India

Implicit hate speech has increasingly been recognized as a significant issue for social media platforms. While much of the research has traditionally focused on harmful speech in general, the need for generalizable techniques to detect veiled and subtle forms of hate has become increasingly pressing. Based on lexicon analysis, we hypothesize that implicit hate speech is already present in publicly available harmful speech datasets but may not have been explicitly recognized or labeled by annotators. Additionally, crowdsourced datasets are prone to mislabeling due to the complexity of the task and often influenced by annotators&#39; subjective interpretations. In this paper, we propose an approach to address the detection of implicit hate speech and enhance generalizability across diverse datasets by leveraging existing harmful speech datasets. Our method comprises three key components: influential sample identification, reannotation, and augmentation using Llama-3 70B and GPT-4o. Experimental results demonstrate the effectiveness of our approach in improving implicit hate detection, achieving a +12.9-point F1 score improvement compared to the baseline.

IJCNLP-AACL 2025

Towards Generalizable Generic Harmful Speech Datasets for Implicit Hate Speech Detection

influential data

implicit hate

hate speech

## Welcome to EMNLP 2022!
I am delighted to welcome you to EMNLP 2022! I believe this conference has been complicated beyond any precedent. Over the past year, it’s been thrilling to see the organization team approach each new puzzle with creativity and enthusiasm. We hope that those participating in Abu Dhabi as well as those joining remotely will leave the conference feeling newly inspired by the program and newly connected to our ever-growing community. Following EMNLP 2021 and major NLP conferences since, EMNLP 2022 is “hybrid,” serving both virtual and in-person participants.

Our key innovations for EMNLP 2022 include:

* EMNLP 2022 is “hybrid” in a second sense, as well: we allowed both direct and rolling review paper submissions, building on the pilot experiment of EMNLP 2021, which considered a small number of ARR submissions. 
* Familiar from NAACL but new to EMNLP, we’ve added an industry track.
* During the conference, “portals” will link virtual poster sessions to in-person conference participants during poster sessions each day.
* The first ACL-family conference in the United Arab Emirates.

 *Message from Noah A. Smith, University of Washington and Allen Institute for AI, Seattle, Washington, USA* 
***EMNLP 2022 General Chair***
 
[![](https://assets.underline.io/uploads/markdown_image/1/image/9eec7d4a287ee18c278b08229290aa83.png)](https://drive.google.com/file/d/1OlPv6QBeo62VVTughj2jkiLeyHd1WnUt/view)
 
[![](https://assets.underline.io/uploads/markdown_image/1/image/a3db7a768409f05192210d98601edb25.png)](https://emnlp2022.rocket.chat/)

To access this site you need to register. Please register [here](https://2022.emnlp.org/registration/).

Register here

EMNLP 2022

Welcome!
EMNLP 2022 will take place in Abu Dhabi from December 7th to December 11th, 2022. And it will be held in hybrid mode, both online and offline.

Implicit hate speech has increasingly been recognized as a significant issue for social media platforms. While much of the research has traditionally focused on harmful speech in general, the need for generalizable techniques to detect veiled and subtle forms of hate has become increasingly pressing. Based on lexicon analysis, we hypothesize that implicit hate speech is already present in publicly available harmful speech datasets but may not have been explicitly recognized or labeled by annotators. Additionally, crowdsourced datasets are prone to mislabeling due to the complexity of the task and often influenced by annotators' subjective interpretations. In this paper, we propose an approach to address the detection of implicit hate speech and enhance generalizability across diverse datasets by leveraging existing harmful speech datasets. Our method comprises three key components: influential sample identification, reannotation, and augmentation using Llama-3 70B and GPT-4o. Experimental results demonstrate the effectiveness of our approach in improving implicit hate detection, achieving a +12.9-point F1 score improvement compared to the baseline.

technical paper

### Welcome to IJCNLP-AACL 2025! 
 It is a great honor to host this joint conference in Mumbai, India, from December 20 to 24, 2025. The joint conferences of IJCNLP and AACL are organized with alternating leadership in the Asia-Pacific region. The event is run by the Asian Federation of Natural Language Processing (AFNLP) in odd years, and by AACL in even years, while it is organized solely by ACL when the annual ACL meeting is held in the region. This year, the conference is primarily organized by AFNLP. 
*Kentaro Inui
MBZUAI, UAE
General Chair, IJCNLP-AACL 2025* 
Read full message and download the Conference Handbook [**here**](https://drive.google.com/file/d/1UTwxkAqSqI-GAoJC3wE1zZt5VP1Y8GX0/view?usp=sharing).

The 14th IJCNLP & 4th AACL will be held in Mumbai, India from December 20th to December 24th, 2025.

Entity disambiguation (ED) links ambiguous mentions in text to entries in a knowledge base and is a core task in entity linking systems. While pretrained decoder-only language models (DLMs) offer strong generalization capabilities, their effective use in ED has been restricted due to sensitivity to candidate order, susceptibility to hallucinated outputs, and potential dataset leakage. We introduce Agnus a zero-shot ED framework that addresses these challenges through three core innovations: (1) order-invariant candidate encoding via shared positional embeddings and modified autoregressive attention masking, which eliminates bias on input ordering; (2) constrained decoding that ensures outputs are restricted to valid candidates, effectively preventing hallucinations; and (3) synthetic dataset creation approach as a diagnostic tool for data contamination detection and mitigation. Agnus eliminates up to 15.2% of F1 variability caused by candidate permutations, delivering consistent and order-robust predictions previously unattainable with autoregressive architectures. In our experiments, Agnus achieves state-of-the-art performance on four standard ED benchmarks, surpassing prior zero-shot approaches by an average 3.7% using small language models. We release code, data including candidate sets, and a synthetic benchmark to support reproducibility and controlled evaluation.

Agnus LLM: Robust and Flexible Entity Disambiguation with decoder-only Language Models

In many real-world settings like journalism, law, medicine, and science communication, information is passed from one person or system to another through multiple rounds of summarization or rewriting. This process, known as multi-hop information transfer, also happens increasingly in workflows involving large language models (LLMs). But while summarization models and factuality metrics have improved, we still don’t fully understand how meaning and factual accuracy hold up across long chains of transformations, especially when both humans and LLMs are involved.

In this paper, we take a fresh look at this problem by combining insights from cognitive science (Bartlett’s serial reproduction) and information theory (Shannon’s noisy-channel model). We build a new dataset of 700 five-step transmission chains that include human-only, LLM-only, mixed human-LLM, and cross-LLM settings across a wide range of source texts. To track how meaning degrades, we introduce three new metrics: Information Degradation Rate (IDR) for semantic drift, Meaning Preservation Entropy (MPE) for uncertainty in factual content, and Cascaded Hallucination Propagation Index (CHPI) for how hallucinations accumulate over time. Our findings reveal that hybrid chains behave asymmetrically. When a human summary is refined by a language model, the final output tends to preserve meaning well, suggesting that models can improve upon human-written summaries.  The code and data will be available at : https://github.com/transtrace6/TransTrace.git.

Who Remembers What? Tracing Information Fidelity in Human-AI Chains

This paper investigates the relationship between large language models’ (LLMs) ability to recognize repetitive input patterns and their performance on in-context learning (ICL). In contrast to prior work that has primarily focused on attention heads, we examine this relationship from the perspective of skill neurons, specifically repetition neurons. Our experiments reveal that the impact of these neurons on ICL performance varies depending on the depth of the layer in which they reside. By comparing the effects of repetition neurons and induction heads, we further identify strategies for reducing repetitive outputs while maintaining strong ICL capabilities.

Understanding and Controlling Repetition Neurons and Induction Heads in In-Context Learning

Human explanations of natural language, *rationales*, form a tool to assess whether models learn a label *for the right reasons* or rely on dataset-specific shortcuts. *Sufficiency* is a common metric for estimating the informativeness of rationales, but it provides limited insight into the effects of rationale information on model performance. We address this limitation by relating sufficiency to two modelling paradigms: the ability of models to identify which tokens are part of the rationale (through token classification) and the ability of improving model performance by incorporating rationales in the input (through attention regularisation). We find that highly informative rationales are not likely to help classify the instance correctly. Sufficiency conversely captures the classification impact of the non-rationalised context, which interferes with rationale information in the same input. We also find that incorporating rationale information in model inputs can boost cross-domain classification, but results are inconsistent per task and model type. Finally, sufficiency and token classification appear to be unrelated. These results exemplify the complexity of rationales, showing that metrics capable of systematically capturing this type of information merit further investigation.

Learning from *Sufficient* Rationales: Analysing the Relationship Between Explanation Faithfulness and Token-level Regularisation Strategies

Instruction-tuning plays a vital role in enhancing the task-solving abilities of large language models (LLMs), improving their usability in generating helpful responses on various tasks. However, previous work has demonstrated that they are sensitive to minor variations in instruction phrasing. In this paper, we explore whether introducing perturbations in instruction-tuning data can enhance LLMs' resistance against noisy instructions. We focus on how instruction-tuning with perturbations, such as removing stop words or shuffling words, affects LLMs' performance on the original and perturbed versions of widely-used benchmarks (MMLU, BBH, GSM8K). We further assess learning dynamics and potential shifts in model behavior. Surprisingly, our results suggest that instruction-tuning on perturbed instructions can, in some cases, improve downstream performance. These findings highlight the importance of including perturbed instructions in instruction-tuning, which can make LLMs more resilient to noisy user inputs.

Fine-Tuning on Noisy Instructions: Effects on Generalization and Performance

We introduce CLASSER, a cross-lingual annotation projection framework enhanced through script similarity, to create fine-grained named entity recognition (FgNER) datasets for low-resource languages. Manual annotation for named entity recognition (NER) is expensive, and distant supervision often produces noisy data that are often limited to high-resource languages. CLASSER employs a two-stage process: first projection of annotations from high-resource NER datasets to target language by using source-to-target parallel corpora and a projection tool built on a multilingual encoder, then refining them by leveraging datasets in script-similar languages. We apply this to five low-resource Indian languages: *Assamese*, *Marathi*, *Nepali*, *Sanskrit*, and *Bodo*, a vulnerable language. The resulting dataset comprises 1.8M sentences, 2.6M entity mentions and 24.7M tokens. Through rigorous analyses, the effectiveness of our method and the high quality of the resulting dataset are ascertained with F1 score improvements of 26% in Marathi and 46% in Sanskrit over the current state-of-the-art. We further extend our analyses to zero-shot and cross-lingual settings, systematically investigating the impact of script similarity and multilingualism on cross-lingual FgNER performance. The dataset is publicly available at [huggingface.co/datasets/prachuryyaIITG/CLASSER](https://huggingface.co/datasets/prachuryyaIITG/CLASSER).

CLASSER: Cross-lingual Annotation Projection enhancement through Script Similarity for Fine-grained Named Entity Recognition

Traditional dialogue summarization primarily focuses on dialogue content, assuming it comprises adequate information for a clear summary. However, this assumption often fails for discussions grounded in shared background, where participants frequently omit context and use implicit references. This results in summaries that are confusing to readers unfamiliar with the background. To address this, we introduce Knowledge-Grounded Discussion Summarization (KGDS), a novel task that produces a supplementary background summary for context and a clear opinion summary with clarified references. To facilitate research, we construct the first KGDS benchmark, featuring news-discussion pairs and expert-created multi-granularity gold annotations for evaluating sub-summaries. We also propose a novel hierarchical evaluation framework with fine-grained and interpretable metrics. Our extensive evaluation of 12 advanced large language models (LLMs) reveals that KGDS remains a significant challenge. The models frequently miss key facts and retain irrelevant ones in background summarization, and often fail to resolve implicit references in opinion summary integration.

What Are They Talking About? A Benchmark of Knowledge-Grounded Discussion Summarization

Trustworthy evaluations of dangerous capabilities are increasingly crucial for determining whether an AI system is safe to deploy. One empirically demonstrated threat is $\textit{sandbagging}$—the strategic underperformance on evaluations by AI models or their developers. A promising defense is to monitor a model's chain-of-thought (CoT) reasoning, as this could reveal its intentions and plans. In this work, we measure the ability of models to sandbag on dangerous capability evaluations against a CoT monitor by prompting them to sandbag while being either monitor-oblivious or monitor-aware. We show that both frontier models and small open-sourced models can $\textit{covertly}$ sandbag against CoT monitoring 0-shot without hints. However, they cannot yet do so reliably: they bypass the monitor 16-36\% of the time when monitor-aware, conditioned on sandbagging successfully. We qualitatively analyzed the uncaught CoTs to understand $\textit{why}$ the monitor failed. We reveal a rich attack surface for CoT monitoring and contribute five covert sandbagging policies generated by models. These results inform potential failure modes of CoT monitoring and may help build more diverse sandbagging model organisms.

LLMs Can Covertly Sandbag on Capability Evaluations Against Chain-of-Thought Monitoring

Audio Large Language Models (AudioLLMs) have achieved strong results in semantic tasks like speech recognition and translation, but remain limited in modeling paralinguistic cues such as emotion. Existing approaches often treat emotion understanding as a classification problem, offering little insight into the underlying rationale behind predictions. In this work, we explore emotion reasoning, a strategy that leverages the generative capabilities of AudioLLMs to enhance emotion recognition by producing semantically aligned, evidence-grounded explanations. To support this in multitask AudioLLMs, we introduce a unified framework combining reasoning-augmented data supervision, dual-encoder architecture, and task-alternating training. This approach enables AudioLLMs to effectively learn different tasks while incorporating emotional reasoning. Experiments on IEMOCAP and MELD show that our approach not only improves emotion prediction accuracy but also enhances the coherence and evidential grounding of the generated responses. Experiments on two out-of-domain datasets demonstrate the generalization capabilities of the resulting model.

Beyond Classification: Towards Speech Emotion Reasoning with Multitask AudioLLMs

Language is, as commonly theorized, largely arbitrary. Yet, systematic relationships between phonetics and semantics have been observed in many specific cases. To what degree could those systematic relationships manifest themselves in large scale, quantitative investigations--both in previously identified and unidentified phenomena? This work undertakes a distributional approach to quantifying phonosemantic iconicity at scale across 6 diverse languages (English, Spanish, Hindi, Finnish, Turkish, and Tamil). In each language, we analyze the alignment of morphemes' phonetic and semantic similarity spaces with a suite of statistical measures, and discover an array of interpretable phonosemantic alignments not previously identified in the literature, along with crosslinguistic patterns. We also analyze 5 previously hypothesized phonosemantic alignments, finding support for some such alignments and mixed results for others.

Downloads

Next from IJCNLP-AACL 2025

Agnus LLM: Robust and Flexible Entity Disambiguation with decoder-only Language Models

Similar lecture

Leveraging World Knowledge in Implicit Hate Speech Detection