Austria

Large language model (LLM) parameters store massive amounts of pre-training knowledge, but it is unclear how to continually update them to reflect changes in the real world. Both dataset and methodology works in existing literature focus on the entity-substitution framework which does not capture the breadth of complex real world updates. In this paper, we first introduce Knowledge Update Playground (KUP) to automatically construct a training corpus with realistic and complex updates to study this problems. Next, we introduce a novel approach, memory recall training (MRT), that trains the model to explicitly surface and condition on &quot;recalled&quot; knowledge both at train and test time. Our results show that the KUP benchmark is challenging for existing continued pre-training (CPT) approaches; LLMs after CPT still prefer outdated knowledge &gt;40% of the time. Using our proposed MRT training approach significantly improves performance across datasets and evaluation settings (58.9 -&gt; 71.0 for LLaMA-3.1-8B). Overall, we show that our KUP dataset, particularly an indirect probing evaluation setting, serves as a challenging testbed to facilitate future research.

ACL 2025

Memorization vs. Reasoning: Updating LLMs with New Knowledge

knowledge update

knowledge conflicts

continued pre-training

knowledge editing

### Welcome to The 63rd Annual Meeting of the Association for Computational Linguistics (ACL 2025)

Message from the General Chair: 
*It is my great pleasure and honor to welcome you to the 63rd Annual Meeting of the Association for Computational Linguistics (ACL 2025), held in beautiful Vienna, Austria, from July 27 to August 1, 2025. ACL2025continues our field’s tradition of excellence in scholarship, innovation, and inclusivity, and I am deeply grateful to the many volunteers who have worked tirelessly to bring this event to life.* 
[Read more](https://drive.google.com/file/d/1GI_hvOpjswAuYdUTromfeDiPpCcqidwg/view?usp=sharing)

To access this event page, you need to log in with the **email address you registered with**. Access credentials will be sent to your email from Underline - subject line "Welcome to ACL 2025". Please be sure to check your spam email folder if you do not see an email confirmation right away.

Please log in

To access this event page, you need to log in with the **email address you registered with**. 

Welcome to The 63rd Annual Meeting of the Association for Computational Linguistics (ACL 2025)

Large language model (LLM) parameters store massive amounts of pre-training knowledge, but it is unclear how to continually update them to reflect changes in the real world. Both dataset and methodology works in existing literature focus on the entity-substitution framework which does not capture the breadth of complex real world updates. In this paper, we first introduce Knowledge Update Playground (KUP) to automatically construct a training corpus with realistic and complex updates to study this problems. Next, we introduce a novel approach, memory recall training (MRT), that trains the model to explicitly surface and condition on "recalled" knowledge both at train and test time. Our results show that the KUP benchmark is challenging for existing continued pre-training (CPT) approaches; LLMs after CPT still prefer outdated knowledge >40% of the time. Using our proposed MRT training approach significantly improves performance across datasets and evaluation settings (58.9 -> 71.0 for LLaMA-3.1-8B). Overall, we show that our KUP dataset, particularly an indirect probing evaluation setting, serves as a challenging testbed to facilitate future research.

poster

As AI advances in text generation, human trust in AI generated content remains constrained by biases that go beyond concerns of accuracy. This study explores how bias shapes the perception of AI versus human generated content. Through three experiments involving text rephrasing, news article summarization, and persuasive writing, we investigated how human raters respond to labeled and unlabeled content. While the raters could not differentiate the two types of texts in the blind test, they overwhelmingly favored content labeled as "Human Generated," over those labeled "AI Generated," by a preference score of over 30%. We observed the same pattern even when the labels were deliberately swapped. This human bias against AI has broader societal and cognitive implications, as it undervalues AI performance. This study highlights the limitations of human judgment in interacting with AI and offers a foundation for improving human-AI collaboration, especially in creative fields.

Human Bias in the Face of AI: The Role of Human Judgement in AI Generated Text Evaluation

This study explores a novel approach to enhance the performance of Large Language Models (LLMs) through the optimization of input data within prompts. While previous research has primarily focused on refining instruction components and augmenting input data with in-context examples, our work investigates the potential benefits of optimizing the input data itself. We introduce a two-pronged strategy for input data optimization: content engineering and structural reformulation. Content engineering involves imputing missing values, removing irrelevant attributes, and enriching profiles by generating additional information inferred from existing attributes. Subsequent to content engineering, structural reformulation is applied to optimize the presentation of the modified content to LLMs, given their sensitivity to input format. Our findings suggest that these optimizations can significantly improve the performance of LLMs in various tasks, offering a promising avenue for future research in prompt engineering. The source code is available at https://github.com/glin2229/Automatic-Data-Optimization.

ADO: Automatic Data Optimization for Inputs in LLM Prompts

Internal world models (WMs) enable agents to understand the world's state and predict transitions, serving as the basis for advanced deliberative reasoning. Recent large Vision-Language Models (VLMs), such as GPT-4o and Gemini, exhibit potential as general-purpose WMs. While the latest studies have evaluated and shown limitations in specific capabilities such as visual understanding, a systematic evaluation of VLMs' fundamental WM abilities remains absent. Drawing on comparative psychology and cognitive science, we propose a two-stage framework that assesses Perception (visual, spatial, temporal, quantitative, and motion) and Prediction (mechanistic simulation, transitive inference, compositional inference) to provide an atomic evaluation of VLMs as WMs. Guided by this framework, we introduce WM-ABench, a large-scale benchmark comprising 23 fine-grained evaluation dimensions across 6 diverse simulated environments with controlled counterfactual simulations. Through 517 controlled experiments on 11 latest commercial and open-source VLMs, we find that these models exhibit striking limitations in basic world model abilities. For instance, all models perform at near-random accuracy when distinguishing motion trajectories. Additionally, they lack disentangled understanding---e.g., they tend to believe blue objects move faster than green ones. More rich results and analyses reveal significant gaps between VLMs and human-level world modeling.

Do Vision-Language Models Have Internal World Models? Towards an Atomic Evaluation

Investigating bias in large language models (LLMs) is crucial for developing trustworthy AI. While prompt-based through prompt engineering is common, its effectiveness relies on the assumption that models inherently understand biases. Our study systematically analyzed this assumption using the BBQ and StereoSet benchmarks on both open-source models as well as commercial GPT model. Experimental results indicate that prompt-based is often superficial; for instance, the Llama2-7B-Chat model misclassified over 90% of unbiased content as biased, despite achieving high accuracy in identifying bias issues on the BBQ dataset. Additionally, specific evaluation and question settings in bias benchmarks often lead LLMs to choose "evasive answers", disregarding the core of the question and the relevance of the response to the context. Moreover, the apparent success of previous methods may stem from flawed evaluation metrics. Our research highlights a potential "false prosperity" in prompt-base efforts and emphasizes the need to rethink bias evaluation metrics to ensure truly trustworthy AI. We will release our data and code upon acceptance.

Rethinking Prompt-based Debiasing in Large Language Model

Large language models (LLMs) demonstrate impressive results in natural language processing tasks, but require a significant amount of computational and memory resources. Structured matrix representations is a promising way for reducing the number of parameters of these models. However, it seems unrealistic to expect that weight matrices of pretrained models can be accurately represented by structured matrices without any fine-tuning. To overcome this issue, we utilize the fact that LLM output is invariant under certain orthogonal transformations of weight matrices. This insight can be leveraged to identify transformations that significantly improve the compressibility of weights within the structured classes. The proposed approach is applicable to various types of structured matrices that support efficient projection operations.

ProcrustesGPT: Compressing LLMs with Structured Matrices and Orthogonal Transformations

Lightweight Large Language Models (LwLLMs) are reduced-parameter, optimized models designed to run efficiently on consumer-grade hardware, offering significant advantages in resource efficiency, cost-effectiveness, and data privacy. However, these models often struggle with limited inference and reasoning capabilities, which restrict their performance on complex tasks and limit their practical applicability. Moreover, existing prompt optimization methods typically rely on extensive manual effort or the meta-cognitive abilities of state-of-the-art LLMs, making them less effective for LwLLMs.

To address these challenges, we introduce DeBoP, a new Direct Behavior Optimization Paradigm, original from the Chain-of-Thought (CoT) prompting technique. Unlike CoT Prompting, DeBoP is an automatic optimization method, which focuses on the optimization directly on the behavior of LwLLMs. In particular, DeBoP transforms the optimization of complex prompts into the optimization of discrete, quantifiable execution sequences using a gradient-free Monte Carlo Tree Search. We evaluate DeBoP on seven challenging tasks where state-of-the-art LLMs excel but LwLLMs generally underperform. Experimental results demonstrate that DeBoP significantly outperforms recent prompt optimization methods on most tasks. In particular, DeBoP-optimized LwLLMs surpass GPT-3.5 on most tasks while reducing computational time by approximately 60% compared to other automatic prompt optimization methods.

Direct Behavior Optimization: Unlocking the Potential of Lightweight LLMs

A NOTAM or NOtice To AirMen is a crucial notice for different aviation stakeholders, particularly flight crews. It delivers essential notifications about abnormal conditions of Aviation System components such as changes to facilities, hazards, service, procedure that are not known far enough in advance to be publicized through other means. NOTAM messages are short, contain acronyms, and look cryptic in most of the cases. Writing and understanding these messages put heavy cognitive load on its end users. In this work, we take up the task of translating NOTAMs into English natural language using LLMs. Since NOTAMs do not adhere to English grammar rules and have their own decoding rules, large language models (LLMs) cannot translate them without effective prompting. In this paper, we develop a framework to come up with effective prompts to achieve the translations. Our approach uses context-aware semantic prompting techniques, paired with domain-specific rules, to improve the accuracy and clarity of translations. The framework is evaluated using comprehensive experiments (6 LLMs of varying sizes, and with 5 different prompting setups for each) and eight evaluation metrics measuring different aspects of the translation. The results demonstrate that our methodology can produce clear translations that accurately convey the information contained in NOTAMs.

Semantics-aware prompting for translating NOtices To AirMen

Vision-language models (VLMs) integrate textual and visual information, enabling the model to process visual inputs and leverage visual information to generate predictions. Such models are demanding for tasks such as visual question answering, image captioning and visual grounding. However, some recent work found that %visual information is not fully utilized by language models. That is, VLMs often rely heavily on textual information, ignoring visual information, but are still able to achieve competitive performance in vision-language (VL) tasks. This survey reviews modality collapse analysis work to provide insights into the reason for this unintended behaviour. It also reviews probing studies for fine-grained vision-language understanding, presenting current findings on information encoded in VL representations and highlight potential directions for future research.

Can VLMs Actually See and Read? A Survey on Modality Collapse in Vision-Language Models

The growing use of large language models (LLMs) for text generation has led to widespread concerns about AI-generated content detection. However, an overlooked challenge is AI-polished text, where human-written content undergoes subtle refinements using AI tools. This raises a critical question: should minimally polished text be classified as AI-generated? Such classification can lead to false plagiarism accusations and misleading claims about AI prevalence in online content. In this study, we systematically evaluate eleven state-of-the-art AI-text detectors using our 'AI-Polished-Text Evaluation (APT-Eval)' dataset, which contains 15K samples refined at varying AI-involvement levels. Our findings reveal that detectors frequently flag even minimally polished text as AI-generated, struggle to differentiate between degrees of AI involvement, and exhibit biases against older and smaller models. These limitations highlight the urgent need for more nuanced detection methodologies.

Almost AI, Almost Human: The Challenge of Detecting AI-Polished Writing

Large Action Models (LAMs) for AI Agents offer incredible potential but face challenges due to the need for high-quality training data, especially for multi-steps tasks that involve planning, executing tool calls, and responding to feedback. To address these issues, we present LAM SIMULATOR, a comprehensive framework designed for online exploration of agentic tasks with high-quality feedback. Our framework features a dynamic task query generator, an extensive collection of tools, and an interactive environment where Large Language Model (LLM) Agents can call tools and receive real-time feedback. This setup enables LLM Agents to explore and solve tasks independently and potentially come up with multiple approaches to tackle any given task. Generated data are then used to create high-quality training datasets for LAMs.

Our research shows that LAM SIMULATOR enables LLM Agents to autonomously solve tasks while automating the creation of high-quality training data. Models trained with these self-generated datasets demonstrated significant performance gains, showing up to a 49.3% improvement over their own baselines. This was especially evident in experiments conducted with the ToolBench and CRMArena environments. The process requires minimal human input during dataset creation, highlighting the LAM SIMULATOR's efficiency and effectiveness in speeding up AI agents' development.

Downloads

Next from ACL 2025

Human Bias in the Face of AI: The Role of Human Judgement in AI Generated Text Evaluation

Similar lecture

How Do LLMs Acquire New Knowledge? A Knowledge Circuits Perspective on Continual Pre-Training