and Jun Suzuki and Toyotaro Suzumura and Kensuke Tachibana and Yu Takagi and Kyosuke Takami and Koichi Takeda and Masashi Takeshita and Masahiro Tanaka and Kenjiro Taura and Arseny Tolmachev and Nobuhiro Ueda and Zhen Wan and Shuntaro Yada and Sakiko Yahata and Yuya Yamamoto and Yusuke Yamauchi and Hitomi Yanaka and Rio Yokota and Koichiro Yoshino},
url = {https://arxiv.org/abs/2407.03963},
year = {2024},
date = {2024-01-01},
urldate = {2024-01-01},
keywords = {},
pubstate = {published},
tppubtype = {misc}
}