23.10.2026
MCML at EMNLP 2026
33 Accepted Papers (21 Main, 11 Findings, and 1 Workshop)
Conference on Empirical Methods in Natural Language Processing, Budapest, Hungary, Oct 24-29, 2026
We are happy to announce that MCML researchers have contributed a total of 33 papers to EMNLP 2026: 21 Main, 11 Findings, and 1 Workshop papers. Congrats to our researchers!
Main Track (21 papers)
J. Bi • Aniri • Y. Wang • D. Yan • W. Huang • Z. Jin • X. Ma • S. Yan • A. Hecker • M. Ye • X. Xiao • H. Schütze • V. Tresp • Y. Ma
PRISM: Self-Pruning Intrinsic Selection Method for Training-Free Multimodal Data Selection.
EMNLP 2026 - Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. Oral Presentation. To be published. Preprint available. arXiv GitHub
PRISM: Self-Pruning Intrinsic Selection Method for Training-Free Multimodal Data Selection.
EMNLP 2026 - Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. Oral Presentation. To be published. Preprint available. arXiv GitHub
B. Chen • P. Hong • Z. Zhang • B. Roth • A. Korhonen • B. Plank
Human Label Variation as Stable Signal: Learning Annotator-Specific Explanation Behavior via Cross-Annotator Preference Optimization.
EMNLP 2026 - Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. Oral Presentation. To be published. Preprint available. arXiv GitHub
Human Label Variation as Stable Signal: Learning Annotator-Specific Explanation Behavior via Cross-Annotator Preference Optimization.
EMNLP 2026 - Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. Oral Presentation. To be published. Preprint available. arXiv GitHub
Y. Ding • M. Li • E. Garces Arias • M. Aßenmacher • C. Heumann • C. Zhang
Breaking the Likelihood Trap: Variance-Calibrated Modulation for Large Language Model Decoding.
EMNLP 2026 - Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. Oral Presentation. To be published. Preprint available. arXiv GitHub
Breaking the Likelihood Trap: Variance-Calibrated Modulation for Large Language Model Decoding.
EMNLP 2026 - Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. Oral Presentation. To be published. Preprint available. arXiv GitHub
L. Edman • D. Dementieva • A. Fraser
From Specialization to Generalization: Instruction-tuned LLMs for Robust Harmful Content Mitigation.
EMNLP 2026 - Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published. Preprint available. arXiv GitHub
From Specialization to Generalization: Instruction-tuned LLMs for Robust Harmful Content Mitigation.
EMNLP 2026 - Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published. Preprint available. arXiv GitHub
S. Gerstner • H. AlQuabeh • K. Inui • H. Schütze
Weakening Neurons: An Input-Output Functionality in Transformers with Outsize Influence.
EMNLP 2026 - Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published. Preprint available. arXiv
Weakening Neurons: An Input-Output Functionality in Transformers with Outsize Influence.
EMNLP 2026 - Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published. Preprint available. arXiv
H. Hu • H. Wu • A. Zhao • L. Ding • P. Yin • Y. Ma • X. Shen
Beyond FLOPs: Benchmarking Real Inference Acceleration of LLM Pruning under a GEMM-Centric Taxonomy.
EMNLP 2026 - Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published. Preprint available. arXiv GitHub
Beyond FLOPs: Benchmarking Real Inference Acceleration of LLM Pruning under a GEMM-Centric Taxonomy.
EMNLP 2026 - Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published. Preprint available. arXiv GitHub
F. Körner • M. Matveev • F. Eichin • G. Kutyniok • B. Plank • M. A. Hedderich
Copy First, Translate Later: Interpreting Translation Dynamics in Multilingual Pretraining.
EMNLP 2026 - Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published. Preprint available. arXiv
Copy First, Translate Later: Interpreting Translation Dynamics in Multilingual Pretraining.
EMNLP 2026 - Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published. Preprint available. arXiv
Y. Liu • J. Han • S. Yan • V. Tresp • Y. Ma
Routing-Free Mixture-of-Experts.
EMNLP 2026 - Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published. Preprint available. arXiv GitHub
Routing-Free Mixture-of-Experts.
EMNLP 2026 - Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published. Preprint available. arXiv GitHub
Y. Lu • Y. Liu • S. Gerstner • L. Hirlimann • J. Rohweder • H. Schütze
Relational Linearity is a Predictor of Hallucinations.
EMNLP 2026 - Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published. Preprint available. arXiv
Relational Linearity is a Predictor of Hallucinations.
EMNLP 2026 - Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published. Preprint available. arXiv
Y. Lu • A. Modarressi • Y. Liu • H. Schütze
Expert-Aware Causal Tracing of Factual Recall in Sparse MoE Language Models.
EMNLP 2026 - Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published. Preprint available. arXiv
Expert-Aware Causal Tracing of Factual Recall in Sparse MoE Language Models.
EMNLP 2026 - Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published. Preprint available. arXiv
T. Maye-Lasserre • Y. Li • B. Jian • M. Ghahremani • B. Wiestler • C. Wachinger
Generating Reports or Repeating Templates? Measuring and Mitigating Template Collapse in 3D CT Report Generation.
EMNLP 2026 - Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published. Preprint available. arXiv
Generating Reports or Repeating Templates? Measuring and Mitigating Template Collapse in 3D CT Report Generation.
EMNLP 2026 - Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published. Preprint available. arXiv
B. Nieth • M. Gracheva • M. Mahlberg • B. M. Eskofier • E. Salin
How Human-Like Are Large Language Models? A Register-Aware Linguistic Evaluation Framework.
EMNLP 2026 - Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. Oral Presentation. To be published. Preprint available. arXiv GitHub
How Human-Like Are Large Language Models? A Register-Aware Linguistic Evaluation Framework.
EMNLP 2026 - Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. Oral Presentation. To be published. Preprint available. arXiv GitHub
K. Rohrbacher • B. Nieth • E. Salin • B. M. Eskofier • M. Mahlberg
How LLMs Build Fictional Worlds: Setting and Narrative Space in AI-Generated Creative Storytelling.
EMNLP 2026 - Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published. Preprint available. arXiv GitHub
How LLMs Build Fictional Worlds: Setting and Narrative Space in AI-Generated Creative Storytelling.
EMNLP 2026 - Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published. Preprint available. arXiv GitHub
M. Sadeghi • P. T. L. Nguyen • A. Triantafyllopoulos • M. Habibpour • R. Richer • L. H. Rupp • L. Schindler-Gmelch • M. Keinert • M. Hager • B. W. Schuller • B. Egger • M. Berking • B. M. Eskofier
Cross-Corpus Depression Detection in Semi-Structured Clinical Interviews.
EMNLP 2026 - Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published.
Cross-Corpus Depression Detection in Semi-Structured Clinical Interviews.
EMNLP 2026 - Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published.
M. Tonneau • N. K. R. Sehgal • N. Malhotra • S. Kazemi • V. Orozco-Olvera • A. M. Muñoz Boudet • L. Subramanian • S. P. Fraiberger • S. C. Guntuku • V. Hofmann
Different Demographic Cues Yield Inconsistent Conclusions About LLM Personalization and Bias.
EMNLP 2026 - Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published. Preprint available.
Different Demographic Cues Yield Inconsistent Conclusions About LLM Personalization and Bias.
EMNLP 2026 - Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published. Preprint available.
J. Song • M. Zhao • Z. Wang • Y. Liu • Q. Li • S. Feng • F. Ren • D. Wang • H. Schütze
A Target-Centric Survey of Quantization-Aware Training.
EMNLP 2026 - Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published. Preprint available. arXiv
A Target-Centric Survey of Quantization-Aware Training.
EMNLP 2026 - Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published. Preprint available. arXiv
M. Tonneau • N. K. R. Sehgal • N. Malhotra • S. Kazemi • V. Orozco-Olvera • A. M. Munoz Boudet • L. Subramanian • S. P. Fraiberger • S. C. Guntuku • V. Hofmann
Different Demographic Cues Yield Inconsistent Conclusions About LLM Personalization and Bias.
EMNLP 2026 - Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published. Preprint available. arXiv
Different Demographic Cues Yield Inconsistent Conclusions About LLM Personalization and Bias.
EMNLP 2026 - Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published. Preprint available. arXiv
Y. Wang • T. Zhang • J. Bi • A. Aniri • W. Ye • B. Liu • S. Yan • S. Wang • X. Zhou • S. Pirk • H. Schütze • Y. Ma
MemWM: Memory-Augmented Text-Based World Model.
EMNLP 2026 - Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published. Preprint available. arXiv
MemWM: Memory-Augmented Text-Based World Model.
EMNLP 2026 - Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published. Preprint available. arXiv
S. Yan • S. Dong • H. Wang • E. Nie • Y. Liu • J. Bi • Y. Xu • S. Schwarzmann • R. Trivisonno • V. Tresp • Y. Ma
PyraVid: Hierarchical Multimodal Memory for Long-Horizon Video Reasoning.
EMNLP 2026 - Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published. Preprint available. arXiv
PyraVid: Hierarchical Multimodal Memory for Long-Horizon Video Reasoning.
EMNLP 2026 - Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published. Preprint available. arXiv
C. Yuan • B. Ma • Z. Zhang • B. Prenkaj • F. Kreuter • G. Kasneci
Moral Lenses, Political Coordinates: Towards Ideological Positioning of Morally Conditioned LLMs.
EMNLP 2026 - Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. Oral Presentation. To be published. Preprint available. arXiv
Moral Lenses, Political Coordinates: Towards Ideological Positioning of Morally Conditioned LLMs.
EMNLP 2026 - Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. Oral Presentation. To be published. Preprint available. arXiv
J. Zhang • Z. Li • R. Zhao • J. Lan • T. Seidl • M. A. Hedderich • H. Schütze • Y. Liu
Beyond Input Understanding: Diagnosing Multilingual Mathematical Reasoning with Directed Acyclic Trace Graphs.
EMNLP 2026 - Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published. Preprint available. arXiv
Beyond Input Understanding: Diagnosing Multilingual Mathematical Reasoning with Directed Acyclic Trace Graphs.
EMNLP 2026 - Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published. Preprint available. arXiv
Findings Track (11 papers)
L. Edman • A. Fraser
Toppling the Hierarchy in Byte-level Language Modeling.
Findings @EMNLP 2026 - Findings of the Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published. Preprint available. arXiv GitHub
Toppling the Hierarchy in Byte-level Language Modeling.
Findings @EMNLP 2026 - Findings of the Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published. Preprint available. arXiv GitHub
Y. Fan • X. Lu • A. Zhao • J. Tong • P. Nie • K. Zou • Y. Ma • W. Zhang • X. Shen
miniReranker: Efficient Multimodal Reranking through Visual Cache Reuse and Interaction Sparsity.
Findings @EMNLP 2026 - Findings of the Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published. Preprint available. arXiv
miniReranker: Efficient Multimodal Reranking through Visual Cache Reuse and Interaction Sparsity.
Findings @EMNLP 2026 - Findings of the Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published. Preprint available. arXiv
A. Fazla • A. Testoni • A. Abu-Hanna • B. Plank • I. Calixto
Uncertainty Is Not a Safety Net for Clinical VQA, but Can It Anticipate Model Failure?
Findings @EMNLP 2026 - Findings of the Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published. Preprint available. arXiv
Uncertainty Is Not a Safety Net for Clinical VQA, but Can It Anticipate Model Failure?
Findings @EMNLP 2026 - Findings of the Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published. Preprint available. arXiv
F. Ghorbanpour • C. Fierro • A. Fraser • A. Søgaard
Latent-Space Intervention for Cross-Lingual Factual Consistency: Consistency Improvements without Accuracy Drops.
Findings @EMNLP 2026 - Findings of the Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published. Preprint available. arXiv
Latent-Space Intervention for Cross-Lingual Factual Consistency: Consistency Improvements without Accuracy Drops.
Findings @EMNLP 2026 - Findings of the Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published. Preprint available. arXiv
A. D. Hakimi • L. Hirlimann • I. Augenstein • H. Schütze
Do We Still Need Humans in the Loop? Comparing Human and LLM Annotation in Active Learning for Hostility Detection.
Findings @EMNLP 2026 - Findings of the Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published. Preprint available. arXiv
Do We Still Need Humans in the Loop? Comparing Human and LLM Annotation in Active Learning for Hostility Detection.
Findings @EMNLP 2026 - Findings of the Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published. Preprint available. arXiv
M. A. Jradi • F. Ghorbanpour • A. Fraser
Attribute-Based Diagnosis of LLM Alignment with Hate Speech Annotations.
Findings @EMNLP 2026 - Findings of the Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published. Preprint available. arXiv
Attribute-Based Diagnosis of LLM Alignment with Hate Speech Annotations.
Findings @EMNLP 2026 - Findings of the Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published. Preprint available. arXiv
B. Muscato • B. Chen • G. Gezici • B. Plank • F. Giannotti
Disagreeing Rationales: Rethinking Classification and Explainability Evaluation in Hate Speech Detection.
Findings @EMNLP 2026 - Findings of the Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published. Preprint available. arXiv
Disagreeing Rationales: Rethinking Classification and Explainability Evaluation in Hate Speech Detection.
Findings @EMNLP 2026 - Findings of the Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published. Preprint available. arXiv
N. Nikeghbal • A. H. Kargaran • S. Kolli • J. Diesner
Who Flips? Self- and Cross-Model Counterarguments Reveal Answer Instability in LLMs.
Findings @EMNLP 2026 - Findings of the Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published. Preprint available. arXiv GitHub
Who Flips? Self- and Cross-Model Counterarguments Reveal Answer Instability in LLMs.
Findings @EMNLP 2026 - Findings of the Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published. Preprint available. arXiv GitHub
R. Pei • Y. Liu • S. Pyysalo • H. Schütze • S. Ji
Reasoning over Grammar: Can Synthetic Linguistic Reasoning Traces Enhance Low-Resource Machine Translation?
Findings @EMNLP 2026 - Findings of the Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published. Preprint available. arXiv GitHub
Reasoning over Grammar: Can Synthetic Linguistic Reasoning Traces Enhance Low-Resource Machine Translation?
Findings @EMNLP 2026 - Findings of the Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published. Preprint available. arXiv GitHub
B. Rabern • P. Mondorf • B. Plank
LogicSkills: A Structured Benchmark for Formal Reasoning in Large Language Models.
Findings @EMNLP 2026 - Findings of the Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published. Preprint available. arXiv
LogicSkills: A Structured Benchmark for Formal Reasoning in Large Language Models.
Findings @EMNLP 2026 - Findings of the Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published. Preprint available. arXiv
F. Sadrieh • M. Štefánik
Prompt-Robust Language Models: Which Training Strategies Work?
Findings @EMNLP 2026 - Findings of the Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published. Preprint available. arXiv
Prompt-Robust Language Models: Which Training Strategies Work?
Findings @EMNLP 2026 - Findings of the Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published. Preprint available. arXiv
Workshops (1 paper)
M. Kandlinger • B. Ma • A.-C. Haensch • M. Aßenmacher
Do LLMs Give Consistent Opinions? Evaluating Response Reliability Under Varying Likert-Scale Formulations in Survey-Style MCQA.
W-NUT @EMNLP 2026 - 11th Workshop on Natural User-generated Text at the Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published.
Do LLMs Give Consistent Opinions? Evaluating Response Reliability Under Varying Likert-Scale Formulations in Survey-Style MCQA.
W-NUT @EMNLP 2026 - 11th Workshop on Natural User-generated Text at the Conference on Empirical Methods in Natural Language Processing. Budapest, Hungary, Oct 24-29, 2026. To be published.
Related
15.10.2026
When Flatter Is Not Always Better
In their ICML 2026 paper, MCML researchers show why the flattest regions found during neural network training are not always the best.
13.10.2026
Björn Schuller: What Your Voice Reveals About Your Health
MCML PI Björn Schuller explores how AI can analyze the human voice to detect signs of disease and reveal insights into our health.
