This paper presents a comprehensive study on the tokenization techniques employed by state-of-the-art large language models (LLMs) and their implications on the cost and availability of services across different languages, especially low resource languages. The analysis considers multiple LLMs, including GPT-4 (using cl100k_base embeddings), GPT-3 (with p50k_base embeddings), and DaVinci (employing r50k_base embeddings), as well as the widely used BERT base tokenizer. The study evaluates the tokenization variability observed across these models and investigates the challenges of linguistic representation in subword tokenization. The research underscores the importance of fostering linguistically-aware development practices, especially for languages that are traditionally under-resourced. Moreover, this paper introduces case studies that highlight the real-world implications of tokenization choices, particularly in the context of electronic health record (EHR) systems. This research aims to promote generalizable Internationalization (I18N) practices in the development of AI services in this domain and beyond, with a strong emphasis on inclusivity, particularly for languages traditionally underrepresented in AI applications.

Gender-Based Violence (GBV) is an increasing problem online, but existing datasets fail to capture the plurality of possible annotator perspectives or ensure the representation of affected groups. We revisit two important stages in the moderation pipeline for GBV: (1) manual data labelling; and (2) automated classification. For (1), we examine two datasets to investigate the relationship between annotator identities and attitudes and the responses they give to two GBV labelling tasks. To this end, we collect demographic and attitudinal information from crowd-sourced annotators using three validated surveys from Social Psychology. We find that higher Right Wing Authoritarianism scores are associated with a higher propensity to label text as sexist, while for Social Dominance Orientation and Neosexist Attitudes, higher scores are associated with a negative tendency to do so. For (2), we conduct classification experiments using Large Language Models and five prompting strategies, including infusing prompts with annotator information. We find: (i) annotator attitudes affect the ability of classifiers to predict their labels; (ii) including attitudinal information can boost performance when we use well-structured brief annotator descriptions; and (iii) models struggle to reflect the increased complexity and imbalanced classes of the new label sets.

Retrieval-Augmented Generation (RAG) improves LLMs by enabling them to incorporate external data during generation. This raises concerns for data owners regarding unauthorized use of their content in RAG systems. Despite its importance, the challenge of detecting such unauthorized usage remains underexplored, with existing datasets and methodologies from adjacent fields being ill-suited for its study. In this work, we take several steps to bridge this gap. To facilitate research on this challenge, we further introduce a novel dataset specifically designed for benchmarking RAG-DI methods under realistic conditions, and propose a set of baseline approaches. Our work provides a foundation for future studies of RAG-DI and highlights LLM watermarks as a promising approach to this problem. Retrieval-Augmented Generation (RAG) has emerged as a popular approach to mitigate limitations of large language models (LLMs), such as hallucinations, the high cost of adapting to new knowledge via fine-tuning, and the inability to back up claims by sources (Lewis et al., 2020). By integrating retrieval, LLMs gain in-context access to large corpora of high-quality, up-to-date data, enabling them to generate more accurate and source-supported responses. To maintain relevance, RAG providers must continuously update their corpus with new data. However, this raises concerns regarding the unauthorized usage of documents, particularly when publicly available documents are used without the owner's permission (Grynbaum & Mac, 2023; Wei et al., 2024a). There is currently no way to conclusively prove such unauthorized usage by a RAG system, and enforce an opt-out by the owner. RAG Dataset Inference (RAG-DI) We formalize the corresponding problem as RAG Dataset Inference (RAG-DI), where a data owner aims to detect unauthorized inclusion of their dataset within a RAG corpus via black-box queries, illustrated in Figure 1.

With the proliferation of Large Language Models (LLMs) in diverse domains, there is a particular need for unified evaluation standards in clinical medical scenarios, where models need to be examined very thoroughly. We present CliMedBench, a comprehensive benchmark with 14 expert-guided core clinical scenarios specifically designed to assess the medical ability of LLMs across 7 pivot dimensions. It comprises 33,735 questions derived from real-world medical reports of top-tier tertiary hospitals and authentic examination exercises. The reliability of this benchmark has been confirmed in several ways. Subsequent experiments with existing LLMs have led to the following findings: (i) Chinese medical LLMs underperform on this benchmark, especially where medical reasoning and factual consistency are vital, underscoring the need for advances in clinical knowledge and diagnostic accuracy. (ii) Several general-domain LLMs demonstrate substantial potential in medical clinics, while the limited input capacity of many medical LLMs hinders their practical use. These findings reveal both the strengths and limitations of LLMs in clinical scenarios and offer critical insights for medical research.

Large language models (LLMs) are stochastic, and not all models give deterministic answers, even when setting temperature to zero with a fixed random seed. However, few benchmark studies attempt to quantify uncertainty, partly due to the time and cost of repeated experiments. We use benchmarks designed for testing LLMs' capacity to reason about cardinal directions to explore the impact of experimental repeats on mean score and prediction interval. We suggest a simple method for cost-effectively quantifying the uncertainty of a benchmark score and make recommendations concerning reproducible LLM evaluation.

The potential effectiveness of counterspeech as a hate speech mitigation strategy is attracting increasing interest in the NLG research community, particularly towards the task of automatically producing it. However, automatically generated responses often lack the argumentative richness which characterises expert-produced counterspeech. In this work, we focus on two aspects of counterspeech generation to produce more cogent responses. First, by investigating the tension between helpfulness and harmlessness of LLMs, we test whether the presence of safety guardrails hinders the quality of the generations. Secondly, we assess whether attacking a specific component of the hate speech results in a more effective argumentative strategy to fight online hate. By conducting an extensive human and automatic evaluation, we show how the presence of safety guardrails can be detrimental also to a task that inherently aims at fostering positive social interactions. Moreover, our results show that attacking a specific component of the hate speech, and in particular its implicit negative stereotype and its hateful parts, leads to higher-quality generations.

MLLM agents demonstrate potential for complex embodied tasks by retrieving multimodal task-relevant trajectory data. However, current retrieval methods primarily focus on surface-level similarities of textual or visual cues in trajectories, neglecting their effectiveness for the specific task at hand. To address this issue, we propose a novel method, MLLM as ReTriever (MART), which enhances the performance of embodied agents by utilizing interaction data to fine-tune an MLLM retriever based on preference learning, such that the retriever fully considers the effectiveness of trajectories and prioritize them for unseen tasks. We also introduce Trajectory Abstraction, a mechanism that leverages MLLMs' summarization capabilities to represent trajectories with fewer tokens while preserving key information, enabling agents to better comprehend milestones in the trajectory. Experimental results across various environments demonstrate our method significantly improves task success rates in unseen scenes compared to baseline methods. This work presents a new paradigm for multimodal retrieval in embodied agents, by fine-tuning a general-purpose MLLM as the retriever to assess trajectory effectiveness. All benchmark task sets and simulator code modifications for action and observation spaces will be released.

Transformer-based language models have shown an excellent ability to effectively capture and utilize contextual information. Although various analysis techniques have been used to quantify and trace the contribution of single contextual cues to a target task such as subject-verb agreement or coreference resolution, scenarios in which multiple relevant cues are available in the context remain underexplored. In this paper, we investigate how language models handle gender agreement when multiple gender cue words are present, each capable of independently disambiguating a target gender pronoun. We analyze two widely used Transformer-based models: BERT, an encoder-based, and GPT-2, a decoder-based model. Our analysis employs two complementary approaches: context mixing analysis, which tracks information flow within the model, and a variant of activation patching, which measures the impact of cues on the model's prediction. We find that BERT tends to prioritize the first cue in the context to form both the target word representations and the model's prediction, while GPT-2 relies more on the final cue. Our findings reveal striking differences in how encoder-based and decoder-based models prioritize and use contextual information for their predictions.

Pre-trained Transformers inherently possess the characteristic of sparse activation, where only a small fraction of the neurons are activated for each token. While sparse activation has been explored through post-training methods, its potential in pre-training remains untapped. In this work, we first study how activation properties change during pre-training. Our examination reveals that Transformers exhibit sparse activation throughout the majority of the pre-training process while the activation correlation keeps evolving as training progresses. Leveraging this observation, we propose Switchable Sparse-Dense Learning (SSD). SSD adaptively switches between the Mixtures-of-Experts (MoE) based sparse training and the conventional dense training during the pre-training process, leveraging the efficiency of sparse training and avoiding the static activation correlation of sparse training. Compared to dense training, SSD achieves comparable performance with identical model size and reduces pre-training costs. Moreover, the models trained with SSD can be directly used as MoE models for sparse inference and achieve the same performance as dense models with up to $2\times$ faster inference speed. Codes are available at https://github.com/thunlp/moefication.

Semantic text embedding is essential to many tasks in Natural Language Processing (NLP). While black-box models are capable of generating high-quality embeddings, their lack of interpretability limits their use in tasks that demand transparency. Recent approaches have improved interpretability by leveraging domain-expert-crafted or LLM-generated questions, but these methods rely heavily on expert input or well-prompt design, which restricts their generalizability and ability to generate discriminative questions across a wide range of tasks. To address these challenges, we introduce CQG-MBQA (Contrastive Question Generation - Multi-task Binary Question Answering), a general framework for producing interpretable semantic text embeddings across diverse tasks. Our framework systematically generates highly discriminative, low cognitive load yes/no questions through the CQG method and answers them efficiently with the MBQA model, resulting in interpretable embeddings in a cost-effective manner. We validate the effectiveness and interpretability of CQG-MBQA through extensive experiments and ablation studies, demonstrating that it delivers embedding quality comparable to many advanced black-box models while maintaining inherently interpretability. Additionally, CQG-MBQA outperforms other interpretable text embedding methods across various downstream tasks. Text embedding is a cornerstone of Natural Language Processing (NLP), transforming texts--whether sentences, paragraphs, or full documents--into embedding vectors that capture their semantic meaning. In semantic embedding spaces, the similarity between texts is represented by the proximity of their embedding vectors, typically measured using distance measures like Euclidean distance, cosine distance, or inner product. Black-box text embedding methods, such as Sentence-BERT (Reimers & Gurevych, 2019), SimCSE (Gao et al., 2021), WhitenedCSE (Zhuo et al., 2023), and AnglE (Li & Li, 2024), excel at generating high-quality embeddings by training on vast amounts of data. These models are highly effective at capturing semantic similarities, making them indispensable for a variety of NLP tasks (Muennighoff et al., 2023). However, their black-box nature leaves the embeddings opaque to human users.