Akira Kawai

Keito Inoshita, Xiaokang Zhou, Akira Kawai et al.

In practical deep learning deployment, the scarcity of data and the imbalance of label distributions often lead to semantically uncovered regions within the real-world data distribution, hindering model training and causing misclassification near class boundaries as well as unstable behaviors in peripheral areas. Although recent large language models (LLMs) show promise for data augmentation, an integrated framework that simultaneously achieves directional control of generation, domain alignment, and quality control has not yet been fully established. To address these challenges, we propose a Cluster-conditioned Interpolative and Extrapolative framework for Geometry-Aware and Domain-aligned data augmentation (CIEGAD), which systematically complements both in-distribution and out-of-distribution semantically uncovered regions. CIEGAD constructs domain profiles through cluster conditioning, allocates generation with a hierarchical frequency-geometric allocation integrating class frequency and geometric indicators, and finely controls generation directions via the coexistence of interpolative and extrapolative synthesis. It further performs quality control through geometry-constrained filtering combined with an LLM-as-a-Judge mechanism. Experiments on multiple classification tasks demonstrate that CIEGAD effectively extends the periphery of real-world data distributions while maintaining high alignment between generated and real-world data as well as semantic diversity. In particular, for long-tailed and multi-class classification tasks, CIEGAD consistently improves F1 and recall, validating the triple harmony of distributional consistency, diversity, and quality. These results indicate that CIEGAD serves as a practically oriented data augmentation framework that complements underrepresented regions while preserving alignment with real-world data.

Keito Inoshita, Xiaokang Zhou, Akira Kawai et al.

Human annotators frequently disagree on emotion labels, yet most evaluations of Large Language Model (LLM) emotion annotation collapse these judgments into a single gold standard, discarding the distributional information that disagreement encodes. We ask whether LLMs capture the structure of this disagreement, not just majority labels, by comparing emotion judgment distributions between human annotators and four zero-shot LLMs, plus a fine-tuned RoBERTa baseline, across two complementary benchmarks: GoEmotions and EmoBank, totaling 640,000 LLM responses. Zero-shot models diverge substantially from human distributions, and in-domain fine-tuning, not model scale, is required to close the gap. We formalize a lexical-grounding gradient through a quantitative transparency score that predicts per-category human--LLM agreement: LLMs reliably capture emotions with explicit lexical markers but systematically fail on pragmatically complex emotions requiring contextual inference, a pattern that replicates across both categorical and continuous emotion frameworks. We further propose three lightweight post-hoc calibration methods that reduce the distributional gap by up to 14\%, and provide actionable guidelines for when LLM emotion annotations can, and cannot, substitute for human labeling.

Mitsumasa Nakajima, Kohki Shibahara, Kohei Ikeda et al.

The explosive growth of global data traffic demands scalable and energy-efficient optical communication systems. Spatial division multiplexing (SDM) using multicore or multimode fibers is a promising solution to overcome the capacity limit of single-mode fibers. However, long-haul SDM transmission faces significant challenges due to modal dispersion, which imposes heavy computational loads on digital signal processing (DSP) for signal equalization. Here, we propose parameterized SDM transmission, where programmable photonic unitary processors are installed at intermediate nodes. Instead of relying on conventional digital equalization only on the receiver side, our approach enables direct optimization of the SDM transmission channel itself by the programmable unitary processor, which reduces digital post-processing loads. We introduce a gradient-based optimization algorithm using a differentiable SDM transmission model to determine the optimal unitary transformation. As a key enabler, we first implemented telecom-grade programmable photonic unitary processor, achieving a low-loss (2.1 dB fiber-to-fiber), wideband (full C-band), polarization-independent, and high-fidelity (R2>96% across the C-band) operation. We experimentally demonstrate 1300-km transmission using a three-mode fiber, achieving strong agreement between simulation and experiment. The optimized photonic processor significantly reduces modal dispersion and post-processing complexity. Our results establish a scalable framework for integrating photonic computation into the optical layer, enabling more efficient, high-capacity optical networks.