MITIGATION STRATEGIES

Post-ASR Correction in Hindi: Comparing Language Models and Large Language Models in Low-Resource Scenarios (2026.eacl-short)

Challenge:	Current mitigation strategies fail to preserve contextual reasoning capabilities in risky scenarios, leading to systemic risks for legal compliance.
Approach:	They propose to use reinforcement learning with a rule-based reward to incentivize contextual reasoning capabilities while enhancing compliance with safety and privacy norms.
Outcome:	The proposed model outperforms Qwen2.5-7B-Instruct model in safety and privacy benchmarks and achieves +8.58% accuracy improvement.

Copied to clipboard

Rishabh Kumar, Amrith Krishna, Ganesh Ramakrishnan, Preethi Jyothi

Spurious Correlations and Beyond: Understanding and Mitigating Shortcut Learning in SDOH Extraction with Large Language Models (2025.acl-short)

Challenge:	Automatic Speech Recognition (ASR) systems for low-resource languages produce erroneous transcripts due to limited annotated data and linguistic complexity.
Approach:	They compare language models and large language models for post-ASR correction in Hindi . they observe a scaling trend under zero-shot ICL where mid-sized LLMs degrade performance before marginal recovery at extreme scales.
Outcome:	The proposed model outperforms larger models in both fine-tuning and in-context learning settings.

Copied to clipboard

Fardin Ahsan Sakib, Ziwei Zhu, Karen Trister Grace, Meliha Yetisgen, Ozlem Uzuner

Large Vision-Language Model Alignment and Misalignment: A Survey Through the Lens of Explainability (2025.findings-emnlp)

Challenge:	Large language models (LLMs) rely on superficial cues leading to spurious predictions . recent work has highlighted how LLMs exploit spurious patterns rather than learning causal, generalizable features.
Approach:	They use a social history annotation corpus dataset to examine drug status extraction . they evaluate prompt engineering and chain-of-thought reasoning to reduce false positives .
Outcome:	The proposed model can predict drug use when alcohol or smoking is not present, while uncovering gender disparities in model performance.

Copied to clipboard

Dong Shu, Haiyan Zhao, Jingyu Hu, Weiru Liu, Ali Payani, Lu Cheng, Mengnan Du

Explainability & Interpretability in NLP

Challenge:	Large Vision-Language Models have demonstrated remarkable capabilities in processing both visual and textual information.
Approach:	They examine the challenge of alignment and misalignment in LVLMs through an explainability lens.
Outcome:	The findings highlight the need for standardized evaluation protocols and in-depth explainability studies.

When Format Changes Meaning: Investigating Semantic Inconsistency of Large Language Models (2025.findings-emnlp)

Copied to clipboard

Cheongwoong Kang, Jongeun Baek, Yeonjea Kim, Jaesik Choi

“Not Aligned” is Not “Malicious”: Being Careful about Hallucinations of Large Language Models’ Jailbreak (2025.coling-main)

Challenge:	Large language models are vulnerable to semantic inconsistency, a study finds . minor formatting variations result in divergent predictions for semantically equivalent inputs.
Approach:	They evaluate LLMs for semantic inconsistency and find they remain vulnerable . they propose to use mechanistic analysis to develop models that improve their reliability .
Outcome:	The proposed model is vulnerable to semantic inconsistency, the authors show . their model is brittle even in state-of-the-art models, they say .

Copied to clipboard

Lingrui Mei, Shenghua Liu, Yiwei Wang, Baolong Bi, Jiayi Mao, Xueqi Cheng

DAST: Difficulty-Adaptive Slow-Thinking for Large Reasoning Models (2025.emnlp-industry)

Challenge:	“Jailbreak” is a major safety concern of Large Language Models (LLMs).
Approach:	They propose a benchmarking framework to evaluate "jailbreak" outputs . they propose specialized validation framework to ensure outputs are useful malicious instructions .
Outcome:	The proposed framework enhances existing benchmarks to ensure outputs are useful . it also aims to evaluate the true potential of jailbroken outputs to cause harm to human society.

Copied to clipboard

Yi Shen, Jian Zhang, Jieyun Huang, Shuming Shi, Wenjing Zhang, Jiangze Yan, Ning Wang, Kai Wang, Zhaoxiang Liu, Shiguo Lian

Green & Sustainable NLP

Challenge:	Recent advances in slow-thinking reasoning models have shown exceptional performance in complex reasoning tasks.
Approach:	They propose a framework that enables models to automatically adjust Chain-of-Thought (CoT) length based on problem difficulty.
Outcome:	The proposed framework penalizes inefficiency on simple problems while incentivizing deep reasoning for complex ones.

Richer Countries and Richer Representations (2022.findings-acl)

Copied to clipboard

Kaitlyn Zhou, Kawin Ethayarajh, Dan Jurafsky

Representation Learning

Challenge:	Using BERT, countries with low frequency in training data are less likely to be invocabulary, and are less frequently predicted in the masked language modeling task.
Approach:	They propose three criteria to characterize the quality of representations for particular entities or groups: consistency, distinctiveness, and recognizability.
Outcome:	The results suggest that frequency is highly correlated with a country’s GDP, perpetuating historic power and wealth inequalities.

Looking at Radiology Report Generation through a Causal Lens: A Survey (2026.acl-long)

Copied to clipboard

Satyam Kumar, Kaustubh Shivshankar Shejole, Pushpak Bhattacharyya

Mitigating Covertly Unsafe Text within Natural Language Systems (2022.findings-emnlp)

Challenge:	Existing surveys on RRG emphasize deep learning while overlooking the critical role of causality.
Approach:	They propose to analyze biases across the RRG pipeline and formalize it as a causal modeling problem and review representative causal techniques from the literature.
Outcome:	The proposed model can mitigate biases and yield fair, reliable systems with clinically meaningful outputs.

Copied to clipboard

Alex Mei, Anisha Kabir, Sharon Levy, Melanie Subbiah, Emily Allaway, John Judge, Desmond Patton, Bruce Bimber, Kathleen McKeown, William Yang Wang

Unveiling Privacy Risks in Multi-modal Large Language Models: Task-specific Vulnerabilities and Mitigation Challenges (2025.findings-acl)

Challenge:	Existing studies on text safety have focused on overtly unsafe, covertly, or indirectly unsafe statements.
Approach:	They propose a method to identify physical harm-causing statements as overtly, covertly or indirectly unsafe and a solution to mitigate the generation of such statements.
Outcome:	The proposed methods identify the type of unsafe language that can cause physical harm and identify mitigation strategies to inspire future researchers to tackle this challenging problem.

Copied to clipboard

Tiejin Chen, Pingzhi Li, Kaixiong Zhou, Tianlong Chen, Hua Wei

Semantic Annotation for Improved Safety in Construction Work (2020.lrec-1)

Challenge:	Privacy risks in text-only Large Language Models are well-documented, especially their tendency to memorize and leak sensitive information.
Approach:	They propose a dataset to assess privacy risks across multi-modal tasks and scenarios . they demonstrate how models leak sensitive data across various tasks .
Outcome:	The proposed model can leak sensitive data embedded in images or stored in memory, exposing privacy risks.

Copied to clipboard

Paul Thompson, Tim Yates, Emrah Inan, Sophia Ananiadou

Information Extraction & Text Mining

Challenge:	a number of documents provide evidence of previous incidents and mitigation strategies . but information about previous projects with similar attributes is often hidden within . a new named entity annotation scheme is being developed for construction safety .
Approach:	a team of four health and safety experts have developed a named entity annotation scheme for construction safety documents.
Outcome:	a new named entity annotation scheme annotates 600 sentences from accident reports . the scheme has an average agreement rate of 0.79 F-Score .

When Audio and Text Disagree: Revealing Text Bias in Large Audio-Language Models (2025.emnlp-main)

Copied to clipboard

Cheng Wang, Gelei Deng, Xianglin Yang, Han Qiu, Tianwei Zhang

Bias Beyond English: Counterfactual Tests for Bias in Sentiment Analysis in Four Languages (2023.findings-acl)

Challenge:	Large Audio-Language Models (LALMs) are augmented with the ability to perceive audio, but their reliability when faced with conflicting inputs remains largely unexplored.
Approach:	They examine how LALMs prioritize information when presented with inconsistent audio-text pairs.
Outcome:	The proposed models display a significant bias toward textual input when presented with inconsistent audio-text pairs.

Copied to clipboard

Seraphina Goldfarb-Tarrant, Adam Lopez, Roi Blanco, Diego Marcheggiani

Sentiment Analysis

Challenge:	Sentiment analysis systems are used in hundreds of products and languages . Gender and racial biases are well-studied in English, but understudied elsewhere .
Approach:	They build a counterfactual evaluation corpus for gender and racial/migrant bias in four languages.
Outcome:	The evaluation corpus reveals which models have less bias and pinpoints changes in model bias behaviour, enabling more targeted mitigation strategies.

Jailbreak Open-Sourced Large Language Models via Enforced Decoding (2024.acl-long)

Copied to clipboard

Hangfan Zhang, Zhimeng Guo, Huaisheng Zhu, Bochuan Cao, Lu Lin, Jinyuan Jia, Jinghui Chen, Dinghao Wu

Semantic Text Processing

Challenge:	Existing studies show that Large Language Models can be misused to generate undesired content.
Approach:	They propose to use large language models to manipulate the generation process to generate undesired content without heavy computations or prompt designs.
Outcome:	The proposed method shows that open-sourced large language models could be misused to generate undesired content without heavy computations or prompt designs.

The Effect of Scaling, Retrieval Augmentation and Form on the Factual Consistency of Language Models (2023.emnlp-main)

Copied to clipboard

Lovisa Hagström, Denitsa Saynova, Tobias Norlund, Moa Johansson, Richard Johansson

Information Retrieval

Challenge:	Large Language Models (LLMs) are useful interfaces to factual knowledge, but their usefulness is limited by their tendency to deliver inconsistent answers to semantically equivalent questions.
Approach:	They evaluate the effectiveness of up-scaling and augmenting the LM with a passage retrieval database to reduce inconsistency.
Outcome:	The proposed models reduce inconsistency but retrieval augmentation is more efficient.

Unveiling Selection Biases: Exploring Order and Token Sensitivity in Large Language Models (2024.findings-acl)

Copied to clipboard

Sheng-Lun Wei, Cheng-Kuang Wu, Hen-Hsen Huang, Hsin-Hsi Chen

Semantic Text Processing

Challenge:	Using zero-shot or few-shot prompting, Large Language Models have been widely adopted in downstream applications.
Approach:	They propose to quantify the impact of option order and token usage on LLMs and propose mitigation strategies to enhance model performance.
Outcome:	The proposed mitigation strategies improve model performance and reduce the impact of token and order sensitivity on LLMs.

Measure and Improve Robustness in NLP Models: A Survey (2022.naacl-main)

Copied to clipboard

Xuezhi Wang, Haohan Wang, Diyi Yang

Robustness in NLP

Challenge:	Despite the performance gains, NLP models are still fragile and brittle to out-of-domain data, adversarial attacks, or small perturbation to the input.
Approach:	They propose a survey of how to define, measure and improve robustness in NLP by connecting multiple definitions of robustness and identifying failures.
Outcome:	The proposed models are robust against unseen or challenging scenarios, but are still fragile and brittle to out-of-domain data and adversarial attacks.

Safety-Utility Conflicts Are Not Global: Surgical Alignment via Head-Level Diagnosis (2026.acl-long)

Copied to clipboard

Wang Cai, Yilin Wen, Jinchang Hou, Du Su, Guoqiu Wang, Zhonghou Lv, Chenfu Bao, Yunfang Wu

Mitigating Hallucinations in Large Vision-Language Models (LVLMs) via Language-Contrastive Decoding (LCD) (2024.findings-acl)

Challenge:	Existing mitigation strategies rely on global gradient geometry to resolve alignment conflicts . however, they overlook Modular Heterogeneity within Transformers, resulting in suboptimal trade-offs . Conflict-Aware Sparse Tuning (CAST) combines head-level diagnosis with sparse fine-tuning .
Approach:	They propose a framework that integrates head-level diagnosis with sparse fine-tuning to address this limitation.
Outcome:	The proposed framework integrates head-level diagnosis with sparse fine-tuning to reduce alignment conflicts in LLMs.

Copied to clipboard

Avshalom Manevich, Reut Tsarfaty

“Yuki Gets Sushi, David Gets Steak?”: Uncovering Gender and Racial Biases in LLM-Based Meal Recommendations (2026.eacl-long)

Challenge:	Large Vision-Language Models (LVLMs) often produce object hallucinations due to their reliance on text cues and learned object co-occurrence biases.
Approach:	They propose a language-contrasting decoding algorithm that adjusts LVLM outputs based on LLM confidence levels to mitigate object hallucinations.
Outcome:	The proposed method shows up to %4 improvement in POPE F1 scores and %36 reduction in CHAIR scores on COCO validation set while improving captioning quality scores.

Copied to clipboard

Xuefeng Wei, Xuan Zhou, Yusuke Sakai, Taro Watanabe

The Reasoning Trap: How Enhancing LLM Reasoning Amplifies Tool Hallucination (2026.acl-long)

Challenge:	Large language models inherit and amplify societal biases related to gender and race.
Approach:	They use a USChainMains dataset to evaluate group bias in Large Language Models . they found that LLMs recommend meals with higher levels of adverse nutrients for names associated with Black, Hispanic, or male individuals .
Outcome:	The proposed model scales improves overall recommendation healthfulness but is insufficient to eliminate the healthfulness gap between demographic groups.

Copied to clipboard

Chenlong Yin, Zeyang Sha, Shiwen Cui, Changhua Meng, Zechao Li

Measuring and Mitigating Constraint Violations of In-Context Learning for Utterance-to-API Semantic Parsing (2023.findings-emnlp)

Challenge:	Recent studies suggest that strengthening reasoning often coincides with increased hallucination . however, no prior work has examined whether reasoning enhancement itself causes tool hallucinism .
Approach:	They propose a diagnostic benchmark measuring tool hallucination in two failure modes . they demonstrate a causal relationship between enhancing reasoning and tool hallubulation .
Outcome:	The proposed benchmark measures tool hallucination in two failure modes: no tool available, and (ii) only distractor tools available.

Copied to clipboard

Shufan Wang, Sébastien Jean, Sailik Sengupta, James Gung, Nikolaos Pappas, Yi Zhang

Semantic Text Processing

Challenge:	In task-oriented semantic parsing, the system aims to translate users’ utterances in natural language to machine-interpretable programs (API calls) However, Large Language Models (LLMs) are known to hallucinate and therefore pose a formidable challenge in constraining generated content.
Approach:	They propose to use large language models to translate user's utterances to machine-interpretable programs (API calls) they identify constraints violations in task-oriented utterrances and define fine-grained metrics that complement traditional ones.
Outcome:	The proposed methods reduce constraints violations and improve quality of the generated API calls, but require careful consideration given their implementation complexity and latency.

Automated Profile Inference with Language Model Agents (2026.findings-acl)

Copied to clipboard

Yuntao Du, Zitao Li, Bolin Ding, Yaliang Li, Hanshen Xiao, Jingren Zhou, Ninghui Li

CausalDetox: Causal Head Selection and Intervention for Language Model Detoxification (2026.findings-acl)

Challenge:	Existing privacy protections for large language models (LLMs) are limited due to the potential for malicious applications.
Approach:	They propose an automated profile inference framework that can extract personal information from public online activities by an adversary with the help of large language model (LLM) based agents.
Outcome:	The proposed framework is highly effective and efficient and the inferred attributes are both identifiable and sensitive, posing significant privacy risks.

Copied to clipboard

Yian Wang, Yuen Chen, Agam Goyal, Hari Sundaram

Fighting Randomness with Randomness: Mitigating Optimisation Instability of Fine-Tuning using Delayed Ensemble and Noisy Interpolation (2024.findings-emnlp)

Challenge:	Large language models (LLMs) frequently generate toxic content, posing significant risks for safe deployment.
Approach:	They propose a framework that identifies and intervenes on the specific attention heads causally responsible for toxic generation.
Outcome:	The proposed framework reduces toxic generation by 5.34% while preserving linguistic fluency and speeding up head selection.

Copied to clipboard

Branislav Pecher, Jan Cegin, Robert Belanec, Jakub Simko, Ivan Srba, Maria Bielikova

To Lie or Not to Lie? Investigating The Biased Spread of Global Lies by LLMs (2026.acl-long)

Challenge:	Despite the success of fine-tuning, it still displays model performance instability, especially with limited data.
Approach:	They propose a new mitigation strategy that leverages the strengths of ensembling, noise regularisation and model interpolation while retaining computational efficiency.
Outcome:	The proposed mitigation strategy outperforms the best performing mitigation strategy (Ensemble) while using only a fraction of its cost.

Copied to clipboard

Zohaib Khan, Mustafa Dogan, Ifeoma Okoh, Pouya Sadeghi, Siddhartha Shrestha, Sergius Justus Chesami Nyah, Mahmoud O. Mokhiamar, Michael J Ryan, Tarek Naous

Among Us: Measuring and Mitigating Malicious Contributions in Model Collaboration Systems (2026.acl-long)

Challenge:	Misinformation is on the rise, and the strong writing capabilities of LLMs lower the barrier for malicious actors to produce and disseminate false information.
Approach:	They introduce a multilingual parallel dataset of 440 misinformation generation prompt templates and 6,867 entities, spanning 8 languages and 195 countries.
Outcome:	The proposed model reduces misinformation generation across languages and countries . it also reduces the risk of misinformation being spread across countries based on the model's performance .

Copied to clipboard

Ziyuan Yang, Wenxuan Ding, Shangbin Feng, Yulia Tsvetkov

Why LLM Safety Guardrails Collapse After Fine-tuning: A Similarity Analysis Between Alignment and Fine-tuning Datasets (2026.acl-long)

Challenge:	Existing research is leveraging multiple language models with diverse skills and strengths to collaborate.
Approach:	They propose mitigation strategies to mitigate the impact of malicious models by employing external supervisors to disable/mask them out to reduce their influence.
Outcome:	The proposed mitigation strategies recover 95.31% of initial performance while making model collaboration systems fully resistant to malicious models remains an open question.

Copied to clipboard

Lei Hsiung, Tianyu Pang, Yung-Chen Tang, Linyue Song, Tsung-Yi Ho, Pin-Yu Chen, Yaoqing Yang

A Picture is Worth a Thousand Words? An Empirical Study of Aggregation Strategies for Visual Financial Document Retrieval (2026.findings-acl)

Challenge:	Existing mitigation strategies focus on reactively addressing jailbreak incidents after safety guardrails have been compromised.
Approach:	They investigate the degradation of safety guardrails through the lens of representation similarity between upstream alignment datasets and downstream fine-tuning tasks.
Outcome:	The proposed model reduces harmfulness score by 10.33% when compared to baseline models.

Copied to clipboard

Ho Hung Lim, Yi Yang

Information Retrieval

Challenge:	Visual RAG is an alternative to traditional RAG, but it requires hundreds of patch tokens per document to retrieve and store information.
Approach:	They propose to aggregate documents into a single vector to avoid semantic loss . they find global texture dominance is the root cause of this loss - they say .
Outcome:	The proposed model shows that aggregation obscures semantic changes in financial documents . global texture dominance is the root cause, and the model scales are consistent across models and embeddings.

Trust Me, I’m Wrong: LLMs Hallucinate with Certainty Despite Knowing the Answer (2025.findings-emnlp)

Copied to clipboard

Adi Simhi, Itay Itzhak, Fazl Barez, Gabriel Stanovsky, Yonatan Belinkov

Unveiling the Spectrum of Data Contamination in Language Model: A Survey from Detection to Remediation (2024.findings-acl)

Challenge:	Prior work on large language model (LLM) hallucinations associated with model uncertainty or inaccurate knowledge.
Approach:	They define and investigate a type of hallucination where a model can answer a question correctly but a perturbation causes it to produce a hallucinous response with high certainty.
Outcome:	The proposed mitigations outperform existing methods on CHOKE hallucinations . the findings highlight the need to understand their origins and improve mitigation strategies .

Copied to clipboard

Chunyuan Deng, Yilun Zhao, Yuzhao Heng, Yitong Li, Jiannan Cao, Xiangru Tang, Arman Cohan

Mitigating Hallucinations in LM-Based TTS Models via Distribution Alignment Using GFlowNets (2025.emnlp-main)

Challenge:	Data contamination is a problem in Large language models due to the reliance on extensive internet-derived training corpora.
Approach:	They present a survey on the topic of data contamination in large language models.
Outcome:	The results of the first survey on data contamination in large language models provide a comprehensive guide for NLP researchers seeking a systematic understanding of the issue.

Copied to clipboard

Chenlin Liu, Minghui Fang, Patrick Zhang, Wei Zhou, Jie Gao, Jiqing Han

BadScientist: Can a Research Agent Write Convincing but Unsound Papers that Fool LLM Reviewers? (2026.acl-long)

Challenge:	Existing mitigation strategies for Text-to-Speech systems require excessive training resources or inference latency.
Approach:	They propose a GFlOwNet-guided distribution AlignmenT framework that mitigates hallucinations without relying on massive resources or inference latency.
Outcome:	The proposed framework reduces over 50% character error rates and lowers uncertainty by up to 58% on challenging test cases.

Copied to clipboard

Fengqing Jiang, Yichen Feng, Yuetai Li, Luyao Niu, Basel Alomair, Radha Poovendran

Challenge:	Existing evidence suggests that LLMs are not able to detect scientifically unsound work from malicious or poorly designed research agents.
Approach:	They develop a framework that evaluates whether fabrication-oriented paper generation agents can deceive multi-model LLM review systems.
Outcome:	The proposed framework shows that fabricated papers achieve acceptance rates up to 18% . the framework shows only marginal improvements, with detection accuracy barely exceeding random chance.

ExtremeAIGC: Benchmarking LMM Vulnerability to AI-Generated Extremist Content (2025.findings-emnlp)

Copied to clipboard

Bhavik Chandna, Mariam Aboujenane, Usman Naseem

Robustness in NLP

Challenge:	Existing datasets for evaluating LMM robustness lack exploration of extremist content . existing models lack diverse image generation models and comprehensive coverage of historical events .
Approach:	They propose a benchmark dataset to assess LMM models against extremist content . ExtremeAIGC simulates real-world events and malicious use cases .
Outcome:	a new benchmark dataset and evaluation framework assesses LMM models against extremist content.

Mixed Signals: Decoding VLMs’ Reasoning and Underlying Bias in Vision-Language Conflict (2025.findings-emnlp)

Copied to clipboard

Pouya Pezeshkpour, Moin Aminnaseri, Estevam Hruschka

Principled Personas: Defining and Measuring the Intended Effects of Persona Prompting on Task Performance (2025.emnlp-main)

Challenge:	Vision-language models have demonstrated impressive performance by effectively integrating visual and textual information to solve complex tasks.
Approach:	They build upon existing benchmarks to create five datasets containing mismatched image-text pairs and examine how they reason over visual and textual data .
Outcome:	The proposed model reasoned over visual and textual data in real-world applications but not in the visual and visual descriptions.

Copied to clipboard

Pedro Henrique Luz de Araujo, Paul Röttger, Dirk Hovy, Benjamin Roth

Why Supervised Fine-Tuning Fails to Learn: A Systematic Study of Incomplete Learning in Large Language Models (2026.acl-long)

Challenge:	Prior work on persona prompting has shown mixed results on its effectiveness . prior work did not consider when and why personas should affect performance .
Approach:	They analyze literature on persona prompting and distill three desiderata for their effectiveness . they propose mitigation strategies to improve robustness but find they only work for the largest, most capable models .
Outcome:	The authors find that expert personas usually lead to positive or non-significant performance changes . they propose mitigation strategies to improve robustness but only for the largest models .

Copied to clipboard

Chao Xue, Yao Wang, Mengqiao Liu, Di Liang, Xingsheng Han, Peiyang Liu, Xianjie Wu, Chenyao Lu, Lei Jiang, Yu Lu, Haibo Shi, Shuang Liang, Minlong Peng, Flora D. Salim

Understanding and Mitigating Bias Inheritance in LLM-based Data Augmentation on Downstream Tasks (2026.acl-long)

Challenge:	Incomplete learning is widespread and heterogeneous in large language models . authors identify five recurrent sources of incomplete learning: missing prerequisite knowledge, conflicts between SFT supervision and pre-training knowledge, internal inconsistencies within SFT data, left-side forgetting during sequential fine-tuning, and insufficient optimization for rare or complex patterns.
Approach:	They propose a diagnostic-first framework that maps incomplete learning to causes . they identify five recurrent sources of incomplete learning: missing prerequisite knowledge, conflicts between supervision and pre-training knowledge, internal inconsistencies, left-side forgetting during sequential fine-tuning, and insufficient optimization for rare or complex patterns.
Outcome:	The proposed framework maps incomplete learning to causes using observable training and inference signals.

Copied to clipboard

Miaomiao Li, Hao Chen, Yang Wang, Tingyuan Zhu, Weijia Zhang, Kaijie Zhu, Kam-Fai Wong, Jindong Wang

Towards Graph-hop Retrieval and Reasoning in Complex Question Answering over Textual Database (2024.lrec-main)

Challenge:	Generating synthetic datasets via large language models (LLMs) has emerged as promising approach to improve LLM performance.
Approach:	They propose three mitigation strategies to mitigate bias inheritance in LLMs by analyzing real and LLM-augmented data.
Outcome:	The proposed methods can work differently on different tasks and biases.

Copied to clipboard

Minjun Zhu, Yixuan Weng, Shizhu He, Kang Liu, Haifeng Liu, Yang jun Jun, Jun Zhao

PRISM: Probing Reasoning, Instruction, and Source Memory in LLM Hallucinations (2026.acl-long)

Challenge:	Existing benchmarks for textual question answering only focus on single-chain or single-hop retrieval . Existing approaches to answer complex questions have limitations .
Approach:	They propose to conduct Graph-Hop, a novel multi-chains and multi-hops retrieval paradigm in complex question answering.
Outcome:	The proposed model provides explicit and fine-grained evidence graphs for complex question to support comprehensive and detailed reasoning.

Copied to clipboard

Yuhe Wu, Guangyu Wang, Yuran Chen, Jiatong Zhang, Yutong Zhang, Yujie Chen, Jiaming Shang, Guang Zhang, Zhuang Liu

Exploring Two-Phase Continual Instruction Fine-tuning for Multilingual Adaptation in Large Language Models (2026.findings-acl)

Challenge:	Existing benchmarks for hallucination evaluation rely on mixed queries and posterior evaluation, which quantifies hallucinosity severity but offers limited insight into where and why they occur.
Approach:	They propose a controlled benchmark that disentangles hallucinations into four dimensions: knowledge missing, knowledge errors, reasoning errors, and instruction-following errors.
Outcome:	The proposed model disentangles hallucinations into four dimensions: knowledge missing, knowledge errors, reasoning errors, and instruction-following errors.

Copied to clipboard

Divyanshu Aggarwal, Sankarshan Damle, Navin Goyal, Satya Lokam, Sunayana Sitaram

Resonant Context Anchoring: Decoupling Attention Routing and Signal Gain at Inference Time (2026.findings-acl)

Challenge:	A key challenge for Large Language Models (LLMs) is improving their Multilingual instruction-following ability over time without deteriorating their ability in languages they already excel at, typically English.
Approach:	They propose a two-phase Continual Fine-tuning setup to improve a model's Multilingual adaptability by comparing an English-only LLM with a multilingual instruction dataset.
Outcome:	The proposed model improves on two-phase Continual Fine-tuning (CFT) setups on a multilingual instruction dataset.

Copied to clipboard

Mingkuan Zhao, Yide Gao, Wentao Hu, Suquan Chen, Tianchen Huang, Zhenhua An, Zetao Chang, Xiayu Sun, Yuheng Min

System-Mediated Attention Imbalances Make Vision-Language Models Say Yes (2026.findings-acl)

Challenge:	Existing mitigation strategies rely on suppressing specific neuron activations or employing computationally expensive contrastive decoding mechanisms, which often result in increased perplexity or significantly elevated inference latency.
Approach:	They propose a lightweight inference-time intervention method grounded in the perspective of residual stream signal dynamics to resolve the signal attenuation of external evidence during its propagation through deep networks.
Outcome:	The proposed method improves contextual faithfulness across multiple factual consistency and strong knowledge-conflict tasks while maintaining the model’s general language understanding capabilities.

Copied to clipboard

Tsan Tsai Chan, Varsha Suresh, Anisha Saha, Michael Hahn, Vera Demberg