Abstract:In generative commonsense reasoning tasks such as CommonGen, generative large language models (LLMs) compose sentences that include all given concepts. However, when focusing on instruction-following capabilities, if a prompt specifies a concept order, LLMs must generate sentences that adhere to the specified order. To address this, we propose Ordered CommonGen, a benchmark designed to evaluate the compositional generalization and instruction-following abilities of LLMs. This benchmark measures ordered coverage to assess whether concepts are generated in the specified order, enabling a simultaneous evaluation of both abilities. We conducted a comprehensive analysis using 36 LLMs and found that, while LLMs generally understand the intent of instructions, biases toward specific concept order patterns often lead to low-diversity outputs or identical results even when the concept order is altered. Moreover, even the most instruction-compliant LLM achieved only about 75% ordered coverage, highlighting the need for improvements in both instruction-following and compositional generalization capabilities.
Abstract:Since self-attention layers in Transformers are permutation invariant by design, positional encodings must be explicitly incorporated to enable spatial understanding. However, fixed-size lookup tables used in traditional learnable position embeddings (PEs) limit extrapolation capabilities beyond pre-trained sequence lengths. Expert-designed methods such as ALiBi and RoPE, mitigate this limitation but demand extensive modifications for adapting to new modalities, underscoring fundamental challenges in adaptability and scalability. In this work, we present SeqPE, a unified and fully learnable position encoding framework that represents each $n$-dimensional position index as a symbolic sequence and employs a lightweight sequential position encoder to learn their embeddings in an end-to-end manner. To regularize SeqPE's embedding space, we introduce two complementary objectives: a contrastive objective that aligns embedding distances with a predefined position-distance function, and a knowledge distillation loss that anchors out-of-distribution position embeddings to in-distribution teacher representations, further enhancing extrapolation performance. Experiments across language modeling, long-context question answering, and 2D image classification demonstrate that SeqPE not only surpasses strong baselines in perplexity, exact match (EM), and accuracy--particularly under context length extrapolation--but also enables seamless generalization to multi-dimensional inputs without requiring manual architectural redesign. We release our code, data, and checkpoints at https://github.com/ghrua/seqpe.
Abstract:Transformers deliver outstanding performance across a wide range of tasks and are now a dominant backbone architecture for large language models (LLMs). Their task-solving performance is improved by increasing parameter size, as shown in the recent studies on parameter scaling laws. Although recent mechanistic-interpretability studies have deepened our understanding of the internal behavior of Transformers by analyzing their residual stream, the relationship between these internal mechanisms and the parameter scaling laws remains unclear. To bridge this gap, we focus on layers and their size, which mainly decide the parameter size of Transformers. For this purpose, we first theoretically investigate the layers within the residual stream through a bias-diversity decomposition. The decomposition separates (i) bias, the error of each layer's output from the ground truth, and (ii) diversity, which indicates how much the outputs of each layer differ from each other. Analyzing Transformers under this theory reveals that performance improves when individual layers make predictions close to the correct answer and remain mutually diverse. We show that diversity becomes especially critical when individual layers' outputs are far from the ground truth. Finally, we introduce an information-theoretic diversity and show our main findings that adding layers enhances performance only when those layers behave differently, i.e., are diverse. We also reveal the performance gains from increasing the number of layers exhibit submodularity: marginal improvements diminish as additional layers increase, mirroring the logarithmic convergence predicted by the parameter scaling laws. Experiments on multiple semantic-understanding tasks with various LLMs empirically confirm the theoretical properties derived in this study.
Abstract:Large Language Models (LLMs) are known to process information using a proficient internal language consistently, referred to as latent language, which may differ from the input or output languages. However, how the discrepancy between the latent language and the input and output language affects downstream task performance remains largely unexplored. While many studies research the latent language of LLMs, few address its importance in influencing task performance. In our study, we hypothesize that thinking in latent language consistently enhances downstream task performance. To validate this, our work varies the input prompt languages across multiple downstream tasks and analyzes the correlation between consistency in latent language and task performance. We create datasets consisting of questions from diverse domains such as translation and geo-culture, which are influenced by the choice of latent language. Experimental results across multiple LLMs on translation and geo-culture tasks, which are sensitive to the choice of language, indicate that maintaining consistency in latent language is not always necessary for optimal downstream task performance. This is because these models adapt their internal representations near the final layers to match the target language, reducing the impact of consistency on overall performance.
Abstract:We introduce gec-metrics, a library for using and developing grammatical error correction (GEC) evaluation metrics through a unified interface. Our library enables fair system comparisons by ensuring that everyone conducts evaluations using a consistent implementation. Moreover, it is designed with a strong focus on API usage, making it highly extensible. It also includes meta-evaluation functionalities and provides analysis and visualization scripts, contributing to developing GEC evaluation metrics. Our code is released under the MIT license and is also distributed as an installable package. The video is available on YouTube.
Abstract:Large-scale Vision Language Models (LVLMs) are increasingly being applied to a wide range of real-world multimodal applications, involving complex visual and linguistic reasoning. As these models become more integrated into practical use, they are expected to handle complex aspects of human interaction. Among these, color perception is a fundamental yet highly variable aspect of visual understanding. It differs across individuals due to biological factors such as Color Vision Deficiencies (CVDs), as well as differences in culture and language. Despite its importance, perceptual diversity has received limited attention. In our study, we evaluate LVLMs' ability to account for individual level perceptual variation using the Ishihara Test, a widely used method for detecting CVDs. Our results show that LVLMs can explain CVDs in natural language, but they cannot simulate how people with CVDs perceive color in image based tasks. These findings highlight the need for multimodal systems that can account for color perceptual diversity and support broader discussions on perceptual inclusiveness and fairness in multimodal AI.
Abstract:Retrieval-Augmented Generation (RAG) has emerged as a way to complement the in-context knowledge of Large Language Models (LLMs) by integrating external documents. However, real-world applications demand not only accuracy but also interpretability. While dense retrieval methods provide high accuracy, they lack interpretability; conversely, sparse retrieval methods offer transparency but often fail to capture the full intent of queries due to their reliance on keyword matching. To address these issues, we introduce IterKey, an LLM-driven iterative keyword generation framework that enhances RAG via sparse retrieval. IterKey consists of three LLM-driven stages: generating keywords for retrieval, generating answers based on retrieved documents, and validating the answers. If validation fails, the process iteratively repeats with refined keywords. Across four QA tasks, experimental results show that IterKey achieves 5% to 20% accuracy improvements over BM25-based RAG and simple baselines. Its performance is comparable to dense retrieval-based RAG and prior iterative query refinement methods using dense models. In summary, IterKey is a novel BM25-based approach leveraging LLMs to iteratively refine RAG, effectively balancing accuracy with interpretability.
Abstract:Generating images from prompts containing specific entities requires models to retain as much entity-specific knowledge as possible. However, fully memorizing such knowledge is impractical due to the vast number of entities and their continuous emergence. To address this, we propose Text-based Intelligent Generation with Entity prompt Refinement (TextTIGER), which augments knowledge on entities included in the prompts and then summarizes the augmented descriptions using Large Language Models (LLMs) to mitigate performance degradation from longer inputs. To evaluate our method, we introduce WiT-Cub (WiT with Captions and Uncomplicated Background-explanations), a dataset comprising captions, images, and an entity list. Experiments on four image generation models and five LLMs show that TextTIGER improves image generation performance in standard metrics (IS, FID, and CLIPScore) compared to caption-only prompts. Additionally, multiple annotators' evaluation confirms that the summarized descriptions are more informative, validating LLMs' ability to generate concise yet rich descriptions. These findings demonstrate that refining prompts with augmented and summarized entity-related descriptions enhances image generation capabilities. The code and dataset will be available upon acceptance.
Abstract:The $k$-nearest-neighbor language model ($k$NN-LM), one of the retrieval-augmented language models, improves the perplexity for given text by directly accessing a large datastore built from any text data during inference. A widely held hypothesis for the success of $k$NN-LM is that its explicit memory, i.e., the datastore, enhances predictions for long-tail phenomena. However, prior works have primarily shown its ability to retrieve long-tail contexts, leaving the model's performance remain underexplored in estimating the probabilities of long-tail target tokens during inference. In this paper, we investigate the behavior of $k$NN-LM on low-frequency tokens, examining prediction probability, retrieval accuracy, token distribution in the datastore, and approximation error of the product quantization. Our experimental results reveal that $k$NN-LM does not improve prediction performance for low-frequency tokens but mainly benefits high-frequency tokens regardless of long-tail contexts in the datastore.
Abstract:One of the goals of automatic evaluation metrics in grammatical error correction (GEC) is to rank GEC systems such that it matches human preferences. However, current automatic evaluations are based on procedures that diverge from human evaluation. Specifically, human evaluation derives rankings by aggregating sentence-level relative evaluation results, e.g., pairwise comparisons, using a rating algorithm, whereas automatic evaluation averages sentence-level absolute scores to obtain corpus-level scores, which are then sorted to determine rankings. In this study, we propose an aggregation method for existing automatic evaluation metrics which aligns with human evaluation methods to bridge this gap. We conducted experiments using various metrics, including edit-based metrics, $n$-gram based metrics, and sentence-level metrics, and show that resolving the gap improves results for the most of metrics on the SEEDA benchmark. We also found that even BERT-based metrics sometimes outperform the metrics of GPT-4. We publish our unified implementation of the metrics and meta-evaluations.