China

This research aims to explore the intersection of Large Language Models and confidence calibration in Entity Matching. To this end, we perform an empirical study to compare baseline RoBERTa confidences for an Entity Matching task against confidences that are calibrated using Temperature Scaling, Monte Carlo Dropout and Ensembles. We use the Abt-Buy, DBLP-ACM, iTunes-Amazon and Company datasets. The findings indicate that the proposed modified RoBERTa model exhibits a slight overconfidence, with Expected Calibration Error scores ranging from 0.0043 to 0.0552 across datasets. We find that this overconfidence can be mitigated using Temperature Scaling, reducing Expected Calibration Error scores by up to 23.83%.

EMNLP 2025

Confidence Calibration in Large Language Model-Based Entity Matching

workshop paper

## 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."<br>

*Message from the General Chair, Dirk Hovy*

[**Link to Conference Handbook**](https://drive.google.com/file/d/1johU5QqVVYO4RfH7QcIORr7qrVBdzdwC/view?usp=sharing)





<br>

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

Large Language Models (LLMs) have demonstrated strong performance in question answering (QA) tasks. However, Multi-Answer Question Answering (MAQA), where a question may have several valid answers, remains challenging. Traditional QA settings often assume consistency across evidences, but MAQA can involve conflicting answers. Constructing datasets that reflect such conflicts is costly and labor-intensive, while existing benchmarks often rely on synthetic data, restrict the task to yes/no questions, or apply unverified automated annotation. To advance research in this area, we extend the conflict-aware MAQA setting to require models not only to identify all valid answers, but also to detect specific conflicting answer pairs, if any. To support this task, we introduce a novel cost-effective methodology for leveraging fact-checking datasets to construct NATCONFQA, a new benchmark for realistic, conflict-aware MAQA, enriched with detailed conflict labels, for all answer pairs. We evaluate eight high-end LLMs on NATCONFQA, revealing their fragility in handling various types of conflicts and the flawed strategies they employ to resolve them.

Consensus or Conflict? Fine-Grained Evaluation of Conflicting Answers in Question-Answering

Recent work has revealed Large Language Models (LLMs) often exhibit undesirable behaviors, such as hallucination and toxicity, limiting their reliability and broader adoption. In this paper, we discover an understudied type of undesirable behavior of LLMs, which we term Verbosity Compensation (VC). VC is similar to the hesitation behavior of humans under uncertainty, compensating with excessive words such as repeating questions, introducing ambiguity, or providing excessive enumeration. We present the first work that analyzes Verbosity Compensation, explores its causes, and proposes a simple mitigating approach. Our experiments on five datasets of knowledge and reasoning-based QA tasks with 14 LLMs, reveal three conclusions. 1) A pervasive presence of VC across all models and all datasets. 2) The large performance gap between verbose and concise responses. We also demonstrate that this difference does not naturally diminish as LLM capability increases. 3) Higher uncertainty exhibited by VC responses across all five datasets, suggesting a strong connection between verbosity and model uncertainty. We propose a simple yet effective cascade algorithm that replaces the verbose responses with the other model-generated responses, alleviating the VC of the Mistral model from 63.81% to 16.16% on the Qasper dataset.

Demystify Verbosity Compensation Behavior of Large Language Models

Quantifying uncertainty in large language models (LLMs) is important for safety-critical applications because it helps spot incorrect answers, known as hallucinations. One major trend of uncertainty quantification methods is based on estimating the entropy of the distribution of the LLM's potential output sequences. This estimation is based on a set of output sequences and associated probabilities obtained by querying the LLM several times. In this paper, we advocate and experimentally and show that the probability of unobserved sequences plays a crucial role, and we recommend future research to integrate it to enhance such LLM uncertainty quantification methods.

On the Role of Unobserved Sequences on Sample-based Uncertainty Quantification for LLMs

We propose a method for confidence estimation in retrieval-augmented generation (RAG) systems that aligns closely with the correctness of large language model (LLM) outputs. Confidence estimation is especially critical in high-stakes domains such as finance and healthcare, where the cost of an incorrect answer outweighs that of not answering the question. Our approach extends prior uncertainty quantification methods by leveraging raw feed-forward network (FFN) activations as auto-regressive signals, avoiding the information loss inherent in token logits and probabilities after projection and softmax normalization. We model confidence prediction as a sequence classification task, and regularize training with a Huber loss term to improve robustness against noisy supervision. Applied in a real-world financial industry customer-support setting with complex knowledge bases, our method outperforms strong baselines and maintains high accuracy under strict latency constraints. Experiments on Llama 3.1 8B model show that using activations from only the 16th layer preserves accuracy while reducing response latency. Our results demonstrate that activation-based confidence modeling offers a scalable, architecture-aware path toward trustworthy RAG deployment.

Confidence-Based Response Abstinence: Improving LLM Trustworthiness via Activation-Based Uncertainty Estimation

Fine-tuning large language models (LLMs) with low-rank adaptaion (LoRA) is a cost-effective way to incorporate information from a specific dataset. However, it is often unclear how well the fine-tuned LLM will generalize, i.e., how well it will perform on unseen datasets. Methods have been proposed to improve generalization by optimizing with in-context prompts, or by using meta-learning to fine-tune LLMs. However, these methods are expensive in memory and computation, requiring either long-context prompts or saving copies of parameters and using second-order gradient updates. To address these challenges, we propose Amortized Bayesian Meta-Learning for LoRA (ABMLL). This method builds on amortized Bayesian meta-learning for smaller models, adapting this approach to LLMs while maintaining its computational efficiency. We reframe task-specific and global parameters in the context of LoRA and use a set of new hyperparameters to balance reconstruction accuracy and the fidelity of task-specific parameters to the global ones. ABMLL provides effective generalization and scales to large models such as \textsc{Llama3-8B}. Furthermore, as a result of using a Bayesian framework, ABMLL provides improved uncertainty quantification. We test ABMLL on Unified-QA and Crossfit datasets and find that it outperforms existing methods on these benchmarks in terms of both accuracy and expected calibration error.

Amortized Bayesian Meta-Learning for Low-Rank Adaptation of Large Language Models

While large, biomedical documents with complex terminology are in need of being understood more easily and efficiently, summarizing this kind of content can be problematic, as Large Language Models (LLMs) aren't always trustworthy. Considering the importance of comprehending Cardiovascular Diseases, we study in depth the ability of different state-of-the-art biomedical LLMs to generate factual and certain summaries in this topic, and examine which generation choices can influence their trustworthiness. To that end, besides using factuality metrics, we employ techniques for token-level uncertainty estimation, an area that has received little attention from the scientific community. Our results reveal dissimilarities between LLMs and generation methods, and highlight connections between factuality and uncertainty metrics, thereby laying the groundwork for further investigation in the area.

Towards Trustworthy Summarization of Cardiovascular Articles: A Factuality-and-Uncertainty-Aware Biomedical LLM Approach

Recent papers show LLMs achieve near-random accuracy in causal relation classification, raising questions about whether such failures arise from limited pretraining exposure or deeper representational gaps. We investigate this under uncertainty-based evaluation, testing whether pretraining exposure to causal examples improves causal understanding using >18K PubMed sentences—half from The Pile corpus, half post-2024—across seven models (Pythia-1.4B/7B/12B, GPT-J-6B, Dolly-7B/12B, Qwen-7B). We analyze model behavior through: (i) causal classification, where the model identifies causal relationships in text, and (ii) verbatim memorization probing, where we assess whether the model prefers previously seen causal statements over their paraphrases. Models perform four-way classification (direct/conditional/correlational/no-relationship) and select between originals and their generated paraphrases. Results show 
almost identical accuracy on seen/unseen sentences (p>0.05), no memorization bias (24.8\% original selection), output distribution over the possible options almost flat --- with entropic values near the maximum (1.35/1.39), confirming random guessing. Instruction-tuned models show severe miscalibration (Qwen: >95\% confidence, 32.8\% accuracy, ECE=0.49). Conditional relations induce highest entropy (+11\% vs direct). These findings suggest that failures in causal understanding arise from the lack of structured causal representation, rather than insufficient exposure to causal examples during pretraining.

Causal Understanding by LLMs: The Role of Uncertainty

Ambiguous words or underspecified references require interlocutors to resolve them, often by relying on shared context and commonsense knowledge. Therefore, we systematically investigate whether Large Language Models (LLMs) can leverage commonsense to resolve referential ambiguity in multi-turn conversations and analyze their behavior when ambiguity persists. Further, we study how requests for simplified language affect this capacity. Using a novel multilingual evaluation dataset, we test DeepSeek v3, GPT-4o, Qwen3-32B, GPT-4o-mini, and Llama-3.1-8B via LLM-as-Judge and human annotations. Our findings indicate that current LLMs struggle to resolve ambiguity effectively: they tend to commit to a single interpretation or cover all possible references, rather than hedging or seeking clarification. This limitation becomes more pronounced under simplification prompts, which drastically reduce the use of commonsense reasoning and diverse response strategies. Fine-tuning Llama-3.1-8B with Direct Preference Optimization substantially improves ambiguity resolution across all request types. These results underscore the need for advanced fine-tuning to improve LLMs' handling of ambiguity and to ensure robust performance across diverse communication styles.

It Depends: Resolving Referential Ambiguity in Minimal Contexts with Commonsense Knowledge

We study the source of uncertainty in DeepSeek R1-32B by analyzing its self-reported verbal confidence on question answering (QA) tasks. In the default answer-then-confidence setting, the model is regularly over-confident, whereas semantic entropy - obtained by sampling many responses - remains reliable. We hypothesize that this is because of semantic entropy's larger test-time compute, which lets us explore the model's predictive distribution. We show that granting DeepSeek the budget to explore its distribution by forcing a long chain-of-thought before the final answer greatly improves its verbal score effectiveness, even on simple fact-retrieval questions that normally require no reasoning. Our analysis concludes that reliable uncertainty estimation requires explicit exploration of the generative space, and self-reported confidence is trustworthy only after such exploration.

Read Your Own Mind: Reasoning Helps Surface Self-Confidence Signals in LLMs

In recent years, we have seen an increased usage of neural ranking models in the information retrieval domain. Although language model-based rankers have shown significant progress in performing ranking tasks, little to no work has addressed the issue of fine-tuning them in the presence of label noise in the training data. In a general learning setting, training models in the presence of noisy labeled data is studied extensively. To this end, confidence calibration approaches have shown significant promise; however, their usage in training neural ranking models is relatively less studied. In this work, we address this gap by adapting and analyzing regularization-based calibration approaches to reduce the effect of label noise in ranking tasks. Specifically, we study label relaxation in neural ranking models. We demonstrate the effectiveness of this approach by performing extensive evaluations comparing the label relaxation approach to standard loss functions. Additionally, we analyze the calibration error associated with the loss functions.
After evaluating on five different noise levels, two different ranking models, and four diverse ranking datasets, the results suggest that label relaxation can improve the performance of the ranking models under noisy labels. Furthermore, we find that label relaxation reduces calibration error, although it suggests a better metric to be used for neural ranking models.

Downloads

Next from EMNLP 2025

Consensus or Conflict? Fine-Grained Evaluation of Conflicting Answers in Question-Answering

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES

.css-70qvj9{display:-webkit-box;display:-webkit-flex;display:-ms-flexbox;display:flex;-webkit-align-items:center;-webkit-box-align:center;-ms-flex-align:center;align-items:center;}Downloads

Next from EMNLP 2025

Consensus or Conflict? Fine-Grained Evaluation of Conflicting Answers in Question-Answering

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES

Downloads