Vocabulary expansion (VE) is the de-facto approach to language adaptation of large language models (LLMs) by adding new tokens and continuing pre-training on target data. While this is effective for base models trained on unlabeled data, it poses challenges for chat models trained to follow instructions through labeled conversation data. Directly adapting the latter with VE on target unlabeled data may result in forgetting chat abilities. While ideal, target chat data is often unavailable or costly to create for low-resource languages, and machine-translated alternatives are not always effective. To address this issue, previous work proposed using a base and chat model from the same family. This method first adapts the base LLM with VE on target unlabeled data and then converts it to a chat model by adding a chat vector (CV) derived from the weight difference between the source base and chat models. We propose ElChat, a new language adaptation method for chat LLMs that adapts a chat model directly on target unlabeled data, without a base model. It elicits chat abilities by injecting information from the source chat model. ElChat offers more robust and competitive target language an
Chat communication is often fast-paced, creating the expectation of quick replies. While the timing of exchanges is known to foster closeness and enjoyment, it remains largely unexplored whether chat partners with strong ties reciprocate each other's response times. Using 3.4 million messages from 889 chats across 97 donations of anonymous WhatsApp and Instagram chats, we analyzed response times, their balance between chat partners, and its stability over time. To our knowledge, this is the first study to examine response speed as an expression of reciprocity, bridging a key aspect of online communication with a fundamental principle of social interactions. We found that around 70% of WhatsApp and 44% of Instagram messages were answered within five minutes, confirming the fast pace of instant messaging. Overall, the response speed between chat partners was similar. The response speed similarity was evident both in the overall response-time distributions of chat partners assessed with Jensen-Shannon distance and in the steep regression slopes (0.786 for WhatsApp and 0.796 for Instagram) linking one person's probability of responding within five minutes to the partner's corresponding
Large language models (LLMs) are expected to follow instructions from users and engage in conversations. Techniques to enhance LLMs' instruction-following capabilities typically fine-tune them using data structured according to a predefined chat template. Although chat templates are shown to be effective in optimizing LLM performance, their impact on safety alignment of LLMs has been less understood, which is crucial for deploying LLMs safely at scale. In this paper, we investigate how chat templates affect safety alignment of LLMs. We identify a common vulnerability, named ChatBug, that is introduced by chat templates. Our key insight to identify ChatBug is that the chat templates provide a rigid format that need to be followed by LLMs, but not by users. Hence, a malicious user may not necessarily follow the chat template when prompting LLMs. Instead, malicious users could leverage their knowledge of the chat template and accordingly craft their prompts to bypass safety alignments of LLMs. We develop two attacks to exploit the ChatBug vulnerability. We demonstrate that a malicious user can exploit the ChatBug vulnerability of eight state-of-the-art (SOTA) LLMs and effectively elic
Roblox is among the most popular online gaming platforms, used by hundreds of millions of users every day. A substantial portion of these users are underage, who are at a greater risk, where abusive users may utilize Roblox's real-time chat interface to make the initial contact with potential victims. Roblox employs automated chat moderation mechanisms to detect potentially abusive messages; however, to date, their effectiveness has not been independently investigated. Toward this goal, we collected approximately 2 million chat messages from four games across multiple age groups and analyzed them to evaluate the moderation system. These messages were collected from public game servers following ethical and legal norms as well as Roblox's terms of service. We use this corpus to qualitatively study which types of unsafe chats escape the moderation system and how policy-violating users evade the moderation system. Given the dataset's scale, it is prohibitively expensive to conduct qualitative content analysis manually. Therefore, we adopt a two-step approach. First, we manually labeled safe and unsafe messages (n=99.8K) and used them as a ground truth to evaluate four locally hosted s
Recently, FuseLLM introduced the concept of knowledge fusion to transfer the collective knowledge of multiple structurally varied LLMs into a target LLM through lightweight continual training. In this report, we extend the scalability and flexibility of the FuseLLM framework to realize the fusion of chat LLMs, resulting in FusionChat. FusionChat comprises two main stages. Firstly, we undertake knowledge fusion for structurally and scale-varied source LLMs to derive multiple target LLMs of identical structure and size via lightweight fine-tuning. Then, these target LLMs are merged within the parameter space, wherein we propose a novel method for determining the merging weights based on the variation ratio of parameter matrices before and after fine-tuning. We validate our approach using three prominent chat LLMs with diverse architectures and scales, namely NH2-Mixtral-8x7B, NH2-Solar-10.7B, and OpenChat-3.5-7B. Experimental results spanning various chat domains demonstrate the superiority of FusionChat-7B across a broad spectrum of chat LLMs at 7B and 34B scales, even surpassing GPT-3.5 (March) and approaching Mixtral-8x7B-Instruct.
Despite being prominent and ubiquitous, message-based interaction is limited in nonverbally conveying emotions. Besides emoticons or stickers, messaging users continue seeking richer options for affective communication. Recent research explored using chat balloons' shape and color to communicate emotional states. However, little work explored whether and how chat-balloon animations could be designed to convey emotions. We present the design of AniBalloons, 30 chat-balloon animations conveying Joy, Anger, Sadness, Surprise, Fear, and Calmness. Using AniBalloons as a research means, we conducted three studies to assess the animations' affect recognizability and emotional properties (N = 40), and probe how animated chat balloons would influence communication experience in typical scenarios including instant messaging (N = 72) and chatbot service (N = 70). Our exploration contributes a set of chat-balloon animations to complement non-nonverbal affective communication for a range of message-based interfaces, and empirical insights into how animated chat balloons might mediate particular conversation experiences (e.g., perceived interpersonal closeness, or chatbot personality).
Generative AI tools such as chatGPT are poised to change the way people engage with online information. Recently, Microsoft announced their "new Bing" search system which incorporates chat and generative AI technology from OpenAI. Google has announced plans to deploy search interfaces that incorporate similar types of technology. These new technologies will transform how people can search for information. The research presented here is an early investigation into how people make use of a generative AI chat system (referred to simply as chat from here on) as part of a search process, and how the incorporation of chat systems with existing search tools may effect users search behaviors and strategies. We report on an exploratory user study with 10 participants who used a combined Chat+Search system that utilized the OpenAI GPT-3.5 API and the Bing Web Search v5 API. Participants completed three search tasks. In this pre-print paper of preliminary results, we report on ways that users integrated AI chat into their search process, things they liked and disliked about the chat system, their trust in the chat responses, and their mental models of how the chat system generated responses.
Large Language Models (LLMs) are increasingly deployed across diverse domains, yet their vulnerability to jailbreak attacks, where adversarial inputs bypass safety mechanisms to elicit harmful outputs, poses significant security risks. While prior work has primarily focused on prompt injection attacks, these approaches often require resource-intensive prompt engineering and overlook other critical components, such as chat templates. This paper introduces TEMPLATEFUZZ, a fine-grained fuzzing framework that systematically exposes vulnerabilities in chat templates, a critical yet underexplored attack surface in LLMs. Specifically, TEMPLATEFUZZ (1) designs a series of element-level mutation rules to generate diverse chat template variants, (2) proposes a heuristic search strategy to guide the chat template generation toward the direction of amplifying the attack success rate (ASR) while preserving model accuracy, and (3) integrates an active learning-based strategy to derive a lightweight rule-based oracle for accurate and efficient jailbreak evaluation. Evaluated on twelve open-source LLMs across multiple attack scenarios, TEMPLATEFUZZ achieves an average ASR of 98.2% with only 1.1% a
Membership inference attacks (MIAs) test whether a target data record belongs to a system's private data, and have become a standard tool to measure privacy leakage in machine learning systems. Prior work has primarily focused on training corpora or retrieval databases. However, MIAs against agent memory have received less attention, even though such memory can contain sensitive user-agent interactions, retrieved facts, and user preferences. Therefore, in this work, we focus on chat agent memory MIAs, where an adversary infers whether a candidate memory unit belongs to the chat agent's memory store. We propose Multi-Recall Memory MIA (MRMMIA), a unified attack that utilizes multiple recall probes to the agent to extract the membership signal across black-box, gray-box, and white-box settings. Our experiments demonstrate that MRMMIA consistently outperforms baselines. Our results expose the privacy risk in agents and provide an initial evaluation framework for membership leakage in chat-agent memory systems.
As large language models (LLMs) become increasingly prevalent, understanding human-LLM interactions is emerging as a central priority in psychological research. Online experiments offer an efficient means to study human-LLM interactions, yet integrating LLMs into established survey platforms remains technically demanding, particularly when aiming for ecologically valid, real-time conversational experiences with strong experimental control. We introduce Simple Chat, an open-source, research-focused chat interface that streamlines LLM integration for platforms such as Qualtrics, oTree, and LimeSurvey, while presenting a unified participant experience across conditions. Simple Chat connects to both commercial providers and open-weights models, supports streaming responses to preserve conversational flow, and offers an administrative interface for fine-grained control of prompts and interface features. By reducing technical barriers, standardizing interfaces, and improving participant experience, Simple Chat helps advance the study of human-LLM interaction. In this article, we outline Simple Chat's key features, provide a step-by-step tutorial, and demonstrate its utility through two i
The complexities of chats pose significant challenges for machine translation models. Recognizing the need for a precise evaluation metric to address the issues of chat translation, this study introduces Multidimensional Quality Metrics for Chat Translation (MQM-Chat). Through the experiments of five models using MQM-Chat, we observed that all models generated certain fundamental errors, while each of them has different shortcomings, such as omission, overly correcting ambiguous source content, and buzzword issues, resulting in the loss of stylized information. Our findings underscore the effectiveness of MQM-Chat in evaluating chat translation, emphasizing the importance of stylized content and dialogue consistency for future studies.
M365 Copilot is used every week by millions of people across more than a million companies around the world as part of their workflows. Uniquely positioned in the AI landscape given its near-exclusive use for work purposes, M365 Copilot can offer a clear picture of how people use AI for work and where that usage may expand next. This paper characterizes that usage through direct classification of user interactions with M365 Copilot Chat. Based on an anonymized and privacy-preserving analysis of a sample of approximately 5.5 million sessions, we combine a learned classification of user intent with a classification of O*NET work activities done with M365 Copilot Chat. We find that M365 Copilot is emerging as an everyday assistant for knowledge work: writing dominates, but users also rely on it for information retrieval, analysis, decision making and strategizing, and evaluating and diagnosing programs and systems, among others. Information seeking tasks remain common, but time trends suggest a relative shift away from ``chat as search'' and toward content and communication-related work. Comparisons across occupational groupings and to work done in the labor market further show that u
Recently, the development of open-source large language models (LLMs) has advanced rapidly. Nevertheless, due to data constraints, the capabilities of most open-source LLMs are primarily focused on English. To address this issue, we introduce the concept of $\textit{chat vector}$ to equip pre-trained language models with instruction following and human value alignment via simple model arithmetic. The chat vector is derived by subtracting the weights of a pre-trained base model (e.g. LLaMA2) from those of its corresponding chat model (e.g. LLaMA2-chat). By simply adding the chat vector to a continual pre-trained model's weights, we can endow the model with chat capabilities in new languages without the need for further training. Our empirical studies demonstrate the superior efficacy of the chat vector from three different aspects: instruction following, toxicity mitigation, and multi-turn dialogue. Moreover, to showcase the adaptability of our approach, we extend our experiments to encompass various languages, base models, and chat vectors. The results underscore the chat vector's simplicity, effectiveness, and wide applicability, making it a compelling solution for efficiently ena
While training large language models (LLMs) from scratch can indeed lead to models with distinct capabilities and strengths, it incurs substantial costs and may lead to redundancy in competencies. Knowledge fusion aims to integrate existing LLMs of diverse architectures and capabilities into a more potent LLM through lightweight continual training, thereby reducing the need for costly LLM development. In this work, we propose a new framework for the knowledge fusion of chat LLMs through two main stages, resulting in FuseChat. Firstly, we conduct pairwise knowledge fusion on source chat LLMs of varying structures and scales to create multiple target LLMs with identical structure and size via lightweight fine-tuning. During this process, a statistics-based token alignment approach is introduced as the cornerstone for fusing LLMs with different structures. Secondly, we merge these target LLMs within the parameter space, where we propose a novel method for determining the merging coefficients based on the magnitude of parameter updates before and after fine-tuning. We implement and validate FuseChat using six prominent chat LLMs with diverse architectures and scales, including OpenChat
Large Language Models (LLMs) such as ChatGPT and Llama have become prevalent in real-world applications, exhibiting impressive text generation performance. LLMs are fundamentally developed from a scenario where the input data remains static and unstructured. To behave interactively, LLM-based chat systems must integrate prior chat history as context into their inputs, following a pre-defined structure. However, LLMs cannot separate user inputs from context, enabling chat history tampering. This paper introduces a systematic methodology to inject user-supplied history into LLM conversations without any prior knowledge of the target model. The key is to utilize prompt templates that can well organize the messages to be injected, leading the target LLM to interpret them as genuine chat history. To automatically search for effective templates in a WebUI black-box setting, we propose the LLM-Guided Genetic Algorithm (LLMGA) that leverages an LLM to generate and iteratively optimize the templates. We apply the proposed method to popular real-world LLMs including ChatGPT and Llama-2/3. The results show that chat history tampering can enhance the malleability of the model's behavior over t
Chat interfaces for intelligent tutoring systems (ITSs) enable interactivity and flexibility. However, when students interact with chat interfaces, they expect dialogue-driven navigation from the system and can express frustration and disinterest if this is not provided. Intent detection systems help students navigate within an ITS, but detecting students' intent during open-ended dialogue is challenging. We designed an intent detection system in a chatbot ITS, classifying a student's intent between continuing the current lesson or switching to a new lesson. We explore the utility of four machine learning approaches for this task - including both conventional classification approaches and fine-tuned large language models - finding that using an intent classifier introduces trade-offs around implementation cost, accuracy, and prediction time. We argue that implementing intent detection in chat interfaces can reduce frustration and support student learning.
Chat assistants increasingly integrate web search functionality, enabling them to retrieve and cite external sources. While this promises more reliable answers, it also raises the risk of amplifying misinformation from low-credibility sources. In this paper, we introduce a novel methodology for evaluating assistants' web search behavior, focusing on source credibility and the groundedness of responses with respect to cited sources. Using 100 claims across five misinformation-prone topics, we assess GPT-4o, GPT-5, Perplexity, and Qwen Chat. Our findings reveal differences between the assistants, with Perplexity achieving the highest source credibility, whereas GPT-4o exhibits elevated citation of non-credibility sources on sensitive topics. This work provides the first systematic comparison of commonly used chat assistants for fact-checking behavior, offering a foundation for evaluating AI systems in high-stakes information environments.
Link prediction in heterogeneous networks is crucial for understanding the intricacies of network structures and forecasting their future developments. Traditional methodologies often face significant obstacles, including over-smoothing-wherein the excessive aggregation of node features leads to the loss of critical structural details-and a dependency on human-defined meta-paths, which necessitate extensive domain knowledge and can be inherently restrictive. These limitations hinder the effective prediction and analysis of complex heterogeneous networks. In response to these challenges, we propose the Contrastive Heterogeneous grAph Transformer (CHAT). CHAT introduces a novel sampling-based graph transformer technique that selectively retains nodes of interest, thereby obviating the need for predefined meta-paths. The method employs an innovative connection-aware transformer to encode node sequences and their interconnections with high fidelity, guided by a dual-faceted loss function specifically designed for heterogeneous network link prediction. Additionally, CHAT incorporates an ensemble link predictor that synthesizes multiple samplings to achieve enhanced prediction accuracy.
Reinforcement learning with verifiable rewards (RLVR) improves language model reasoning by using rule-based rewards in verifiable domains such as mathematics and code. However, RLVR leads to limited generalization for open-ended tasks -- such as writing outline essays or making meal plans -- where humans reason routinely. This paper shows that the RLVR paradigm is effective beyond verifiable domains, and introduces **RL** with **M**odel-rewarded **T**hinking (**RLMT**) for general-purpose chat capabilities. Using diverse real-world prompts, RLMT requires LMs to generate long CoT reasoning before response, and optimizes them with online RL against a preference-based reward model used in RLHF. Across 40 training runs on Llama-3.1-8B and Qwen-2.5-7B (both base and instruct) and multiple optimization algorithms (DPO, PPO, and GRPO), RLMT consistently outperforms standard RLHF pipelines. This includes substantial gains of 3-7 points on three chat benchmarks (AlpacaEval2, WildBench, and ArenaHardV2), along with 1-3 point improvements on other tasks like creative writing and general knowledge. Our best 8B model surpasses GPT-4o in chat and creative writing and rivals Claude-3.7-Sonnet (Th
Chat models, such as ChatGPT, have shown impressive capabilities and have been rapidly adopted across numerous domains. However, these models are only accessible through a restricted API, creating barriers for new research and progress in the field. We propose a pipeline that can automatically generate a high-quality multi-turn chat corpus by leveraging ChatGPT to engage in a conversation with itself. Subsequently, we employ parameter-efficient tuning to enhance LLaMA, an open-source large language model. The resulting model, named Baize, demonstrates good performance in multi-turn dialogues with guardrails that minimize potential risks. Furthermore, we propose a new technique called Self-Distill with Feedback, to further improve the performance of the Baize models with feedback from ChatGPT. The Baize models and data are released for research purposes only at https://github.com/project-baize/baize-chatbot. An online demo is also available at https://huggingface.co/spaces/project-baize/chat-with-baize.