Papers by Andreas Vlachos

Improving Scheduled Sampling with Elastic Weight Consolidation for Neural Machine Translation (2022.findings-emnlp)

Challenge:	Conformity is a form of social influence that affects the way people respond to information.
Approach:	They adapt psychological experiments to examine the extent of conformity in large language models.
Outcome:	The proposed interventions mitigate conformity by reducing the naturalness of majority tones and reducing instruction-tuned models.

Copied to clipboard

Michalis Korakakis, Andreas Vlachos

Machine Translation

Challenge:	Autoregressive models trained with maximum likelihood estimation suffer from exposure bias, i.e. the discrepancy between ground-truth prefixes used during training and model-generated prefix at inference time.
Approach:	They propose to use Elastic Weight Consolidation to better balance mitigating exposure bias with retaining performance.
Outcome:	The proposed method significantly outperforms maximum likelihood estimation and scheduled sampling baselines on four translation datasets.

ProoFVer: Natural Logic Theorem Proving for Fact Verification (2022.tacl-1)

Copied to clipboard

Amrith Krishna, Sebastian Riedel, Andreas Vlachos

Dis2Dis: Explaining Ambiguity in Fact-Checking (2025.findings-naacl)

Challenge:	Recent fact verification systems rely on neural network classifiers for veracity prediction, which lack explainability.
Approach:	They propose a model that generates natural logic-based inferences as proofs using lexical mutations between spans in the claim and the evidence retrieved.
Outcome:	The proposed model has highest label accuracy and second best score in the FEVER leaderboard.

Copied to clipboard

Ieva Staliunaite, Andreas Vlachos

Merge and Label: A Novel Neural Network Architecture for Nested NER (P19-1)

Challenge:	Ambiguity is a linguistic tool for encoding information efficiently, yet it also causes misunderstandings and disagreements.
Approach:	They propose a constrained generation task for explaining ambiguous claims in fact-checking by editing them to spell out an interpretation that can be unequivocally supported by the given evidence.
Outcome:	The proposed model disambiguates claims 72% of the time compared to a simple copy baseline and a Large Language Model baseline.

Copied to clipboard

Joseph Fisher, Andreas Vlachos

Strong Baselines for Complex Word Identification across Multiple Languages (N19-1)

Challenge:	Named entity recognition (NER) is one of the best studied tasks in natural language processing.
Approach:	They propose a neural network architecture that merges tokens and/or entities into nested entities and labels them independently.
Outcome:	The proposed approach achieves state-of-the-art F1 of 74.6 and improves with contextual embeddings to 82.4.

Copied to clipboard

Pierre Finnimore, Elisabeth Fritzsch, Daniel King, Alison Sneyd, Aneeq Ur Rehman, Fernando Alva-Manchego, Andreas Vlachos

Multilinguality

Challenge:	Complex Word Identification (CWI) is the task of identifying which words or phrases in a sentence are difficult to understand by a specific type of reader.
Approach:	They propose to use monolingual and cross-lingual CWI models to make predictions for languages not seen during training.
Outcome:	The proposed models perform as well as (or better than) most models submitted to the latest CWI Shared Task.

Survival text regression for time-to-event prediction in conversations (2021.findings-acl)

Copied to clipboard

Do We Need Language-Specific Fact-Checking Models? The Case of Chinese (2024.emnlp-main)

Challenge:	a recent study has modelled time-to-event prediction tasks as classification tasks . authors: this is contrived and less informative than traditional classification models .
Approach:	They propose to frame time-to-event prediction tasks as classification tasks . they use survival regression techniques commonly used in healthcare and reliability engineering .
Outcome:	The proposed models outperform text regression methods and comparable classification models on three datasets.

Copied to clipboard

Caiqi Zhang, Zhijiang Guo, Andreas Vlachos

Demystifying Multi-Agent Debate: The Role of Confidence and Diversity (2026.findings-acl)

Challenge:	Existing fact-checking models in other languages lack grounding in real-world claims . current models are constrained to a single domain, like COVID-19 .
Approach:	They propose a Chinese document-level evidence retriever that can be translated into Chinese . they then construct an adversarial dataset that is more robust toward biases .
Outcome:	The proposed method outperforms translation-based methods and multilingual language models and is more robust toward biases.

Copied to clipboard

Xiaochen Zhu, Caiqi Zhang, Yizhou Chi, Tom Stafford, Nigel Collier, Andreas Vlachos

Challenge:	Multi-agent debate (MAD) is widely used to improve large language models' (LLMs) reasoning and test-time scaling.
Approach:	They propose a diversity-aware initialisation that selects a more diverse pool of candidate answers, increasing the likelihood that a correct hypothesis is present at the start of debate.
Outcome:	The proposed protocol outperforms vanilla MAD and majority vote on six reasoning-oriented QA benchmarks.

Document-level Claim Extraction and Decontextualisation for Fact-Checking (2024.acl-long)

Copied to clipboard

Zhenyun Deng, Michael Schlichtkrull, Andreas Vlachos

Uncertainty Quantification for Evaluating Gender Bias in Machine Translation (2026.findings-eacl)

Challenge:	Existing methods for document-level claim extraction focus on identifying and extracting claims from individual sentences.
Approach:	They propose a method for document-level claim extraction for fact-checking which aims to extract check-worthy claims from documents and decontextualise them so they can be understood out of context.
Outcome:	The proposed method extracts check-worthy claims from documents and decontextualises them so they can be understood out of context.

Copied to clipboard

Ieva Staliunaite, Julius Cheng, Andreas Vlachos

Multilinguality

Challenge:	Existing models can reproduce existing social inequalities but cannot be reduced.
Approach:	They propose that models should maintain uncertainty when input is ambiguous to avoid reinforcing biases.
Outcome:	The proposed model can detect gender bias when translated to ambiguous and unambiguous sources and shows that it does not correlate with high translation accuracy and debiases the two cases differently.

Natural Logic-guided Autoregressive Multi-hop Document Retrieval for Fact Verification (2022.emnlp-main)

Copied to clipboard

Rami Aly, Andreas Vlachos

An LLM Feature-based Framework for Dialogue Constructiveness Assessment (2024.emnlp-main)

Challenge:	Recent evidence retrieval approaches rely on heuristics and assume hyperlinks between documents.
Approach:	They propose a retrieval method that combines a retriever and a proof system that reranks documents and reorders them .
Outcome:	The proposed method exceeds or is on par with the current state-of-the-art on FEVER, HoVer and FEVEROUS-S while using 5 to 10 times less memory than competing systems.

Copied to clipboard

Lexin Zhou, Youmna Farag, Andreas Vlachos

Zero-Shot Fact Verification via Natural Logic and Large Language Models (2024.findings-emnlp)

Challenge:	Existing studies on dialogue constructiveness assessment focus on analysing conversational factors that influence individuals to take specific actions, win debates, change their perspectives or broaden their open-mindedness.
Approach:	They propose an LLM feature-based framework for dialogue constructiveness assessment that combines the strengths of feature- and neural approaches while mitigating their downsides.
Outcome:	The proposed framework outperforms standard feature-based models and neural models on three dialogue constructiveness datasets.

Copied to clipboard

Marek Strong, Rami Aly, Andreas Vlachos

Multimodal Automated Fact-Checking: A Survey (2023.findings-emnlp)

Challenge:	Recent advancements in fact verification systems with natural logic have enhanced their explainability by aligning claims with evidence through set-theoretic operators, providing faithful justifications.
Approach:	They propose a method that utilizes the generalization capabilities of instruction-tuned large language models to provide faithful justifications.
Outcome:	The proposed method outperforms other systems that were not specifically trained on natural logic data, and achieves an average accuracy improvement of 8.96 points over the baseline.

Copied to clipboard

Mubashara Akhtar, Michael Schlichtkrull, Zhijiang Guo, Oana Cocarascu, Elena Simperl, Andreas Vlachos

Multimodality

Challenge:	Existing studies on automated fact-checking focus on text, but they focus on a single modality, text . multimodal misinformation is perceived as more credible by humans and spreads faster than text-only counterparts.
Approach:	They propose a framework for automated fact-checking that includes subtasks unique to multimodal misinformation.
Outcome:	The proposed framework includes subtasks unique to multimodal misinformation.

Cross-Policy Compliance Detection via Question Answering (2021.emnlp-main)

Copied to clipboard

Marzieh Saeidi, Majid Yazdani, Andreas Vlachos

Question Answering

Challenge:	Policy compliance detection is the task of ensuring that a scenario conforms to a policy.
Approach:	They propose to decompose policy compliance detection into question answering . they propose to use an existing dataset to augment expert annotations .
Outcome:	The proposed approach improves accuracy in cross-policy setups, especially when policies are unseen in training.

ALVIN: Active Learning Via INterpolation (2024.emnlp-main)

Copied to clipboard

Michalis Korakakis, Andreas Vlachos, Adrian Weller

Automated Fact Checking: Task Formulations, Methods and Future Directions (C18-1)

Challenge:	Experimental results show that Active Learning methods ignore example groups whose prevalence may vary . supervised fine-tuning remains a critical component of model development, authors say .
Approach:	They propose an approach that uses interpolations to create anchors between examples . they propose to use the model to identify informative examples that counteract shortcuts .
Outcome:	The proposed model outperforms state-of-the-art active learning methods on six datasets . it prioritizes high-certainty instances that integrate representations from different example groups .

Copied to clipboard

James Thorne, Andreas Vlachos

QA-NatVer: Question Answering for Natural Logic-based Fact Verification (2023.emnlp-main)

Challenge:	Recent research on fact checking has focused on misinformation . however, relevant papers and articles have been published in research communities that are unaware of each other and use inconsistent terminology.
Approach:	They propose avenues for future NLP research on automated fact checking . they highlight the use of evidence as an important distinguishing factor .
Outcome:	The proposed methods unify the task formulations and methodologies across papers and authors.

Copied to clipboard

Rami Aly, Marek Strong, Andreas Vlachos

AnchorAL: Computationally Efficient Active Learning for Large and Imbalanced Datasets (2024.naacl-long)

Challenge:	Recent work has focused on natural logic, which operates directly on natural language by capturing the semantic relation of spans between an aligned claim and its evidence via set-theoretic operators.
Approach:	They propose to use question answering to predict natural logic operators using generalization capabilities of instruction-tuned language models.
Outcome:	The proposed approach outperforms the best baseline on a Danish verification dataset by 4.3 accuracy points.

Copied to clipboard

Pietro Lesci, Andreas Vlachos

TCP: a Benchmark for Temporal Constraint-Based Planning (2025.emnlp-main)

Challenge:	Standard pool-based active learning is computationally expensive on large pools and often reaches low accuracy by overfitting the initial decision boundary.
Approach:	They propose a pool-based active learning method that selects class-specific instances from a labelled set and retrieves the most similar unlabelled instances from the pool.
Outcome:	Experiments with AnchorAL show that it is faster, often reducing runtime from hours to minutes, and trains more performant models.

Copied to clipboard

Zifeng Ding, Sikuan Yan, Moy Yuan, Xianglong Hu, Fangru Lin, Andreas Vlachos

Incorporating Label Dependencies in Multilabel Stance Detection (D19-1)

Challenge:	Existing benchmarks evaluate temporal reasoning and planning in isolation and under limited forms of complexity.
Approach:	They propose a temporal constraint-based planning benchmark that assesses temporal reasoning and planning capabilities in large language models.
Outcome:	The proposed model fails to perform well under limited constraints and lacks temporal grounding.

Copied to clipboard

William Ferreira, Andreas Vlachos

Sentiment Analysis

Challenge:	Stance detection is a well-studied task in social media, but previous work focused on multiclass versions of the problem where labels are mutually exclusive.
Approach:	They propose a method that explicitly incorporates label dependencies in the training objective and reduces multilabel to multiclass learning.
Outcome:	The proposed method improves on two out of three datasets and reduces multilabel to multiclass learning.

A Survey on Automated Fact-Checking (2022.tacl-1)

Copied to clipboard

Zhijiang Guo, Michael Schlichtkrull, Andreas Vlachos

Improving Zero-shot Sentence Decontextualisation with Content Selection and Planning (2025.emnlp-main)

Challenge:	Fact-checking is an essential task in journalism due to the speed with which information and misinformation can spread in the media ecosystem.
Approach:	They propose to use natural language processing to automate fact-checking by identifying common concepts and defining definitions.
Outcome:	The proposed method can predict the veracity of claims using natural language processing, machine learning, and databases.

Copied to clipboard

Zhenyun Deng, Yulong Chen, Andreas Vlachos

Leveraging Wikipedia article evolution for promotional tone detection (2022.acl-long)

Challenge:	Existing methods for extracting sentences from documents leave some implicit discourse information in the sentence unresolved due to their lack of context.
Approach:	They propose a content selection framework for zero-shot decontextualisation which determines what content should be mentioned and in what order for a sentence to be understood out of context.
Outcome:	The proposed framework outperforms existing methods in rewriting sentences that lack context while maintaining original meaning.

Copied to clipboard

TSVer: A Benchmark for Fact Verification Against Time-Series Evidence (2025.emnlp-main)

Challenge:	Detecting biased language is useful for a variety of applications, authors say . a dataset for document-level promotional tone detection is available for WikiEvolve .
Approach:	They propose a dataset for document-level promotional tone detection using Wikipedia . they use a gradient reversal framework to encode two versions simultaneously .
Outcome:	The proposed dataset improves on in-domain and out-of-domain evaluations.

Copied to clipboard

Marek Strong, Andreas Vlachos

Reasoning

Challenge:	Existing systems for fact-checking lack structured evidence, provide insufficient justifications for verdicts, or rely on synthetic claims.
Approach:	They propose a temporal and numerical reasoning dataset based on time-series evidence that is annotated with time frames and a verdict and justifications reflecting how the evidence is used to reach the verdict.
Outcome:	The proposed dataset improves the quality of the annotations and achieves an inter-annotator agreement of = 0.745 on verdicts.

Mitigating Shortcut Learning with InterpoLated Learning (2025.acl-long)

Copied to clipboard

Michalis Korakakis, Andreas Vlachos, Adrian Weller

Generating Fact Checking Briefs (2020.emnlp-main)

Challenge:	Existing shortcut mitigation approaches are model-specific, difficult to tune, computationally expensive, and fail to improve learned representations.
Approach:	They propose to interpolate representations of majority examples to include features from intra-class minority examples with shortcut-mitigating patterns.
Outcome:	The proposed method improves minority generalization over ERM and state-of-the-art mitigation methods on multiple natural language understanding tasks while preserving accuracy on majority examples.

Copied to clipboard

Angela Fan, Aleksandra Piktus, Fabio Petroni, Guillaume Wenzek, Marzieh Saeidi, Andreas Vlachos, Antoine Bordes, Sebastian Riedel

FEVER: a Large-scale Dataset for Fact Extraction and VERification (N18-1)

Challenge:	Existing work has framed fact checking as classification, often supported by a claim as input.
Approach:	They propose to use natural language briefs to increase the accuracy of fact checking . they show that QABriefer increases the accuracy by 10% while QABries reduce time .
Outcome:	The proposed model increases the accuracy of crowdworkers by 10% while reducing the time required by 20%.

Copied to clipboard

James Thorne, Andreas Vlachos, Christos Christodoulopoulos, Arpit Mittal

Topic or Style? Exploring the Most Useful Features for Authorship Attribution (C18-1)

Challenge:	185,445 claims generated by altering sentences from Wikipedia are verified without knowledge of the sentence they were derived from.
Approach:	They propose a publicly available dataset for verification against textual sources, FEVER: Fact Extraction and VERification.
Outcome:	The proposed dataset achieves 31.87% accuracy on labeling a claim accompanied by the correct evidence, compared to 50.91% if we ignore the evidence.

Copied to clipboard

Yunita Sari, Mark Stevenson, Andreas Vlachos

Representation Learning

Challenge:	Existing approaches to authorship attribution rely on individual's writing style and/or preferred topics.
Approach:	They analyse four widely used datasets to explore how different types of features affect authorship attribution accuracy under varying conditions.
Outcome:	The proposed model outperforms the state-of-the-art on two out of the four datasets used.

Generating Token-Level Explanations for Natural Language Inference (N19-1)

Copied to clipboard

James Thorne, Andreas Vlachos, Christos Christodoulopoulos, Arpit Mittal

Ev2R: Evaluating Evidence Retrieval in Automated Fact-Checking (2026.tacl-1)

Challenge:	Existing methods to generate token-level explanations for NLI on single sentences have not been tested.
Approach:	They propose to generate token-level explanations for NLI without explicitly annotating training data.
Outcome:	The proposed approach is faster and more accurate than the black-box methods.

Copied to clipboard

Mubashara Akhtar, Michael Schlichtkrull, Andreas Vlachos

Neural Generative Rhetorical Structure Parsing (D19-1)

Challenge:	Current methods for automated fact-checking rely on relying on other evaluation metrics and closed knowledge sources.
Approach:	They propose a method which combines evidence evaluation with verdict-level proxy scoring.
Outcome:	The proposed method outperforms existing methods in accuracy and robustness against human ratings and adversarial tests.

Copied to clipboard

Amandla Mabona, Laura Rimell, Stephen Clark, Andreas Vlachos

I Beg to Differ: A study of constructive disagreement in online conversations (2021.eacl-main)

Challenge:	Existing approaches to RST parsing use discriminative models, but these are less sample efficient and datasets are typically small.
Approach:	They propose a document-level RNN grammar with a bottom-up traversal order and a beam search algorithm that keeps track of both structure-and word-generating actions without exhibit-ing this branching bias.
Outcome:	The proposed model outperforms a discriminative model with the same features by 2.6 F1points and achieves comparable performance to the state-of-the-art.

Copied to clipboard

Segment-Level Diffusion: A Framework for Controllable Long-Form Generation with Diffusion Language Models (2025.acl-long)

Challenge:	Disagreements are pervasive in human communication.
Approach:	They construct a corpus of Wikipedia Talk page conversations that contain content disputes and define the task of predicting whether disagreements will be escalated to mediation by a moderator.
Outcome:	The proposed model outperforms feature-based models in predicting whether disagreements will escalate to mediation by a moderator.

Copied to clipboard

Xiaochen Zhu, Georgi Karadzhov, Chenxi Whitehouse, Andreas Vlachos

Robustness in NLP

Challenge:	Diffusion models have shown promise in text generation, but often struggle with generating long, coherent, and contextually accurate text.
Approach:	They propose a framework that enhances diffusion-based text generation through text segmentation, robust representation training with adversarial and contrastive learning, and improved latent-space guidance.
Outcome:	The proposed framework improves diffusion-based text generation and improves scalability and fluency.

Incremental Beam Manipulation for Natural Language Generation (2021.eacl-main)

Copied to clipboard

James Hargreaves, Andreas Vlachos, Guy Emerson

Measuring Uncertainty in Neural Machine Translation with Similarity-Sensitive Entropy (2024.eacl-long)

Challenge:	a larger beam size can lead to deteriorating performance of natural language generation systems due to model errors . performance of NLG systems can plateau or even decrease when beam sizes larger than 10 are used .
Approach:	They propose to rerank the output of beam search to produce a good set of hypotheses . they propose incremental beam manipulation to discarded hypothese .
Outcome:	The proposed method outperforms a strong reranker on the E2E and WebNLG datasets while being on par with the existing method.

Copied to clipboard

Julius Cheng, Andreas Vlachos

Multilinguality

Challenge:	Uncertainty estimation is an important diagnostic tool for statistical models.
Approach:	They propose to adapt similarity-sensitive Shannon entropy (S3E) for NMT by incorporating a concept borrowed from theoretical ecology.
Outcome:	The proposed framework improves quality estimation and named entity recall, and improves translation quality.

PledgeTracker: A System for Monitoring the Fulfilment of Pledges (2025.emnlp-demos)

Copied to clipboard

Yulong Chen, Michael Sejr Schlichtkrull, Zhenyun Deng, David Corney, Nasim Asl, Joshua Salisbury, Andrew Dudfield, Andreas Vlachos

Automated Fact-Checking in Dialogue: Are Specialized Models Needed? (2023.emnlp-main)

Challenge:	Existing methods simplify pledge verification into document classification task, overlooking its dynamic temporal and multi-document nature.
Approach:	They propose a system that reformulates pledge verification into structured event timeline construction.
Outcome:	The proposed system shows that it can be used in real-world workflows and reduces human verification effort.

Copied to clipboard

Eric Chamoun, Marzieh Saeidi, Andreas Vlachos

Ethical NLP

Challenge:	Prior work has shown that typical fact-checking models struggle with claims made in conversation.
Approach:	They propose to fine-tune models for dialogue on conversational data to improve performance on typical fact-checking.
Outcome:	The proposed models perform better on stand-alone claims than state-of-the-art models for dialogue while maintaining their performance on standalone claim.

A Bayesian Optimization Approach to Machine Translation Reranking (2025.naacl-long)

Copied to clipboard

Julius Cheng, Maike Züfle, Vilém Zouhar, Andreas Vlachos

Machine Translation

Challenge:	reranking is a method of improving prediction quality but can add computational cost.
Approach:	They propose to score a list of prediction candidates with an external scoring model and return the highest-scoring candidate.
Outcome:	The proposed method achieves the same CometKiwi score using 70 evaluations on average compared to scoring a subset of 180 candidates.

LoVeC: Reinforcement Learning for Better Verbalized Confidence in Long-Form Generation (2026.acl-long)

Copied to clipboard

Caiqi Zhang, Xiaochen Zhu, Chengzu Li, Nigel Collier, Andreas Vlachos

Faster Minimum Bayes Risk Decoding with Confidence-based Pruning (2023.emnlp-main)

Challenge:	Existing methods for hallucination detection are limited to short-form question answering tasks and do not generalize well to open-ended generation.
Approach:	They propose a method that trains LLMs to append a numerical confidence score to each generated statement during long-form generation.
Outcome:	The proposed method is 20 faster than traditional self-consistency methods while achieving better calibration.

Copied to clipboard

Julius Cheng, Andreas Vlachos

Model-Agnostic Meta-Learning for Relation Classification with Limited Supervision (P19-1)

Challenge:	Minimum Bayes risk (MBR) decoding is a decision rule for conditional sequence generation tasks.
Approach:	They propose an algorithm which grows the number of samples used to estimate utility . it prunes hypotheses that are unlikely to have the highest utility based on bootstrap sampling .
Outcome:	The proposed method outperforms beam search in conditional language generation and neural machine translation tasks while being statistically indistinguishable from other proposed methods.

Copied to clipboard

Abiola Obamuyide, Andreas Vlachos

Leveraging Type Descriptions for Zero-shot Named Entity Recognition and Classification (2021.acl-long)

Challenge:	Existing methods for relation classification are based on labelled training data, which can be expensive to obtain and available only in limited quantities.
Approach:	They propose a model-agnostic meta-learning protocol for training relation classifiers to achieve enhanced predictive performance in limited supervision settings.
Outcome:	The proposed protocol improves the performance of two state-of-the-art supervised relation classification models on two relation classification datasets.

Copied to clipboard

Rami Aly, Andreas Vlachos, Ryan McDonald

Guided Neural Language Generation for Abstractive Summarization using Abstract Meaning Representation (D18-1)

Challenge:	Named entity recognition and classification (NERC) tasks require annotated data for the target classes during training.
Approach:	They propose a novel approach that leverages the fact that textual descriptions for many entity classes occur naturally.
Outcome:	The proposed approach outperforms baselines adapted from machine reading comprehension and zero-shot text classification.

Copied to clipboard

Hardy Hardy, Andreas Vlachos

Varifocal Question Generation for Fact-checking (2022.emnlp-main)

Challenge:	Recent work on abstractive summarization has made progress with neural encoder-decoder architectures, but these models lack explicit semantic modeling of the source document and its summary.
Approach:	They extend previous work on abstractive summarization using Abstract Meaning Representation (AMR) with a neural language generation stage which they guide using the source document.
Outcome:	The proposed approach improves summarization performance by 7.4 and 10.5 points in ROUGE-2 using gold standard AMR parses and parses obtained from an off-the-shelf parser respectively.

Copied to clipboard

Nedjma Ousidhoum, Zhangdie Yuan, Andreas Vlachos

Evaluating adversarial attacks against multiple fact verification systems (D19-1)

Challenge:	Recent question generation approaches assume that the answer is known . however, such passages are what is being sought when verifying a claim.
Approach:	They propose a method that generates questions based on different focal points within a claim . they demonstrate that the method generates more relevant and informative questions .
Outcome:	The proposed method outperforms previous work on a fact-checking question generation dataset on measurable evaluation metrics.

Copied to clipboard

James Thorne, Andreas Vlachos, Christos Christodoulopoulos, Arpit Mittal

The Intended Uses of Automated Fact-Checking Artefacts: Why, How and Who (2023.findings-emnlp)

Challenge:	Automated fact verification is progressing due to advances in modeling and availability of large datasets.
Approach:	They propose two scoring metrics which take into account the correctness of adversarial instances.
Outcome:	The proposed method and paraphrasing method have higher potency and higher resilience than baselines.

Copied to clipboard

Michael Schlichtkrull, Nedjma Ousidhoum, Andreas Vlachos

Causal Estimation of Tokenisation Bias (2025.acl-long)

Challenge:	Automated fact-checking is often presented as an epistemic tool fact-seekers, social media consumers, and other stakeholders can use to fight misinformation.
Approach:	They analyse 100 highly-cited papers and annotate epistemic elements related to intended use, i.e., means, ends, and stakeholders.
Outcome:	The proposed strategies are often left out of the literature and lack empirical backing.

Copied to clipboard

Pietro Lesci, Clara Meister, Thomas Hofmann, Andreas Vlachos, Tiago Pimentel

Syntactic Text Processing

Challenge:	Modern language models define probabilities over character-strings, but in practice, it does . Ideally, the choice of the tokeniser should not affect the probability assigned to the underlying character- string.
Approach:	They quantify a type of tokenisation bias by framing it as a causal effect and estimating it using the regression discontinuity design.
Outcome:	The proposed model can estimate tokenisation bias by comparing subwords around arbitrary cutoff points.

How to disagree well: Investigating the dispute tactics used on Wikipedia (2022.emnlp-main)

Copied to clipboard

Improving the robustness of NLI models with minimax training (2023.acl-long)

Challenge:	Disagreements are often studied from the perspective of toxicity or analysing argument structure.
Approach:	They propose a dispute tactics framework which unifies both perspectives . they annotate 213 disagreements from Wikipedia Talk pages .
Outcome:	The proposed framework can be used to predict disagreements with a transformer-based model.

Copied to clipboard

Michalis Korakakis, Andreas Vlachos

Robustness in NLP

Challenge:	Experimental results show that our method consistently outperforms other robustness enhancement techniques on out-of-distribution adversarial test sets, while maintaining high in-distance accuracy.
Approach:	They propose a minimax objective between a learner model being trained for the task and an auxiliary model aiming to maximize the learner's loss by up-weighting underrepresented "hard" examples with patterns that contradict the shortcuts learned from the prevailing "easy" examples.
Outcome:	The proposed method outperforms other robustness enhancement techniques on out-of-distribution adversarial test sets while maintaining high in-distance accuracy.

Social Good or Scientific Curiosity? Uncovering the Research Framing Behind NLP Artefacts (2025.emnlp-main)

Copied to clipboard

Eric Chamoun, Nedjma Ousidhoum, Michael Sejr Schlichtkrull, Andreas Vlachos

Elastic weight consolidation for better bias inoculation (2021.eacl-main)

Challenge:	Recent studies show that few papers explicitly identify key stakeholders, intended uses, or appropriate contexts.
Approach:	They propose to automate analysis of NLP research by extracting key elements and linking them through interpretable rules and contextual reasoning.
Outcome:	The proposed system improves on two domains of fact-checking and hate speech detection.

Copied to clipboard

James Thorne, Andreas Vlachos

Evidence-based Factual Error Correction (2021.acl-long)

Challenge:	Recent studies have shown that the lack of suitable inductive biases in sentence-pair classification models can cause misclassifications on training datasets.
Approach:	They propose to use elastic weight consolidation (EWC) to fine-tune models to mitigate biases while being less susceptible to catastrophic forgetting.
Outcome:	The proposed model improves on fact verification and stress tests while maintaining the original task accuracy.

Copied to clipboard

James Thorne, Andreas Vlachos

Reasoning

Challenge:	Existing methods to correct factual errors are limited to labeled claims . a recent task of fact verification has attracted significant attention .
Approach:	They propose a task of factual error correction that performs edits to a claim so that the generated rewrite is better supported by evidence.
Outcome:	The proposed method produces accurate factual error corrections for 5x more instances in human evaluation and a .125 increase in SARI score.

The FEVER2.0 Shared Task (D19-66)

Copied to clipboard

James Thorne, Andreas Vlachos, Oana Cocarascu, Christos Christodoulopoulos, Arpit Mittal

Opening up Minds with Argumentative Dialogues (2022.findings-emnlp)

Challenge:	Existing deep neural models are becoming more complex and difficult to understand and characterize their behaviour.
Approach:	They present the results of the second Fact Extraction and VERification (FEVER2.0) Shared Task.
Outcome:	The proposed task was based on the second Fact Extraction and VERification (FEVER2.0) shared task.

Copied to clipboard

Youmna Farag, Charlotte Brand, Jacopo Amidei, Paul Piwek, Tom Stafford, Svetlana Stoyanchev, Andreas Vlachos