China

Large language models (LLMs) have demonstrated remarkable capabilities in natural language processing tasks. However, they remain vulnerable to semantic inconsistency, where minor, semantically equivalent variations in input formatting result in divergent predictions. Our comprehensive evaluation reveals that this brittleness persists even in state-of-the-art models such as GPT-4o, posing a serious challenge to their reliability. Through a mechanistic analysis, we attribute this phenomenon to deep representational failures, whereby semantic-equivalent input changes induce instability in the model’s internal representations. We further examine standard mitigation strategies and uncover their fundamental limitations. In particular, even direct fine-tuning on format variations frequently fails to yield format-invariant semantic representations, highlighting the difficulty of the problem. By explaining the failure of existing methods through our representational diagnosis, we underscore the need for representation-aware strategies to achieve robust and reliable LLM behavior.

EMNLP 2025

When Format Changes Meaning: Investigating Semantic Inconsistency of Large Language Models

semantic inconsistency

interpretability

generalization

### Welcome to 2025 Annual Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics

Welcome to the 2025 meeting of the Nations of the Americas Chapter of the Association for Computational Linguistics! I am proud to help organize the first NAACL conference to carry the new name of our organization, one that emphasizes inclusion for all of the Americas. I am also pleased to welcome you to Albuquerque, New Mexico, a state whose unique blend of cultural influences will make for an excellent backdrop for NAACL 2025, especially with this year’s special theme on NLP in a Multicultural World. 
**[Continue reading...](https://drive.google.com/file/d/1jX-qGhqVSZZCIrAnJaz798pflu5Irrdn/view?usp=sharing)**

*- Colin Cherry, Google, NAACL 2025 General Chair* 

[![](https://assets.underline.io/markdown_image/1/image/b087f8a4dc5816d6e1a6514e59c59ac3.png)](https://drive.google.com/file/d/1T96GzPqObXrMln2BMByCSXTSizjTg69P/view?usp=sharing)

You need to log in with the email address you registered with. Access credentials have been sent to your email. 

Please be sure to check your spam and other email folders if you do not see an email confirmation right away.

Please log in to explore this event.

To access NAACL 2025 event page you are required to register. Please follow [**this link**](https://2025.naacl.org/registration/registration/) to register. Access will depend on your registration type.

Please register!

NAACL 2025

Welcome to 2025 Annual Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics

poster

## Welcome!
"I am excited to welcome you to this year’s edition of the Conference on Empirical Methods in Natural Language Processing! Importantly, it marks the 30th edition of EMNLP. With over 8,000 submissions, more than 3,000 accepted papers, and thousands of attendees, we have come a long way from that first
workshop, which had 14 accepted papers. As the field looks ahead, Suzhou is the fitting location for celebrating this milestone: rooted in a long literary tradition, yet modern and forward-looking, and home to a large share of the NLP community." 

*Message from the General Chair, Dirk Hovy*

Celebrate 30 Years of EMNLP! 
EMNLP 2025 will be held in Suzhou, China from November 5th to November 9th, 2025.

Existing LLM red-teaming approaches prioritize high attack success rate, often resulting in high-perplexity prompts. This focus overlooks low-perplexity attacks that are more difficult to filter, more likely to arise during benign usage, and more impactful as negative downstream training examples. In response, we introduce ASTPrompter, a single-step optimization method that uses contrastive preference learning to train an attacker to maintain low perplexity while achieving a high attack success rate (ASR). ASTPrompter achieves an attack success rate 5.1 times higher on Llama-8.1B while using inputs that are 2.1 times more likely to occur according to the frozen LLM. Furthermore, our attack transfers to Mistral-7B, Qwen-7B, and TinyLlama in both black- and white-box settings. Lastly, by tuning a single hyperparameter in our method, we discover successful attack prefixes along an efficient frontier between ASR and perplexity, highlighting perplexity as a previously under-considered factor in red-teaming.

ASTPrompter: Preference-Aligned Automated Language Model Red-Teaming to Generate Low-Perplexity Unsafe Prompts

The absence of explicit communication channels between automated vehicles (AVs) and other road users requires the use of external Human-Machine Interfaces (eHMIs) to convey messages effectively in uncertain scenarios. Currently, most eHMI studies employ predefined text messages and manually designed actions to perform these messages, which limits the real-world deployment of eHMIs, where adaptability in dynamic scenarios is essential. Given the generalizability and versatility of large language models (LLMs), they could potentially serve as automated action designers for the message-action design task. To validate this idea, we make three contributions: (1) We propose a pipeline that integrates LLMs and 3D renderers, using LLMs as action designers to generate executable actions for controlling eHMIs and rendering action clips. (2) We collect a user-rated Action-Design Scoring dataset comprising a total of 320 action sequences for eight intended messages and four representative eHMI modalities. The dataset validates that LLMs can translate intended messages into actions close to a human level, particularly for reasoning-enabled LLMs. (3) We introduce two automated raters, Action Reference Score (ARS) and Vision-Language Models (VLMs), to benchmark 18 LLMs, finding that the VLM aligns with human preferences yet varies across eHMI modalities. The source code, prompts, Blender scenarios, and rendered clips are available at https://anonymous.4open.science/r/eHMI_action_design/.

Automating eHMI Action Design with LLMs for Automated Vehicle Communication

Large Language Model (LLM)-based self-refinement methods have significantly enhanced data analysis performance, especially in correcting errors for Text-to-SQL. However, their effectiveness diminishes when addressing SQL semantic errors, since LLM hallucinations cause persistent biases in the semantic understanding of questions, leading to an uncorrectable situation. To solve this problem, we propose \textbfTest-driven \textbfSelf-refinement for Text-to-SQL \textbf(TS-SQL). It leverages a collaborative LLM agent framework to automatically synthesize high-quality test cases, including test data and test code. The test cases are further employed to provide execution feedback for LLM self-refinement towards SQL semantic errors. Rigorous evaluation shows the superiority of TS-SQL: for BIRD-dev, TS-SQL improves at least \textbf6\\% over existing SQL self-refinement methods; for Spider-dev, TS-SQL identifies and corrects \textbf131 gold SQL errors, exposing system flaws in benchmark rigor. For reproducibility, we release the modified Spider-dev benchmark to foster further research.

TS-SQL: Test-driven Self-refinement for Text-to-SQL

Fine-tuning LLMs with datasets containing stealthy backdoors from publishers poses security risks to downstream applications. Mainstream detection methods either identify poisoned samples by analyzing the prediction probability of poisoned classification models or rely on the rewriting model to eliminate the stealthy triggers. However, the former cannot be applied to generation tasks, while the latter may degrade generation performance and introduce new triggers. Therefore, efficiently eliminating stealthy poisoned samples for LLMs remains an urgent problem. We observe that after applying TF-IDF clustering to the sample response, there are notable differences in the intra-class distances between clean and poisoned samples. Poisoned samples tend to cluster closely because of their specific malicious outputs, whereas clean samples are more scattered due to their more varied responses. Thus, in this paper, we propose a stealthy backdoor sample detection method based on Reference-Filtration and Tfidf-Clustering mechanisms (RFTC). Specifically, we first compare the sample response with the reference model's outputs and consider the sample suspicious if there's a significant discrepancy. And then we perform TF-IDF clustering on these suspicious samples to identify the true poisoned samples based on the intra-class distance. Experiments on two machine translation datasets and one QA dataset demonstrate that RFTC outperforms baselines in backdoor detection and model performance. Further analysis of different reference models also confirms the effectiveness of our Reference-Filtration.

Detecting Stealthy Backdoor Samples based on Intra-class Distance for Large Language Models

Addressing the intellectual property protection challenges in commercial deployment of large language models (LLMs), existing black-box fingerprinting techniques face dual challenges from incremental fine-tuning erasure and feature-space defense due to their reliance on overfitting high-perplexity trigger patterns. We firstly reveal that, model editing in the fingerprint domain exhibits unique advantages including significantly lower false positive rates, enhanced harmlessness, and superior robustness. Building on this foundation, this paper innovatively proposes a textbfPrefix-textbfenhanced Fingerprint textbfEditing Framework (PREE), which encodes copyright information into parameter offsets through dual-channel knowledge edit to achieve covert embedding of fingerprint features. Experimental results demonstrate that the proposed solution achieves the 90% trigger precision in mainstream architectures including LLaMA-3 and Qwen-2.5. The minimal parameter offset (change rate < 0.03) effectively preserves original knowledge representation while demonstrating strong robustness against incremental fine-tuning and multi-dimensional defense strategies, maintaining zero false positive rate throughout evaluations.

PREE: Towards Harmless and Adaptive Fingerprint Editing in Large Language Models via Knowledge Prefix Enhancement

Large Reason Models (LRMs) extend long reasoning process to solve complex tasks. However, due to the lack of fine-grained control, they often suffer from overthinking and erroneous reasoning problems, risking accuracy loss. To address this issue, we introduce Reasoning Direction Steering (RDS) to enable fine-grained control over LRMs' reasoning behaviors by aligning reasoning trajectories with specific cognitive patterns. We develop a simple yet effective paradigm, Thinking Intervention, which explores two key dimensions - intervention positions and intervention styles - to achieve integration intervention throughout model reasoning processes. To validate the effectiveness of our approach, we conduct comprehensive experiments on multi-hop question answering tasks using state-of-the-art LRMs, including Qwen3-Series and R1-Series models. Experimental results demonstrate the efficacy of Thinking Intervention with 9.4% average improvement on R1-Series models and 1.9% improvement on Qwen3-Series models.

Can We Steer Reasoning Direction by Thinking Intervention?

The automatic control of mobile devices is essential for efficiently performing complex tasks that involve multiple sequential steps. However, these tasks pose significant challenges due to the limited environmental information available at each step, primarily through visual observations. As a result, current approaches, which typically rely on reactive policies, focus solely on immediate observations and often lead to suboptimal decision-making. To address this problem, we propose Foresighted Planning with World Model-Driven Code Execution (FPWC),a framework that prioritizes natural language understanding and structured reasoning to enhance the agent's global understanding of the environment by developing a task-oriented, refinable world model at the outset of the task. Foresighted actions are subsequently generated through iterative planning within this world model, executed in the form of executable code. Extensive experiments conducted in simulated environments and on real mobile devices demonstrate that our method outperforms previous approaches, particularly achieving a 44.4% relative improvement in task success rate compared to the state-of-the-art in the simulated environment.

Unlocking Smarter Device Control: Foresighted Planning with a World Model-Driven Code Execution Approach

While Multimodal Large Language Models (MLLMs) have achieved remarkable progress in open-ended visual question answering, they remain vulnerable to hallucinations. These are outputs that contradict or misrepresent input semantics, posing a critical challenge to the reliability and factual consistency. Existing methods often rely on external verification or post-hoc correction, lacking an internal mechanism to validate outputs directly during training. To bridge this gap, we propose ReLoop, a unified closed-loop training framework that encourages multimodal consistency for cross-modal understanding in MLLMs. ReLoop adopts a ring-shaped structure that integrates three complementary consistency feedback mechanisms, obliging MLLMs to "seeing twice and thinking backwards". Specifically, ReLoop employs the frozen Consistency Feedback Plugin (CFP), comprising semantic reconstruction, visual description, and an attention supervision module for attention alignment. These components collectively enforce semantic reversibility, visual consistency, and interpretable attention, enabling the model to correct its outputs during training. Extensive evaluations and analyses demonstrate the effectiveness of ReLoop in reducing hallucination rates across multiple benchmarks, establishing a robust method for hallucination mitigation in MLLMs. We will release our source code and data in the camera-ready version.

ReLoop: "Seeing Twice and Thinking Backwards" via Closed-loop Training to Mitigate Hallucinations in Multimodal understanding

Large language models (LLMs) have demonstrated remarkable reasoning capabilities, including in financial question answering (FQA). However, the performance in FQA remains limited, particularly in questions that require deep financial knowledge and complex numerical reasoning. While supervised fine-tuning and closed-source LLMs have shown promise, they are often constrained by high costs or computational inefficiency. In this paper, we propose a low-cost yet effective framework, named FinMAN (Financial multi-agent framework), that enables small LLMs (e.g., 8B) to perform complex reasoning tasks without relying on expensive models or task-specific fine-tuning. FinMAN improves formula selection, extraction, and calculation to help small-scale models solve FQA tasks more accurately, with a lightweight verification mechanism to correct common errors. Experimental results show that FinMAN outperforms the best open-source model on BizBench by 10.46% and achieves competitive performance to GPT-3.5 using significantly fewer parameters. Our code and data are publicly available at https://anonymous.4open.science/r/MultiAgentFin-EE42.

David vs. Goliath: Cost-Efficient Financial QA via Cascaded Multi-Agent Reasoning

LLM search methods, such as Chain of Thought (CoT) and Tree of Thought (ToT), enhance LLM reasoning by exploring multiple reasoning paths. When combined with search algorithms like MCTS and Bandit methods, their effectiveness relies heavily on uncertainty estimation to prioritize paths that align with specific search objectives. \emph{However, it remains unclear whether existing LLM uncertainty metrics adequately capture the diverse types of uncertainty required to guide different search objectives.} In this work, we introduce a framework for uncertainty benchmarking, identifying four distinct uncertainty types: Answer, Correctness, Aleatoric, and Epistemic Uncertainty. Each type serves different optimization goals in search. Our experiments demonstrate that current metrics often align with only a subset of these uncertainty types, limiting their effectiveness for objective-aligned search in some cases. These findings highlight the need for additional target-aware uncertainty estimators that can adapt to various optimization goals in LLM search. Code is available at \href{https://anonymous.4open.science/r/Benchmarking-Uncertainty-Metrics-for-LLM-Sampling-Search-4AAC/README.md}{link}.

Downloads

Next from EMNLP 2025

ASTPrompter: Preference-Aligned Automated Language Model Red-Teaming to Generate Low-Perplexity Unsafe Prompts

Similar lecture

LLMs Are Biased Towards Output Formats! Systematically Evaluating and Mitigating Output Format Bias of LLMs