Large-scale Vision-Language Models (LVLMs) process both images and text, excelling in multimodal tasks such as image captioning and description generation. However, while these models excel at generating factual content, their ability to generate and evaluate texts reflecting perspectives on the same image, depending on the context, has not been sufficiently explored. To address this, we propose IRR: Image Review Rank, a novel evaluation framework designed to assess critic review texts from multiple perspectives. IRR evaluates LVLMs by measuring how closely their judgments align with human interpretations. We validate it using a dataset of images from 15 categories, each with five critic review texts and annotated rankings in both English and Japanese, totaling over 2,000 data instances. The datasets are available at https://hf.co/datasets/naist-nlp/Wiki-ImageReview1.0. Our results indicate that, although LVLMs exhibited consistent performance across languages, their correlation with human annotations was insufficient, highlighting the need for further advancements. These findings highlight the limitations of current evaluation methods and the need for approaches that better capture human reasoning in Vision & Language tasks.

Existing multilingual neural machine translation (MNMT) approaches mainly focus on improving models with the encoder-decoder architecture to translate multiple languages. However, decoder-only architecture has been explored less in MNMT due to its underperformance when trained on parallel data solely. In this work, we attribute the issue of the decoder-only architecture to its lack of language transfer capability. Specifically, the decoder-only architecture is insufficient in encoding source tokens with the target language features. We propose dividing the decoding process into two stages so that target tokens are explicitly excluded in the first stage to implicitly boost the transfer capability across languages. Additionally, we impose contrastive learning on translation instructions, resulting in improved performance in zero-shot translation. We conduct experiments on TED-19 and OPUS-100 datasets, considering both training from scratch and fine-tuning scenarios. Experimental results show that, compared to the encoder-decoder architecture, our methods not only perform competitively in supervised translations but also achieve improvements of up to 3.39 BLEU, 6.99 chrF++, 3.22 BERTScore, and 4.81 COMET in zero-shot translations.

Vision Language Models (VLMs) often struggle with culture-specific knowledge, particularly in languages other than English and in underrepresented cultural contexts. To evaluate their understanding of such knowledge, we introduce WorldCuisines, a massive-scale benchmark for multilingual and multicultural, visually grounded language understanding. This benchmark includes a visual question answering (VQA) dataset with text-image pairs across 30 languages and dialects, spanning 9 language families and featuring over 1 million data points, making it the largest multicultural VQA benchmark to date. It includes tasks for identifying dish names and their origins. We provide evaluation datasets in two sizes (12k and 60k instances) alongside a training dataset (1 million instances). Our findings show that while VLMs perform better with correct location context, they struggle with adversarial contexts and predicting specific regional cuisines and languages. To support future research, we release a knowledge base with annotated food entries and images along with the VQA data.

Adapting named entity recognition (NER) methods to new domains poses significant challenges. We introduce RapidNER, a framework designed for the rapid deployment of NER systems through efficient dataset construction. RapidNER operates through three key steps: (1) extracting domain-specific sub-graphs and triples from a general knowledge graph, (2) collecting and leveraging texts from various sources to build the NERsocial dataset, which focuses on entities typical in human-robot interaction, and (3) implementing an annotation scheme using Elasticsearch (ES) to enhance efficiency. NERsocial, validated by human annotators, includes six entity types, 153K tokens, and 99.4K sentences, demonstrating RapidNER's capability to expedite dataset creation.

The tasks of lexical complexity prediction (LCP) and complex word identification (CWI) commonly presuppose that difficult to understand words are shared by the target population. Meanwhile, personalization methods have also been proposed to adapt models to individual needs. We verify that a recent Japanese LCP dataset is representative of its target population by partially replicating the annotation. By another reannotation we show that native Chinese speakers perceive the complexity differently due to Sino-Japanese vocabulary. To explore the possibilities of personalization, we compare competitive baselines trained on the group mean ratings and individual ratings in terms of performance for an individual. We show that the model trained on a group mean performs similarly to an individual model in the CWI task, while achieving good LCP performance for an individual is difficult. We also experiment with adapting a finetuned BERT model, which results only in marginal improvements across all settings.

What kinds of and how much data is necessary for language models to induce grammatical knowledge to judge sentence acceptability? Recent language models still have much room for improvement in their data efficiency compared to humans. This paper investigates whether language models efficiently use indirect data (indirect evidence), from which they infer sentence acceptability. In contrast, humans use indirect evidence efficiently, which is considered one of the inductive biases contributing to efficient language acquisition. To explore this question, we introduce the Wug InDirect Evidence Test (WIDET), a dataset consisting of training instances inserted into the pre-training data and evaluation instances. We inject synthetic instances with newly coined wug words into pretraining data and explore the model's behavior on evaluation data that assesses grammatical acceptability regarding those words. We prepare the injected instances by varying their levels of indirectness and quantity. Our experiments surprisingly show that language models do not induce grammatical knowledge even after repeated exposure to instances with the same structure but differing only in lexical items from evaluation instances in certain language phenomena. Our findings suggest a potential direction for future research: developing models that use latent indirect evidence to induce grammatical knowledge.

Previous studies on sequence-based extraction of human movement trajectories have an issue of inadequate trajectory representation. Specifically, a pair of locations may not be lined up in a sequence especially when one location includes the other geographically. In this study, we propose a graph representation that retains information on the geographic hierarchy as well as the temporal order of visited locations, and have constructed a benchmark dataset for graph-structured trajectory extraction. The experiments with our baselines have demonstrated that it is possible to accurately predict visited locations and the order among them, but it remains a challenge to predict the hierarchical relations.

Text generation commonly relies on greedy and beam decoding that limit the search space and degrade output quality. Minimum Bayes Risk (MBR) decoding can mitigate this problem by utilizing automatic evaluation metrics and model-generated pseudo-references. Previous studies have conducted empirical analyses to reveal the improvement by MBR decoding, and reported various observations. However, despite these observations, the theoretical relationship between them remains uncertain. To address this, we present a novel theoretical interpretation of MBR decoding from the perspective of bias-diversity decomposition. We decompose errors in the estimated quality of generated hypotheses in MBR decoding into two key factors: bias, which reflects the closeness between utility functions and human evaluations, and diversity, which represents the variation in the estimated quality of utility functions. Our theoretical analysis reveals the difficulty in simultaneously improving both bias and diversity, and highlights the effectiveness of increasing diversity to enhance MBR decoding performance. This analysis verifies the alignment between our theoretical insights and the empirical results reported in previous work. Furthermore, to support our theoretical findings, we propose a new metric, pseudo-bias, which approximates the bias term using gold references. We also introduce a new MBR approach, Metric-augmented MBR (MAMBR), which increases diversity by adjusting the behavior of utility functions without altering the pseudo-references. Experimental results across multiple NLP tasks show that the decomposed terms in the bias-diversity decomposition correlate well with performance, and that MAMBR improves text generation quality by modifying utility function behavior. Our code will be available at https://github.com/naist-nlp/mbr-bias-diversity.

A large part of human communication relies on nonverbal cues such as facial expressions, eye contact, and body language. Unlike language or sign language, such nonverbal communication lacks formal rules, requiring complex reasoning based on commonsense understanding. Enabling current Video Large Language Models (VideoLLMs) to accurately interpret body language is a crucial challenge, as human unconscious actions can easily cause the model to misinterpret their intent. To address this, we propose a dataset, BQA, a body language question answering dataset, to validate whether the model can correctly interpret emotions from short clips of body language comprising 26 emotion labels of videos of body language. We evaluated various VideoLLMs on BQA and revealed that understanding body language is challenging, and our analyses of the wrong answers by VideoLLMs show that certain VideoLLMs made significantly biased answers depending on the age group and ethnicity of the individuals in the video. The dataset is available.

The context window of large language models (LLMs) has been extended significantly in recent years. However, while the context length that the LLM can process has grown, the capability of the model to accurately reason over that context degrades noticeably. This occurs because modern LLMs often become overwhelmed by the vast amount of information in the context; when answering questions, the model must identify and reason over relevant evidence sparsely distributed throughout the text. To alleviate the challenge of long-context reasoning, we develop a retrieve-then-reason framework, enabling LLMs to reason over relevant evidence collected during an intermediate retrieval step. We find that modern LLMs struggle to accurately retrieve relevant facts and instead, often hallucinate "retrieved facts", resulting in flawed reasoning and the production of incorrect answers. Through extensive experiments on long-context QA benchmarks, we find our method to outperform competitive baselines by large margins, achieving at least 8.4 and 7.9 EM gains on the long-context versions of HotpotQA and SQuAD datasets, respectively. While these developments are promising, in our preliminary study, we show that the long-context performance of LLMs varied significantly across different tasks. We observe that, when tasked to generate answers by directly reasoning over the full context, performance degrades as the input context grows. In contrast, when tasked with retrieving the set of evidence relevant to the question, the performance of LLMs is only mildly affected by the growth of the input context.