Showing 1–10 of 10 results for author: Horie, T

On Fairness of Task Arithmetic: The Role of Task Vectors

Authors: Hiroki Naganuma, Kotaro Yoshida, Laura Gomezjurado Gonzalez, Takafumi Horie, Yuji Naraki, Ryotaro Shimizu

Abstract: Model editing techniques, particularly task arithmetic using task vectors, have shown promise in efficiently modifying pre-trained models through arithmetic operations like task addition and negation. Despite computational advantages, these methods may inadvertently affect model fairness, creating risks in sensitive applications like hate speech detection. However, the fairness implications of tas… ▽ More Model editing techniques, particularly task arithmetic using task vectors, have shown promise in efficiently modifying pre-trained models through arithmetic operations like task addition and negation. Despite computational advantages, these methods may inadvertently affect model fairness, creating risks in sensitive applications like hate speech detection. However, the fairness implications of task arithmetic remain largely unexplored, presenting a critical gap in the existing literature. We systematically examine how manipulating task vectors affects fairness metrics, including Demographic Parity and Equalized Odds. To rigorously assess these effects, we benchmark task arithmetic against full fine-tuning, a costly but widely used baseline, and Low-Rank Adaptation (LoRA), a prevalent parameter-efficient fine-tuning method. Additionally, we explore merging task vectors from models fine-tuned on demographic subgroups vulnerable to hate speech, investigating whether fairness outcomes can be controlled by adjusting task vector coefficients, potentially enabling tailored model behavior. Our results offer novel insights into the fairness implications of model editing and establish a foundation for fairness-aware and responsible model editing practices. △ Less

Submitted 30 May, 2025; originally announced May 2025.

Augmenting NER Datasets with LLMs: Towards Automated and Refined Annotation

Authors: Yuji Naraki, Ryosuke Yamaki, Yoshikazu Ikeda, Takafumi Horie, Kotaro Yoshida, Ryotaro Shimizu, Hiroki Naganuma

Abstract: In the field of Natural Language Processing (NLP), Named Entity Recognition (NER) is recognized as a critical technology, employed across a wide array of applications. Traditional methodologies for annotating datasets for NER models are challenged by high costs and variations in dataset quality. This research introduces a novel hybrid annotation approach that synergizes human effort with the capab… ▽ More In the field of Natural Language Processing (NLP), Named Entity Recognition (NER) is recognized as a critical technology, employed across a wide array of applications. Traditional methodologies for annotating datasets for NER models are challenged by high costs and variations in dataset quality. This research introduces a novel hybrid annotation approach that synergizes human effort with the capabilities of Large Language Models (LLMs). This approach not only aims to ameliorate the noise inherent in manual annotations, such as omissions, thereby enhancing the performance of NER models, but also achieves this in a cost-effective manner. Additionally, by employing a label mixing strategy, it addresses the issue of class imbalance encountered in LLM-based annotations. Through an analysis across multiple datasets, this method has been consistently shown to provide superior performance compared to traditional annotation methods, even under constrained budget conditions. This study illuminates the potential of leveraging LLMs to improve dataset quality, introduces a novel technique to mitigate class imbalances, and demonstrates the feasibility of achieving high-performance NER in a cost-effective way. △ Less

Submitted 30 December, 2024; v1 submitted 30 March, 2024; originally announced April 2024.

Symbol emergence as interpersonal cross-situational learning: the emergence of lexical knowledge with combinatoriality

Authors: Yoshinobu Hagiwara, Kazuma Furukawa, Takafumi Horie, Akira Taniguchi, Tadahiro Taniguchi

Abstract: We present a computational model for a symbol emergence system that enables the emergence of lexical knowledge with combinatoriality among agents through a Metropolis-Hastings naming game and cross-situational learning. Many computational models have been proposed to investigate combinatoriality in emergent communication and symbol emergence in cognitive and developmental robotics. However, existi… ▽ More We present a computational model for a symbol emergence system that enables the emergence of lexical knowledge with combinatoriality among agents through a Metropolis-Hastings naming game and cross-situational learning. Many computational models have been proposed to investigate combinatoriality in emergent communication and symbol emergence in cognitive and developmental robotics. However, existing models do not sufficiently address category formation based on sensory-motor information and semiotic communication through the exchange of word sequences within a single integrated model. Our proposed model facilitates the emergence of lexical knowledge with combinatoriality by performing category formation using multimodal sensory-motor information and enabling semiotic communication through the exchange of word sequences among agents in a unified model. Furthermore, the model enables an agent to predict sensory-motor information for unobserved situations by combining words associated with categories in each modality. We conducted two experiments with two humanoid robots in a simulated environment to evaluate our proposed model. The results demonstrated that the agents can acquire lexical knowledge with combinatoriality through interpersonal cross-situational learning based on the Metropolis-Hastings naming game and cross-situational learning. Furthermore, our results indicate that the lexical knowledge developed using our proposed model exhibits generalization performance for novel situations through interpersonal cross-modal inference. △ Less

Submitted 27 June, 2023; originally announced June 2023.

Survey of Period Variations of Superhumps in SU UMa-Type Dwarf Novae. IX: The Ninth Year (2016-2017)

Authors: Taichi Kato, Keisuke Isogai, Franz-Josef Hambsch, Tonny Vanmunster, Hiroshi Itoh, Berto Monard, Tamaas Tordai, Mariko Kimura, Yasuyuki Wakamatsu, Seiichiro Kiyota, Ian Miller, Peter Starr, Kiyoshi Kasai, Sergey Yu. Shugarov, Drahomir Chochol, Natalia Katysheva, Anna M. Zaostrojnykh, Matej Sekeras, Yuliana G. Kuznyetsova, Eugenia S. Kalinicheva, Polina Golysheva, Viktoriia Krushevska, Yutaka Maeda, Pavol A. Dubovsky, Igor Kudzej , et al. (54 additional authors not shown)

Abstract: Continuing the project described by Kato et al. (2009, arXiv:0905.1757), we collected times of superhump maxima for 127 SU UMa-type dwarf novae observed mainly during the 2016--2017 season and characterized these objects. We provide updated statistics of relation between the orbital period and the variation of superhumps, the relation between period variations and the rebrightening type in WZ Sge-… ▽ More Continuing the project described by Kato et al. (2009, arXiv:0905.1757), we collected times of superhump maxima for 127 SU UMa-type dwarf novae observed mainly during the 2016--2017 season and characterized these objects. We provide updated statistics of relation between the orbital period and the variation of superhumps, the relation between period variations and the rebrightening type in WZ Sge-type objects. We obtained the period minimum of 0.05290(2)d and confirmed the presence of the period gap above the orbital period ~0.09d. We note that four objects (NY Her, 1RXS J161659.5+620014, CRTS J033349.8-282244 and SDSS J153015.04+094946.3) have supercycles shorter than 100d but show infrequent normal outbursts. We consider that these objects are similar to V503 Cyg, whose normal outbursts are likely suppressed by a disk tilt. These four objects are excellent candidates to search for negative superhumps. DDE 48 appears to be a member of ER UMa-type dwarf novae. We identified a new eclipsing SU UMa-type object MASTER OT J220559.40-341434.9. We observed 21 WZ Sge-type dwarf novae during this interval and reported 18 out of them in this paper. Among them, ASASSN-16js is a good candidate for a period bouncer. ASASSN-16ia showed a precursor outburst for the first time in a WZ Sge-type superoutburst. ASASSN-16kg, CRTS J000130.5+050624 and SDSS J113551.09+532246.2 are located in the period gap. We have newly obtained 15 orbital periods, including periods from early superhumps. △ Less

Submitted 29 June, 2017; v1 submitted 12 June, 2017; originally announced June 2017.

Comments: 126 pages. 168 figures, accepted for publication in PASJ. For readers' convenience, combined version with Supplementary Information is provided. See Ancillary files for the PASJ version (correction in author list)

Survey of Period Variations of Superhumps in SU UMa-Type Dwarf Novae. VIII: The Eighth Year (2015-2016)

Authors: Taichi Kato, Franz-Josef Hambsch, Berto Monard, Tonny Vanmunster, Yutaka Maeda, Ian Miller, Hiroshi Itoh, Seiichiro Kiyota, Keisuke Isogai, Mariko Kimura, Akira Imada, Tamas Tordai, Hidehiko Akazawa, Kenji Tanabe, Noritoshi Otani, Minako Ogi, Kazuko Ando, Naoki Takigawa, Pavol A. Dubovsky, Igor Kudzej, Sergey Yu. Shugarov, Natalia Katysheva, Polina Golysheva, Natalia Gladilina, Drahomir Chochol , et al. (53 additional authors not shown)

Abstract: Continuing the project described by Kato et al. (2009, arXiv:0905.1757), we collected times of superhump maxima for 128 SU UMa-type dwarf novae observed mainly during the 2015-2016 season and characterized these objects. The data have improved the distribution of orbital periods, the relation between the orbital period and the variation of superhumps, the relation between period variations and the… ▽ More Continuing the project described by Kato et al. (2009, arXiv:0905.1757), we collected times of superhump maxima for 128 SU UMa-type dwarf novae observed mainly during the 2015-2016 season and characterized these objects. The data have improved the distribution of orbital periods, the relation between the orbital period and the variation of superhumps, the relation between period variations and the rebrightening type in WZ Sge-type objects. Coupled with new measurements of mass ratios using growing stages of superhumps, we now have a clearer and statistically greatly improved evolutionary path near the terminal stage of evolution of cataclysmic variables. Three objects (V452 Cas, KK Tel, ASASSN-15cl) appear to have slowly growing superhumps, which is proposed to reflect the slow growth of the 3:1 resonance near the stability border. ASASSN-15sl, ASASSN-15ux, SDSS J074859.55+312512.6 and CRTS J200331.3-284941 are newly identified eclipsing SU UMa-type (or WZ Sge-type) dwarf novae. ASASSN-15cy has a short (~0.050 d) superhump period and appears to belong to EI Psc-type objects with compact secondaries having an evolved core. ASASSN-15gn, ASASSN-15hn, ASASSN-15kh and ASASSN-16bu are candidate period bouncers with superhump periods longer than 0.06 d. We have newly obtained superhump periods for 79 objects and 13 orbital periods, including periods from early superhumps. In order that the future observations will be more astrophysically beneficial and rewarding to observers, we propose guidelines how to organize observations of various superoutbursts. △ Less

Submitted 20 May, 2016; originally announced May 2016.

Comments: 123 pages, 162 figures, 119 tables, accepted for publication in PASJ (including supplementary information)

Hidden photon CDM search at Tokyo

Authors: Jun'ya Suzuki, Yoshizumi Inoue, Tomoki Horie, Makoto Minowa

Abstract: We report on a search for hidden photon cold dark matter (HP CDM) using a novel technique with a dish antenna. We constructed two independent apparatus: one is aiming at the detection of the HP with a mass of $\sim\,\rm{eV}$ which employs optical instruments, and the other is for a mass of $\sim5\times10^{-5}\, \rm{eV}$ utilizing a commercially available parabolic antenna facing on a plane reflect… ▽ More We report on a search for hidden photon cold dark matter (HP CDM) using a novel technique with a dish antenna. We constructed two independent apparatus: one is aiming at the detection of the HP with a mass of $\sim\,\rm{eV}$ which employs optical instruments, and the other is for a mass of $\sim5\times10^{-5}\, \rm{eV}$ utilizing a commercially available parabolic antenna facing on a plane reflector. From the result of the measurements, we found no evidence for the existence of HP CDM and set upper limits on the photon-HP mixing parameter $χ$. △ Less

Submitted 2 September, 2015; originally announced September 2015.

Comments: Contributed to the 11th Patras Workshop on Axions, WIMPs and WISPs, Zaragoza, June 22 to 26, 2015

Superfield description of (4+2n)-dimensional SYM theories and their mixtures on magnetized tori

Authors: Hiroyuki Abe, Tomoharu Horie, Keigo Sumita

Abstract: We provide a systematic way of dimensional reduction for $(4+2n)$-dimensional $U(N)$ supersymmetric Yang-Mills (SYM) theories ($n=0,1,2,3$) and their mixtures compactified on two-dimensional tori with background magnetic fluxes, which preserve a partial ${\mathcal N}=1$ supersymmetry out of full ${\mathcal N}=2,3$ or $4$ in the original SYM theories. It is formulated in an $\mathcal N=1$ superspac… ▽ More We provide a systematic way of dimensional reduction for $(4+2n)$-dimensional $U(N)$ supersymmetric Yang-Mills (SYM) theories ($n=0,1,2,3$) and their mixtures compactified on two-dimensional tori with background magnetic fluxes, which preserve a partial ${\mathcal N}=1$ supersymmetry out of full ${\mathcal N}=2,3$ or $4$ in the original SYM theories. It is formulated in an $\mathcal N=1$ superspace respecting the unbroken supersymmetry, and the four-dimensional effective action is written in terms of superfields representing $\mathcal N=1$ vector and chiral multiplets, those arise from the higher-dimensional SYM theories. We also identify the dilaton and geometric moduli dependence of matter Kähler metrics and superpotential couplings as well as of gauge kinetic functions in the effective action. The results would be useful for various phenomenological/cosmological model buildings with SYM theories or D-branes wrapping magnetized tori, especially, with mixture configurations of them with different dimensionalities from each other. △ Less

Submitted 9 July, 2015; originally announced July 2015.

Comments: 37pages

Report number: WU-HEP-15-08

Experimental Search for Hidden Photon CDM in the eV mass range with a Dish Antenna

Authors: J. Suzuki, T. Horie, Y. Inoue, M. Minowa

Abstract: A search for hidden photon cold dark matter (HP CDM) using a new technique with a dish antenna is reported. From the result of the measurement, we found no evidence for the existence of HP CDM and set an upper limit on the photon-HP mixing parameter $χ$ of $\sim 6\times 10^{-12}$ for the hidden photon mass $m_γ= 3.1 \pm 1.2$ eV. A search for hidden photon cold dark matter (HP CDM) using a new technique with a dish antenna is reported. From the result of the measurement, we found no evidence for the existence of HP CDM and set an upper limit on the photon-HP mixing parameter $χ$ of $\sim 6\times 10^{-12}$ for the hidden photon mass $m_γ= 3.1 \pm 1.2$ eV. △ Less

Submitted 1 April, 2015; originally announced April 2015.

Comments: 10 pages, 5 figures. To be submitted to JCAP

Search for Hidden photons with Sumico

Authors: Yoshizumi Inoue, Tetsuya Mizumoto, Ryosuke Ohta, Tomoki Horie, Jun'ya Suzuki, Makoto Minowa

Abstract: We searched for solar hidden photons in the visible photon energy range using a hidden photon detector add-on attached to Sumico. It consists of a parabolic mirror of dia. 0.5m and f=1m installed in a vacuum chamber, and a low noise photomultiplier tube at the focal point. No evidence for the existence of hidden photons was found in the latest measurement giving a new limit on the photon-hidden ph… ▽ More We searched for solar hidden photons in the visible photon energy range using a hidden photon detector add-on attached to Sumico. It consists of a parabolic mirror of dia. 0.5m and f=1m installed in a vacuum chamber, and a low noise photomultiplier tube at the focal point. No evidence for the existence of hidden photons was found in the latest measurement giving a new limit on the photon-hidden photon mixing parameter in the hidden photon mass range 0.001-1eV. △ Less

Submitted 27 December, 2013; originally announced December 2013.

Comments: 6 pages. Contributed to the 9th Patras Workshop on Axions, WIMPs and WISPs, Mainz, June 24-28, 2013

Experimental search for solar hidden photons in the eV energy range using kinetic mixing with photons

Authors: T. Mizumoto, R. Ohta, T. Horie, J. Suzuki, Y. Inoue, M. Minowa

Abstract: We have searched for solar hidden photons in the eV energy range using a dedicated hidden photon detector. The detector consisted of a parabolic mirror with a diameter of 500mm and a focal length of 1007mm installed in a vacuum chamber, and a photomultiplier tube at its focal point. The detector was attached to the Tokyo axion helioscope, Sumico which has a mechanism to track the sun. From the res… ▽ More We have searched for solar hidden photons in the eV energy range using a dedicated hidden photon detector. The detector consisted of a parabolic mirror with a diameter of 500mm and a focal length of 1007mm installed in a vacuum chamber, and a photomultiplier tube at its focal point. The detector was attached to the Tokyo axion helioscope, Sumico which has a mechanism to track the sun. From the result of the measurement, we found no evidence for the existence of hidden photons and set a limit on the photon-hidden photon mixing parameter χdepending on the hidden photon mass m_{γ'}. △ Less

Submitted 18 June, 2013; v1 submitted 5 February, 2013; originally announced February 2013.

Comments: 11 pages, 4 figures. To be submitted to JCAP. Added fig.2. Added references to some recent theoretical developments. Added results of new references to the exclusion plot(fig.5). Our results remain unchanged

Report number: RESCEU-18/13

Journal ref: JCAP07(2013)013