Austria

One of the most widely used tasks to evaluate Large Language Models (LLMs) is Multiple-Choice Question Answering (MCQA). While open-ended question answering tasks are more challenging to evaluate, MCQA tasks are, in principle, easier to assess, as the model&#39;s answer is thought to be simple to extract and is directly compared to a set of predefined choices. However, recent studies have started to question the reliability of MCQA evaluation, showing that multiple factors can significantly impact the reported performance of LLMs, especially when the model generates free-form text before selecting one of the answer choices. In this work, we shed light on the inconsistencies of MCQA evaluation strategies, which can lead to inaccurate and misleading model comparisons. We systematically analyze whether existing answer extraction methods are aligned with human judgment, and how they are influenced by answer constraints in the prompt across different domains. Our experiments demonstrate that traditional evaluation strategies often underestimate LLM capabilities, while LLM-based answer extractors are prone to systematic errors. Moreover, we reveal a fundamental trade-off between including format constraints in the prompt to simplify answer extraction and allowing models to generate free-form text to improve reasoning. Our findings call for standardized evaluation methodologies and highlight the need for more reliable and consistent MCQA evaluation practices.

ACL 2025

Right Answer, Wrong Score: Uncovering the Inconsistencies of LLM Evaluation in Multiple-Choice Question Answering

regex

answer extractor

mcqa

evaluation

adversarial

### Welcome!
The 62nd Annual Meeting of the Association for Computational Linguistics (ACL 2024) will take place in Bangkok, Thailand from August 11th to 16th, 2024. Our Virtual Poster Sessions will take place online Thursday, August 22, 2024.

You are required to register for this event. **Please register [here](https://2024.aclweb.org/registration). **

If you have already registered, please check your inbox for an email from Underline granting you access to ACL 2024 content.

Please register!

ACL 2024

The 62nd Annual Meeting of the Association for Computational Linguistics (ACL 2024) will take place in Bangkok, Thailand from August 11th to 16th, 2024. More information will be announced soon.

One of the most widely used tasks to evaluate Large Language Models (LLMs) is Multiple-Choice Question Answering (MCQA). While open-ended question answering tasks are more challenging to evaluate, MCQA tasks are, in principle, easier to assess, as the model's answer is thought to be simple to extract and is directly compared to a set of predefined choices. However, recent studies have started to question the reliability of MCQA evaluation, showing that multiple factors can significantly impact the reported performance of LLMs, especially when the model generates free-form text before selecting one of the answer choices. In this work, we shed light on the inconsistencies of MCQA evaluation strategies, which can lead to inaccurate and misleading model comparisons. We systematically analyze whether existing answer extraction methods are aligned with human judgment, and how they are influenced by answer constraints in the prompt across different domains. Our experiments demonstrate that traditional evaluation strategies often underestimate LLM capabilities, while LLM-based answer extractors are prone to systematic errors. Moreover, we reveal a fundamental trade-off between including format constraints in the prompt to simplify answer extraction and allowing models to generate free-form text to improve reasoning. Our findings call for standardized evaluation methodologies and highlight the need for more reliable and consistent MCQA evaluation practices.

poster

### 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)

In the last couple years, there has been a flood of interest in studying the extent to which language models (LMs) have a theory of mind (ToM) — the ability to ascribe mental states to themselves and others. The results provide an unclear picture of the current state of the art, with some finding near-human performance and others near-zero. To make sense of this landscape, we perform a survey of 15 recent studies aimed at measuring ToM in LMs and find that, while almost all perform checks for human identifiable issues, less than half do so for patterns only a machine might exploit. Among those which do perform such validation, none identify LMs to exceed human performance. We conclude that these datasets are easier than their peers, likely due to the presence of spurious patterns in the data, and we caution against building ToM benchmarks relying solely on human validation of the data.

Machine Theory of Mind Needs Machine Validation

Detecting ambiguity is important for language understanding, including uncertainty estimation, humour detection, and processing garden path sentences. We assess language models' sensitivity to ambiguity by introducing an adversarial ambiguity dataset that includes syntactic, lexical, and phonological ambiguities along with adversarial variations (e.g., word-order changes, synonym replacements, and random-based alterations). Our findings show that direct prompting fails to robustly identify ambiguity, while linear probes trained on model representations can decode ambiguity with high accuracy, sometimes exceeding 90%. Our results offer insights into the prompting paradigm and how language models encode ambiguity at different layers.

Trick or Neat: Adversarial Ambiguity and Language Model Evaluation

We investigate static and contextualized embeddings for English pseudowords across a variety of Large Language Models (LLMs), to study (i) how these models represent semantic attributes of strings they encounter for the very first time and how (ii) these representations interact with sentence context. We zoom in a key semantic attribute, valence, which plays an important role in theories of language processing, acquisition, and evolution. In three experiments, we show that pseudoword valence is encoded in meaningful ways both in isolation and in context, and that pseudowords affect the representation of whole sentences similarly to words. This highlights how, at least for most LLMs we surveyed, pseudowords and words are not qualitatively different constructs. Our study confirms that LLMs capture systematic mappings between form and valence, and shows that embeddings of novel words are sufficiently stable in context, largely preserving their valence connotations. Our findings provide a computational exploration of how sub-lexical distributional patterns influence the valence of novel strings, offering useful insights for theories on the form-meaning interface and how it affects language learning and processing.

A puyfred feels less of a puyfred if you say it's cute, but it still feels bad: context-dependent form-meaning systematicity in LLMs

Large language models (LLMs) exhibit pronounced conservative bias in relation extraction tasks, frequently defaulting to no_relation label when an appropriate option is unavailable. While this behavior helps prevent incorrect relation assignments, our analysis reveals that it also leads to significant information loss when reasoning is not explicitly included in the output. We systematically evaluate this trade-off across multiple prompts, datasets, and relation types, introducing the concept of Hobson's choice to capture scenarios where models opt for safe but uninformative labels over hallucinated ones. Our findings suggest that conservative bias occurs twice as often as hallucination. To quantify this effect, we use SBERT and LLM prompts to capture the semantic similarity between conservative bias behaviors in constrained prompts and labels generated from semi-constrained and open-ended prompts.

Conservative Bias in Large Language Models: Measuring Relation Predictions

Large-scale multilingual evaluations, such as MEGA, often include only a handful of African languages due to the scarcity of high-quality evaluation data and the limited discoverability of existing African datasets. This lack of representation hinders comprehensive LLM evaluation across a diverse range of languages and tasks. To address these challenges, we introduce AfroBench - a multi-task benchmark for evaluating the performance of LLMs across 64 African languages, 15 tasks and 22 datasets. AfroBench consists of nine natural language understanding datasets, five text generation datasets, five knowledge and question answering tasks, and one mathematical reasoning task. We present results comparing the performance of prompting LLMs to fine-tuned baselines based on BERT and T5-style models. Our results suggest large gaps in performance between high-resource languages, such as English, and African languages across most tasks; but performance also varies based on the availability of monolingual data resources. Our findings confirm that performance on African languages continues to remain a hurdle for current LLMs, underscoring the need for additional efforts to close this gap

AfroBench: How Good are Large Language Models on African Languages?

Large language models are trained on massive scrapes of the web, as required by current scaling laws. Most progress is made for English, given its abundance of high-quality pretraining data. For most other languages, however, such high quality pretraining data is unavailable. In this work, we study how to boost pretrained model performance in a target language with insufficient pretraining data for training a high performing language model by enlisting data from an auxiliary language for which high quality data is available. We study this by quantifying the performance gap between training with data in a data-rich auxiliary language compared with training in the target language, exploring the benefits of translation systems, studying the limitations of model scaling when data is limited in the target languages, and proposing new methods for upsampling data from the auxiliary language. Our results show that stronger auxiliary datasets result in performance gains without modification to the model or training objective for close languages, and, in particular, that performance gains due to the development of more information-rich English pretraining datasets can extend to targeted language settings with limited data.

Training Bilingual LMs with Data Constraints in the Targeted Language

Query expansion methods powered by large language models (LLMs) have demonstrated effectiveness in zero-shot retrieval tasks. These methods assume that LLMs can generate hypothetical documents that, when incorporated into a query vector, enhance the retrieval of real evidence. However, we challenge this assumption by investigating whether knowledge leakage in benchmarks contributes to the observed performance gains. Using fact verification as a testbed, we analyzed whether the generated documents contained information entailed by ground truth evidence and assessed their impact on performance. Our findings indicate that performance improvements occurred consistently only for claims whose generated documents included sentences entailed by ground truth evidence. This suggests that knowledge leakage may be present in these benchmarks, potentially inflating the perceived performance of query expansion methods, particularly in real-world scenarios that require retrieving niche or novel knowledge.

Hypothetical Documents or Knowledge Leakage? Rethinking LLM-based Query Expansion

This study investigates the linguistic understanding of Large Language Models (LLMs) regarding signifier (form) and signified (meaning) by distinguishing two LLM assessment paradigms: psycholinguistic and neurolinguistic. Traditional psycholinguistic evaluations often reflect statistical rules that may not accurately represent LLMs' true linguistic competence. We introduce a neurolinguistic approach, utilizing a novel method that combines minimal pair and diagnostic probing to analyze activation patterns across model layers. This method allows for a detailed examination of how LLMs represent form and meaning, and whether these representations are consistent across languages. We found: (1) Psycholinguistic and neurolinguistic methods reveal that language performance and competence are distinct; (2) Direct probability measurement may not accurately assess linguistic competence; (3) Instruction tuning won't change much competence but improve performance; (4) LLMs exhibit higher competence and performance in form compared to meaning. Additionally, we introduce new conceptual minimal pair datasets for Chinese (COMPS-ZH) and German (COMPS-DE), complementing existing English datasets.

Large Language Models as Neurolinguistic Subjects: Discrepancy in Performance and Competence for Form and Meaning

Knowing which latent conditions lead to a particular outcome is useful for critically examining claims made about complex event outcomes. Identifying implied conditions and examining their influence on an outcome is challenging. We handle this by combining and augmenting annotations from two existing datasets consisting of goals and states, and explore the influence of conditions through our research questions and Condition-based Reasoning tasks. We examine open and closed LLMs of varying sizes and intent-alignment on our reasoning tasks and find that conditions are useful when not all context is available. Models differ widely in their ability to generate and identify outcome-variant conditions, which affects their performance on outcome validation, when conditions are used to replace missing context. Larger models like GPT-4o, are more cautious in such less constrained situations.

CoRE: Condition-based Reasoning for Identifying Outcome Variance in Complex Events

Homonymy can easily raise lexical ambiguity due to the misunderstanding of its multiple senses. Correct recognition of homonym sense greatly relies on its surrounding context. This ambiguous nature makes homonyms an appropriate testbed for examining the contextualization capability of pre-trained (PLM) and large language models (LLMs). Considering the impact of part-of-speech (POS) on homonym disambiguation and the dominance of English in word embedding research, this study provides a comprehensive layer-wise analysis of homonym representations in both English and Chinese, spanning same and different POS categories, across four families of PLMs/LLMs (BERT, GPT-2, Llama 3, Qwen 2.5). Through the creation of a synthetic dataset and computation of disambiguation score (D-Score), we found that: (1) no universal layer depth was found which excels in differentiating homonym representations; (2) bidirectional models produce better contextualized homonym representations compared to much larger autoregressive models; (3) most importantly, POS affects homonym representations in models in ways that slightly differ from human research findings. The individual differences between LLMs uncovered in our study challenge the simplistic understanding of their inner workings. This reveals a compelling research frontier: conducting controlled experiments with purposefully manipulated inputs to enhance the interpretability of LLMs. We have made our dataset and codes available publicly at https://anonymous.4open.science/r/ehril/.

Downloads

Next from ACL 2025

Machine Theory of Mind Needs Machine Validation

Similar lecture

Artifacts or Abduction: How Do LLMs Answer Multiple-Choice Questions Without the Question?