arXiv subjects

Wataru Hirota

Publications and source records attributed to Wataru Hirota.

6 recordsLinked to original sources

arXiv2026

Why Expert Alignment Is Hard: Evidence from Subjective Evaluation

Aligning large language models with expert judgment is especially difficult in subjective evaluation tasks, where experts may disagree, rely on tacit criteria, and change their judgments over time. In this paper, we study expert alignment as a way to understand this difficulty. Using expert evaluations and follow-up questionnaires, we examine how different forms of expert information affect alignment and what this reveals about subjective judgment. Our findings show four consistent patterns. First, alignment difficulty varies substantially across experts, suggesting that expert evaluation styles differ widely in their distance from a model's prior behavior. Second, explicit criteria and reasoning do not always improve alignment, indicating that expert judgment is not fully captured by verbalized rules. Third, editing is sensitive to both the number and the identity of examples, with small numbers of edits providing useful but unstable gains. Fourth, alignment difficulty differs across evaluation dimensions: dimensions grounded more directly in proposal content are easier to align, while dimensions requiring external knowledge or value-based judgment remain harder. Taken together, these results suggest that expert alignment is difficult not only because of model limitations, but also because subjective evaluation is inherently heterogeneous, partly tacit, dimension-dependent, and temporally unstable.

cs.CL↗

arXiv2026

Aggregate vs. Personalized Judges in Business Idea Evaluation: Evidence from Expert Disagreement

Evaluating LLM-generated business ideas is often harder to scale than generating them. Unlike standard NLP benchmarks, business idea evaluation relies on multi-dimensional criteria such as feasibility, novelty, differentiation, user need, and market size, and expert judgments often disagree. This paper studies a methodological question raised by such disagreement: should an automatic judge approximate an aggregate consensus, or model evaluators individually? We introduce PBIG-DATA, a dataset of approximately 3,000 individual scores across 300 patent-grounded product ideas, provided by domain experts on six business-oriented dimensions: specificity, technical validity, innovativeness, competitive advantage, need validity, and market size. Analyses show substantial expert disagreement on fine-grained ordinal scores, while agreement is higher under coarse selection, suggesting structured heterogeneity rather than random noise. We then compare three judge configurations: a rubric-only zero-shot judge, an aggregate judge conditioned on mixed evaluator histories, and a personalized judge conditioned on the target evaluator's scoring history. Across dimensions and model sizes, personalized judges align more closely with the corresponding evaluator than aggregate judges, and evaluator agreement correlates with similarity of judge-generated reasoning only under personalized conditioning. These results indicate that pooled labels can be a fragile target in pluralistic evaluation settings and motivate evaluator-conditioned judge designs for business idea assessment.

cs.CL↗

arXiv2025

Exploring Design of Multi-Agent LLM Dialogues for Research Ideation

Large language models (LLMs) are increasingly used to support creative tasks such as research idea generation. While recent work has shown that structured dialogues between LLMs can improve the novelty and feasibility of generated ideas, the optimal design of such interactions remains unclear. In this study, we conduct a comprehensive analysis of multi-agent LLM dialogues for scientific ideation. We compare different configurations of agent roles, number of agents, and dialogue depth to understand how these factors influence the novelty and feasibility of generated ideas. Our experimental setup includes settings where one agent generates ideas and another critiques them, enabling iterative improvement. Our results show that enlarging the agent cohort, deepening the interaction depth, and broadening agent persona heterogeneity each enrich the diversity of generated ideas. Moreover, specifically increasing critic-side diversity within the ideation-critique-revision loop further boosts the feasibility of the final proposals. Our findings offer practical guidelines for building effective multi-agent LLM systems for scientific ideation. Our code is available at https://github.com/g6000/MultiAgent-Research-Ideator.

cs.CL↗

arXiv2022

Machop: an End-to-End Generalized Entity Matching Framework

Real-world applications frequently seek to solve a general form of the Entity Matching (EM) problem to find associated entities. Such scenarios include matching jobs to candidates in job targeting, matching students with courses in online education, matching products with user reviews on e-commercial websites, and beyond. These tasks impose new requirements such as matching data entries with diverse formats or having a flexible and semantics-rich matching definition, which are beyond the current EM task formulation or approaches. In this paper, we introduce the problem of Generalized Entity Matching (GEM) that satisfies these practical requirements and presents an end-to-end pipeline Machop as the solution. Machop allows end-users to define new matching tasks from scratch and apply them to new domains in a step-by-step manner. Machop casts the GEM problem as sequence pair classification so as to utilize the language understanding capability of Transformers-based language models (LMs) such as BERT. Moreover, it features a novel external knowledge injection approach with structure-aware pooling methods that allow domain experts to guide the LM to focus on the key matching information thus further contributing to the overall performance. Our experiments and case studies on real-world datasets from a popular recruiting platform show a significant 17.1% gain in F1 score against state-of-the-art methods along with meaningful matching results that are human-understandable.

cs.DB↗

arXiv2021

Machamp: A Generalized Entity Matching Benchmark

Entity Matching (EM) refers to the problem of determining whether two different data representations refer to the same real-world entity. It has been a long-standing interest of the data management community and many efforts have been paid in creating benchmark tasks as well as in developing advanced matching techniques. However, existing benchmark tasks for EM are limited to the case where the two data collections of entities are structured tables with the same schema. Meanwhile, the data collections for matching could be structured, semi-structured, or unstructured in real-world scenarios of data science. In this paper, we come up with a new research problem -- Generalized Entity Matching to satisfy this requirement and create a benchmark Machamp for it. Machamp consists of seven tasks having diverse characteristics and thus provides good coverage of use cases in real applications. We summarize existing EM benchmark tasks for structured tables and conduct a series of processing and cleaning efforts to transform them into matching tasks between tables with different structures. Based on that, we further conduct comprehensive profiling of the proposed benchmark tasks and evaluate popular entity matching approaches on them. With the help of Machamp, it is the first time that researchers can evaluate EM techniques between data collections with different structures.

cs.DB↗

arXiv2019

Emu: Enhancing Multilingual Sentence Embeddings with Semantic Specialization

We present Emu, a system that semantically enhances multilingual sentence embeddings. Our framework fine-tunes pre-trained multilingual sentence embeddings using two main components: a semantic classifier and a language discriminator. The semantic classifier improves the semantic similarity of related sentences, whereas the language discriminator enhances the multilinguality of the embeddings via multilingual adversarial training. Our experimental results based on several language pairs show that our specialized embeddings outperform the state-of-the-art multilingual sentence embedding model on the task of cross-lingual intent classification using only monolingual labeled data.

cs.CL↗