Austria

What makes a difference in the post-training of LLMs? We investigate the training patterns of different layers in large language models (LLMs) through the lens of the gradient. We are specifically interested in how fast vs. slow thinking affects the layer-wise gradients, given the recent popularity of training LLMs on reasoning paths such as chain-of-thoughts (CoT) and process rewards. In our study, fast thinking without CoT leads to larger gradients and larger differences of gradients across layers than slow thinking (Detailed CoT), indicating the learning stability brought by the latter. Additionally, we study whether the gradient patterns can reflect the correctness of responses when training different LLMs using slow vs. fast thinking paths. The results show that the gradients of slow thinking can distinguish correct and irrelevant reasoning paths. As a comparison, we conduct similar gradient analyses on non-reasoning knowledge learning tasks, on which, however, trivially increasing the response length does not lead to similar behaviors of slow thinking. Our study strengthens fundamental understandings of LLM training and sheds novel insights on its efficiency and stability, which pave the way towards building a generalizable System-2 agent.

ACL 2025

What Happened in LLMs Layers when Trained for Fast vs. Slow Thinking: A Gradient Perspective

gradient analysis

interpretability

reasoning

The purpose of the AAAI conference is to promote research in artificial intelligence (AI) and scientific exchange among AI researchers, practitioners, scientists, and engineers in affiliated disciplines. 

AAAI 2022

technical paper

### 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)

Code LLMs have been widely used in various domains, including code generation, logical reasoning, and agent systems. However, open-access code LLMs mostly only release weights, lacking key features such as reproducible data pipelines and transparent training protocols, which are crucial for advancing deeper, more reliable investigations. To address the gap, we introduce OpenCoder, a top-tier code LLM that not only achieves performance comparable to leading models but also serves as an ``open cookbook'' for the research community. Unlike most prior efforts, we release not only model weights and inference code, but also the reproducible training data, complete data processing pipeline, rigorous experimental ablation results, and detailed training protocols for open scientific research. Our work identifies the key ingredients for building a top-tier code LLM: optimized heuristic rules for data cleaning and deduplication, effective recall of code-related text corpus, and high-quality synthetic data for both annealing and supervised fine-tuning stages. By offering this level of openness, we aim to broaden access to all aspects of a top-tier code LLM, with OpenCoder serving as both a powerful model and an open foundation to accelerate research and enable reproducible advancements in code intelligence. The released resource is available at https://opencoder-llm.github.io.

OpenCoder: The Open Cookbook for Top-Tier Code Large Language Models

Many sensitive domains — such as the clinical domain — lack widely available datasets due to privacy risks. The increasing generative capabilities of large language models (LLMs) have made synthetic datasets a viable path forward. In this study, we domain-adapt LLMs to the clinical domain and generate synthetic clinical texts that are machine-annotated with tags for personally identifiable information using capable encoder-based NER models. The synthetic corpora are then used to train synthetic NER models. The results show that training NER models using synthetic corpora incurs only a small drop in predictive performance. The limits of this process are investigated in a systematic ablation study — using both Swedish and Spanish data. Our analysis shows that smaller datasets can be sufficient for domain-adapting LLMs for data synthesis. Instead, the effectiveness of this process is almost entirely contingent on the performance of the machine-annotating NER models trained using the original data.

Data-Constrained Synthesis of Training Data for De-Identification

Despite the increasing use of large language models (LLMs) for context-grounded tasks like summarization and question-answering, understanding what makes an LLM produce a certain response is challenging. We propose Multi-Level Explanations for Generative Language Models (MExGen), a technique to provide explanations for context-grounded text generation. MExGen assigns scores to parts of the context to quantify their influence on the model's output. It extends attribution methods like LIME and SHAP to LLMs used in context-grounded tasks where (1) inference cost is high, (2) input text is long, and (3) the output is text. We conduct a systematic evaluation, both automated and human, of perturbation-based attribution methods for summarization and question answering. The results show that our framework can provide more faithful explanations of generated output than available alternatives, including LLM self-explanations. We open-source code for MExGen as part of the ICX360 toolkit: https://github.com/IBM/ICX360.

Multi-Level Explanations for Generative Language Models

Language models excel at following instructions but often struggle with the collaborative aspects of conversation that humans naturally employ. This limitation in grounding---the process by which conversation participants establish mutual understanding---can lead to outcomes ranging from frustrated users to serious consequences in high-stakes scenarios. To systematically study grounding challenges in human-LLM interactions, we analyze logs from three human-assistant datasets: WildChat, MultiWOZ, and Bing Chat. We develop a taxonomy of grounding acts and build models to annotate and forecast grounding behavior. Our findings reveal significant differences in human-human and human-LLM grounding: LLMs were three times less likely to initiate clarification and sixteen times less likely to provide follow-up requests than humans. Additionally, we find that early grounding failures predict later interaction breakdowns. Building on these insights, we introduce Rifts, a benchmark derived from publicly available LLM interaction data containing situations where LLMs fail to initiate grounding. We note that current frontier models perform poorly on Rifts, highlighting the need to reconsider how we train and prompt LLMs for human interaction. To this end, we develop a preliminary intervention aimed at mitigating grounding failures.

Navigating Rifts in Human-LLM Grounding: Study and Benchmark

Large-scale deployment of large language models (LLMs) in various applications, such as chatbots and virtual assistants, requires LLMs to be culturally sensitive to the user to ensure inclusivity.  Culture has been widely studied in psychology and anthropology, and there has been a recent surge in research on making LLMs more culturally inclusive, going beyond multilinguality and building on findings from psychology and anthropology. In this paper, we survey efforts towards incorporating cultural awareness into text-based and multimodal LLMs. We start by defining cultural awareness in LLMs, taking definitions of culture from the anthropology and psychology literature as a point of departure. We then examine methodologies adopted for creating cross-cultural datasets, strategies for cultural inclusion in downstream tasks, and methodologies that have been used for benchmarking cultural awareness in LLMs. Further, we discuss the ethical implications of cultural alignment, the role of Human-Computer Interaction (HCI) in driving cultural inclusion in LLMs, and the role of cultural alignment in driving social science research. We finally provide pointers to future research based on our findings about gaps in the literature.

Survey of Cultural Awareness in Language Models: Text and Beyond

Accurate mapping of queries to product categories is crucial for efficient retrieval and ranking of relevant products in e-commerce search. Conventionally, such query classification models rely on supervised learning using historical user interactions, but their effectiveness diminishes in cold-start scenarios, where new categories or products lack sufficient training data. This results in poor query-to-category mappings, negatively affecting retrieval and ranking. Synthetic query generation has emerged as a promising solution by augmenting training data; however, existing methods do not incorporate feedback from the query relevance model, limiting their ability to generate queries that enhance product retrieval. To address this, we propose an adversarial reinforcement learning framework that optimizes an LLM-based generator to expose weaknesses in query classification models. The generator produces synthetic queries to augment the classifier’s training set, ultimately improving its performance. Additionally, we introduce a structured reward signal to ensure stable training. Experiments on public datasets show an average PR-AUC improvement of +1.82\% on benchmarks and +3.26\% on a proprietary dataset, demonstrating the framework’s effectiveness in enhancing query classification and mitigating cold-start challenges.

Reinforcement Learning for Adversarial Query Generation to Enhance Relevance in Cold-Start Product Search

Speaker diarization aims to segment an audio stream into homogeneous partitions based on speaker identity, playing a crucial role in speech comprehension and analysis. Mainstream speaker diarization systems rely only on acoustic information, making the task particularly challenging in complex acoustic environments in real-world applications. Recently, significant efforts have been devoted to audio-visual or audio-semantic multimodal modeling to enhance speaker diarization performance; however, these approaches still struggle to address the complexities of speaker diarization on spontaneous and unstructured multi-party conversations. To fully exploit meaningful dialogue patterns, we propose a novel multimodal approach that jointly utilizes audio, visual, and semantic cues to enhance speaker diarization. Our approach structures visual cues among active speakers and semantic cues in spoken content into a cohesive format known as pairwise constraints, and employs a semi-supervised clustering technique based on pairwise constrained propagation. Extensive experiments conducted on multiple multimodal datasets demonstrate that our approach effectively integrates audio-visual-semantic information into the clustering process for acoustic speaker embeddings and consistently outperforms state-of-the-art speaker diarization methods, while largely preserving the overall system framework.

Integrating Audio, Visual, and Semantic Information for Enhanced Multimodal Speaker Diarization on Multi-party Conversation

Often when we interact with a first-person account of events, we consider whether or not the narrator, the primary speaker of the text, is reliable. In this paper, we propose using computational methods to identify unreliable narrators, i.e. those who unintentionally misrepresent information. Borrowing literary theory from narratology to define different types of unreliable narrators based on a variety of textual phenomena, we present TUNa, a human-annotated dataset of narratives from multiple domains, including blog posts, subreddit posts, hotel reviews, and works of literature. We define classification tasks for intra-narrational, inter-narrational, and inter-textual unreliabilities and analyze the performance of popular open-weight and proprietary LLMs for each. We propose learning from literature to perform unreliable narrator classification on real-world text data. To this end, we experiment with few-shot, fine-tuning, and curriculum learning settings. Our results show that this task is very challenging, and there is potential for using LLMs to identify unreliable narrators. We release our expert-annotated dataset and code at https://github.com/adbrei/unreliable-narrators and invite future research in this area.

Classifying Unreliable Narrators with Large Language Models

State-of-the-art automatic event detection struggles with interpretability and adaptability to evolving large-scale key events---unlike episodic structures, which excel in these areas. Often overlooked, episodes represent cohesive clusters of core entities performing actions at a specific time and location; a partially ordered sequence of episodes can represent a key event. This paper introduces a novel task, **episode detection**, which identifies episodes within a news corpus of key event articles. Detecting episodes poses unique challenges, as they lack explicit temporal or locational markers and cannot be merged using semantic similarity alone. While large language models (LLMs) can aid with these reasoning difficulties, they suffer with long contexts typical of news corpora. To address these challenges, we introduce **EpiMine**, an unsupervised framework that identifies a key event's candidate episodes by leveraging natural episodic partitions in articles, estimated through shifts in discriminative term combinations. These candidate episodes are more cohesive and representative of true episodes, synergizing with LLMs to better interpret and refine them into final episodes. We apply EpiMine to our three diverse, real-world event datasets annotated at the episode level, where it achieves a 59.2% average gain across all metrics compared to baselines.

Synergizing Unsupervised Episode Detection with LLMs for Large-Scale News Events

Automatic language identification is a core problem of many Natural Language
Processing (NLP) pipelines. A wide variety of architectures and benchmarks have
been proposed with often near-perfect performance. Although previous
studies have focused on certain challenging setups (i.e. cross-domain, short
inputs), a systematic comparison is missing. We propose a benchmark that 
allows us to test for the effect of input size, training data size, domain, number of
languages, scripts, and language families on performance. We evaluate
five popular models on this benchmark and identify which open challenges
remain for this task as well as which architectures achieve robust performance. We
find that cross-domain setups are the most challenging (although arguably most
relevant), and that number of languages, variety in scripts, and variety in
language families have only a small impact on performance. We also contribute
practical takeaways: training with 1,000 instances per language and a maximum
input length of 100 characters is enough for robust language identification.
Based on our findings, we train an accurate (94.41%) multi-domain language
identification model on 2,034 languages, for which we also provide an analysis
of the remaining errors.

Downloads

Next from ACL 2025

OpenCoder: The Open Cookbook for Top-Tier Code Large Language Models

Similar lecture

Thinking Fast and Slow in AI

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES