论文检索

Ikhyun Cho, Julia Hockenmaier

Sparse autoencoders (SAEs) have emerged as a powerful analytical tool in mechanistic interpretability for large language models (LLMs), with growing success in applications beyond interpretability. Building on this momentum, we present a novel approach that leverages SAEs to enhance the general in-context learning (ICL) performance of LLMs.Specifically, we introduce Feature Detection through Prompt Variation (FDPV), which leverages the SAE’s remarkable ability to capture subtle differences between prompts, enabling efficient feature selection for downstream steering. In addition, we propose a novel steering method tailored to ICL—Selective In-Context Steering (SISTER)—grounded in recent insights from ICL research that LLMs utilize label words as key anchors. Our method yields a 3.5% average performance improvement across diverse text classification tasks and exhibits greater robustness to hyperparameter variations compared to standard steering approaches. Our code is available at https://github.com/ihcho2/SAE-ICL.

Tongtong Liu, Zhaohui Wang, Meiyue Qin, Zenghui Lu, Xudong Chen, Yuekui Yang, Peng Shu

The integration of Large Language Models (LLMs) with retrieval systems has shown promising potential in retrieving documents (docs) or advertisements (ads) for a given query. Existing LLM-based retrieval methods generate numeric or content-based DocIDs to retrieve docs/ads. However, the one-to-few mapping between numeric IDs and docs, along with the time-consuming content extraction, leads to semantic inefficiency and limits the scalability of existing methods on large-scale corpora. In this paper, we propose the **R**eal-time **A**d **RE**trieval (RARE) framework, which leverages LLM-generated text called Commercial Intentions (CIs) as an intermediate semantic representation to directly retrieve ads for queries in real-time. These CIs are generated by a customized LLM injected with commercial knowledge, enhancing its domain relevance. Each CI corresponds to multiple ads, yielding a lightweight and scalable set of CIs. RARE has been implemented in a real-world online system, handling daily search volumes in billions. The online implementation has yielded significant benefits: a 5.04% increase in consumption, a 6.37% rise in Gross Merchandise Volume (GMV), a 1.28% enhancement in click-through rate (CTR) and a 5.29% increase in shallow conversions. Extensive offline experiments show RARE’s superiority over ten competitive baselines in four major categories.

Ridwan Mahbub, Mohammed Saidul Islam, Mir Tafseer Nayeem, Md Tahmid Rahman Laskar, Mizanur Rahman, Shafiq Joty, Enamul Hoque

Charts are very common for exploring dataand communicating insights, but extracting key takeaways from charts and articulating them in natural language can be challenging. The chart-to-text task aims to automate this process by generating textual summaries of charts. While with the rapid advancement of large Vision-Language Models (VLMs), we have witnessed great progress in this domain, little to no attention has been given to potential biases in their outputs. This paper investigates how VLMs can amplify geo-economic biases when generating chart summaries, potentially causing societal harm. Specifically, we conduct a large-scale evaluation of geo-economic biases in VLM-generated chart summaries across 6,000 chart-country pairs from six widely used proprietary and open-source models to understand how a country’s economic status influences the sentiment of generated summaries. Our analysis reveals that existing VLMs tend to produce more positive descriptions for high-income countries compared to middle- or low-income countries, even when country attribution is the only variable changed. We also find that models such as GPT-4o-mini, Gemini-1.5-Flash, and Phi-3.5 exhibit varying degrees of bias. We further explore inference-time prompt-based debiasing techniques using positive distractors but find them only partially effective, underscoring the complexity of the issue and the need for more robust debiasing strategies. Our code and dataset are available at <redacted>.

Younwoo Choi, Changling Li, Yongjin Yang, Zhijing Jin

As large language models (LLMs) are increasingly integrated into multi-agent and human-AI systems, understanding their awareness of both self-context and conversational partners is essential for ensuring reliable performance and robust safety. While prior work has extensively studied situational awareness which refers to an LLM’s ability to recognize its operating phase and constraints, it has largely overlooked the complementary capacity to identify and adapt to the identity and characteristics of a dialogue partner. In this paper, we formalize this latter capability as interlocutor awareness and present the first systematic evaluation of its emergence in contemporary LLMs. We examine interlocutor inference across three dimensions—reasoning patterns, linguistic style, and alignment preferences—and show that LLMs reliably identify same-family peers and certain prominent model families, such as GPT and Claude. To demonstrate its practical significance, we develop three case studies in which interlocutor awareness both enhances multi-LLM collaboration through prompt adaptation and introduces new alignment and safety vulnerabilities, including reward-hacking behaviors and increased jailbreak susceptibility. Our findings highlight the dual promise and peril of identity—sensitive behavior in LLMs, underscoring the need for further understanding of interlocutor awareness and new safeguards in multi-agent deployments.

Hanqi Duan, Yao Cheng, Jianxiang Yu, Yao Liu, Xiang Li

Graph Neural Networks (GNNs) have emerged as powerful models for learning from graph-structured data. However, GNNs lack the inherent semantic understanding capability of rich textual node attributes, limiting their effectiveness in applications. On the other hand, we empirically observe that for existing GNN models, no one can consistently outperforms others across diverse datasets. In this paper, we study whether LLMs can act as an ensembler for multi-GNNs and propose the LensGNN model. The model first aligns multiple GNNs, mapping the representations of different GNNs into the same space. Then, through LoRA fine-tuning, it aligns the space between the GNN and the LLM, injecting graph tokens and textual information into LLMs. This allows LensGNN to ensemble multiple GNNs and take advantage of the strengths of LLM, leading to a deeper understanding of both textual semantic information and graph structural information. The experimental results show that LensGNN outperforms existing models. This research advances text-attributed graph ensemble learning by providing a robust and superior solution for integrating semantic and structural information. We provide our code and data here: https://github.com/AquariusAQ/LensGNN.

Yejin Lee, Joonghyuk Hahn, Hyeseon Ahn, Yo-Sub Han

Implicit hate speech detection is challenging due to its subtlety and reliance on contextual interpretation rather than explicit offensive words. Current approaches rely on contrastive learning, which are shown to be effective on distinguishing hate and non-hate sentences. Humans, however, detect implicit hate speech by first identifying specific targets within the text and subsequently interpreting how these target relate to their surrounding context. Motivated by this reasoning process, we propose AmpleHate, a novel approach designed to mirror human inference for implicit hate detection. AmpleHate identifies explicit target using a pretrained Named Entity Recognition model and capture implicit target information via [CLS] tokens. It computes attention-based relationships between explicit, implicit targets and sentence context and then, directly injects these relational vectors into the final sentence representation. This amplifies the critical signals of target-context relations for determining implicit hate. Experiments demonstrate that AmpleHate achieves state-of-the-art performance, outperforming contrastive learning baselines by an average of 82.14% and achieve faster convergence. Qualitative analyses further reveal that attention patterns produced by AmpleHate closely align with human judgement, underscoring its interpretability and robustness.

Ruohao Guo, Wei Xu, Alan Ritter

As Large Language Models (LLMs) are widely deployed in diverse scenarios, the extent to which they could tacitly spread misinformation emerges as a critical safety concern. Current research primarily evaluates LLMs on explicit false statements, overlooking how misinformation often manifests subtly as unchallenged premises in real-world interactions. We curated EchoMist, the first comprehensive benchmark for implicit misinformation, where false assumptions are embedded in the query to LLMs. EchoMist targets circulated, harmful, and ever-evolving implicit misinformation from diverse sources, including realistic human-AI conversations and social media interactions. Through extensive empirical studies on 15 state-of-the-art LLMs, we find that current models perform alarmingly poorly on this task, often failing to detect false premises and generating counterfactual explanations. We also investigate two mitigation methods, i.e., Self-Alert and RAG, to enhance LLMs’ capability to counter implicit misinformation. Our findings indicate that EchoMist remains a persistent challenge and underscore the critical need to safeguard against the risk of implicit misinformation.

Xin Su, Phillip Howard, Steven Bethard

Temporal information extraction (IE) aims to extract structured temporal information from unstructured text, thereby uncovering the implicit timelines within. This technique is applied across domains such as healthcare, newswire, and intelligence analysis, aiding models in these areas to perform temporal reasoning and enabling human users to grasp the temporal structure of text. Transformer-based pre-trained language models have produced revolutionary advancements in natural language processing, demonstrating exceptional performance across a multitude of tasks. Despite the achievements garnered by Transformer-based approaches in temporal IE, there is a lack of comprehensive reviews on these endeavors. In this paper, we aim to bridge this gap by systematically summarizing and analyzing the body of work on temporal IE using Transformers while highlighting potential future research directions.

Keonwoo Roh, Yeong-Joon Ju, Seong-Whan Lee

Large Language Models (LLMs) have shown significant progress in Open-domain question answering (ODQA), yet most evaluations focus on English and assume locale-invariant answers across languages. This assumption neglects the cultural and regional variations that affect question understanding and answer, leading to biased evaluation in multilingual benchmarks. To address these limitations, we introduce XLQA, a novel benchmark explicitly designed for locale-sensitive multilingual ODQA. XLQA contains 3,000 English seed questions expanded to eight languages, with careful filtering for semantic consistency and human-verified annotations distinguishing locale-invariant and locale-sensitive cases. Our evaluation of five state-of-the-art multilingual LLMs reveals notable failures on locale-sensitive questions, exposing gaps between English and other languages due to a lack of locale-grounding knowledge. We provide a systematic framework and scalable methodology for assessing multilingual QA under diverse cultural contexts, offering a critical resource to advance real-world applicability of multilingual ODQA systems. Our findings suggest that disparities in training data distribution contribute to differences in both linguistic competence and locale-awareness across models.

Zihan Liang, Ziwen Pan, Ruoxuan Xiong

Clinical notes contain rich patient information, such as diagnoses or medications, making them valuable for patient representation learning. Recent advances in large language models have further improved the ability to extract meaningful representations from clinical texts. However, clinical notes are often missing. For example, in our analysis of the MIMIC-IV dataset, 24.5% of patients have no available discharge summaries. In such cases, representations can be learned from other modalities such as structured data, chest X-rays, or radiology reports. Yet the availability of these modalities is influenced by clinical decision-making and varies across patients, resulting in modality missing-not-at-random (MMNAR) patterns. We propose a causal representation learning framework that leverages observed data and informative missingness in multimodal clinical records. It consists of: (1) an MMNAR-aware modality fusion component that integrates structured data, imaging, and text while conditioning on missingness patterns to capture patient health and clinician-driven assignment; (2) a modality reconstruction component with contrastive learning to ensure semantic sufficiency in representation learning; and (3) a multitask outcome prediction model with a rectifier that corrects for residual bias from specific modality observation patterns. Comprehensive evaluations across MIMIC-IV and eICU show consistent gains over the strongest baselines, achieving up to 13.8% improvement for hospital readmission and 13.1% for ICU admission (AUC, relative to best baseline).

Xiaohan Ding, Kaike Ping, Buse Çarık, Eugenia Rho

Understanding causal language in informal discourse is a core yet underexplored challenge in NLP. Existing datasets largely focus on explicit causality in structured text, providing limited support for detecting implicit causal expressions, particularly those found in informal, user-generated social media posts. We introduce CausalTalk, a multi-level dataset of five years of Reddit posts (2020–2024) discussing public health related to the COVID-19 pandemic, among which 10,120 posts are annotated across four causal tasks: (1) binary causal classification, (2) explicit vs. implicit causality, (3) cause–effect span extraction, and (4) causal gist generation. Annotations comprise both gold-standard labels created by domain experts and silver-standard labels generated by GPT-4o and verified by human annotators.CausalTalk bridges fine-grained causal detection and gist-based reasoning over informal text. It enables benchmarking across both discriminative and generative models, and provides a rich resource for studying causal reasoning in social media contexts.

Yu-Ang Lee, Guan-Ting Yi, Mei-Yi Liu, Jui-Chao Lu, Guan-Bo Yang, Yun-Nung Chen

Recent advancements in large language models (LLMs) and AI systems have led to a paradigm shift in the design and optimization of complex AI workflows. By integrating multiple components, compound AI systems have become increasingly adept at performing sophisticated tasks. However, as these systems grow in complexity, new challenges arise in optimizing not only individual components but also their interactions. While traditional optimization methods such as supervised fine-tuning (SFT) and reinforcement learning (RL) remain foundational, the rise of natural language feedback introduces promising new approaches, especially for optimizing non-differentiable systems. This paper provides a systematic review of recent progress in optimizing compound AI systems, encompassing both numerical and language-based techniques. We formalize the notion of compound AI system optimization, classify existing methods along several key dimensions, and highlight open research challenges and future directions in this rapidly evolving field.

Advaith Malladi, Rakesh R Menon, Yuvraj Jain, Shashank Srivastava

With the growing adoption of machine learning models in critical domains, techniques for explaining differences between models have become essential for trust, debugging, and informed deployment. Previous approaches address this by identifying input transformations that cause divergent predictions or by learning joint surrogate models to align and contrast behaviors. These methods often require access to training data and do not produce natural language explanations. In this paper, we introduce SLED, a framework that generates faithful natural language explanations of when and how two ML models converge or diverge in their predictions. SLED first uses gradient-based optimization to synthesize input samples that highlight divergence and convergence patterns, and then leverages a large language model (LLM) to generate explanations grounded in these synthetic samples. Across both text-based (3 tasks, 7 models) and structured (10 tasks, 4 models) classification tasks, we show that SLED explanations are 18–24% more faithful than the strongest baselines. User studies also indicate that SLED explanations achieve a real-world simulatability of 63.5%. Importantly, SLED requires minimal access to training data and generalizes well to real-world samples, enabling transparent and data-efficient model comparison.

Junhyuk Choi, Ro-hoon Oh, Jihwan Seol, Bugeun Kim

We introduce VoiceBBQ, a spoken extension of the BBQ (Bias Benchmark for Question answering) - a dataset that measures social bias by presenting ambiguous or disambiguated contexts followed by questions that may elicit stereotypical responses. Due to the nature of speech modality, social bias in Spoken Language Models (SLMs) can emerge from two distinct sources: 1) content aspect and 2) acoustic aspect. The dataset converts every BBQ context into controlled voice conditions, enabling per-axis accuracy, bias, and consistency scores that remain comparable to the original text benchmark. Using VoiceBBQ, we evaluate two SLMs—LLaMA-Omni and Qwen2-Audio—and observe architectural contrasts: LLaMA-Omni retains strong acoustic sensitivity, amplifying gender and accent bias, whereas Qwen2-Audio substantially dampens these cues while preserving content fidelity. VoiceBBQ thus provides a compact, drop-in testbed for jointly diagnosing content and acoustic bias across spoken language models.

Özge Alacam, Sanne Hoeken, Andreas Säuberli, Hannes Gröner, Diego Frassinelli, Sina Zarrieß, Barbara Plank

Variation is inherent in opinion-based annotation tasks like sentiment or hate speech analysis. It does not only arise from errors, fatigue, or sentence ambiguity but also from genuine differences in opinion shaped by background, experience, and culture. In this paper, first, we show how annotators’ confidence ratings can be great use for disentangling subjective variation from uncertainty, without relying on specific features present in the data (text, gaze, etc.). Our goal is to establish distinctive dimensions of variation which are often not clearly separated in existing work on modeling annotator variation. We illustrate our approach through a hate speech detection task, demonstrating that models are affected differently by instances of uncertainty and subjectivity. In addition, we show that human gaze patterns offer valuable indicators of subjective evaluation and uncertainty. Disclaimer: This paper contains sentences that may be offensive.

Jeongwoo Na, Jun Kwon, Eunseong Choi, Jongwuk Lee

Recent advances in large language models (LLMs) have shown impressive performance in passage reranking tasks. Despite their success, LLM-based methods still face challenges in efficiency and sensitivity to external biases. (1) Existing models rely mostly on autoregressive generation and sliding window strategies to rank passages, which incur heavy computational overhead as the number of passages increases. (2) External biases, such as position or selection bias, hinder the model’s ability to accurately represent passages and increase input-order sensitivity. To address these limitations, we introduce a novel passage reranking model, called Multi-View-guided Passage Reranking (MVP). MVP is a non-generative LLM-based reranking method that encodes query-passage information into diverse view embeddings without being influenced by external biases. For each view, it combines query-aware passage embeddings to produce a distinct anchor vector, which is then used to directly compute relevance scores in a single decoding step. In addition, it employs an orthogonal loss to make the views more distinctive. Extensive experiments demonstrate that MVP, with just 220M parameters, matches the performance of much larger 7B-scale fine-tuned models while achieving a 100x reduction in inference latency. Notably, the 3B-parameter variant of MVP achieves state-of-the-art performance on both in-domain and out-of-domain benchmarks. The source code is available at: https://github.com/bulbna/MVP.

Zhuo Chen, Xinyu Wang, Yong Jiang, Zhen Zhang, Xinyu Geng, Pengjun Xie, Fei Huang, Kewei Tu

Despite the advancements made in Vision Large Language Models (VLLMs), like text Large Language Models (LLMs), they have limitations in addressing questions that require real-time information or are knowledge-intensive. Indiscriminately adopting Retrieval Augmented Generation (RAG) techniques is an effective yet expensive way to enable models to answer queries beyond their knowledge scopes. To mitigate the dependence on retrieval and simultaneously maintain, or even improve, the performance benefits provided by retrieval, we propose a method to detect the knowledge boundary of VLLMs, allowing for more efficient use of techniques like RAG. Specifically, we propose a method with two variants that fine-tune a VLLM on an automatically constructed dataset for boundary identification. Experimental results on various types of Visual Question Answering datasets show that our method successfully depicts a VLLM’s knowledge boundary, based on which we are able to reduce indiscriminate retrieval while maintaining or improving the performance. In addition, we show that the knowledge boundary identified by our method for one VLLM can be used as a surrogate boundary for other VLLMs. Code will be released at https://github.com/Chord-Chen-30/VLLM-KnowledgeBoundary

Shashank Srivastava

Large language models are reshaping the norms of human communication, sometimes decou- pling words from genuine human thought. This transformation is deep, and undermines norms historically tied to authorship of text. We draw from linguistic philosophy and AI ethics to detail how large-scale text genera- tion can induce semantic drift, erode account- ability, and obfuscate intent and authorship. Our work here introduces hybrid authorship graphs (modeling humans, LLMs, and texts in a provenance network), epistemic doppel- gängers (LLM-generated texts that are indis- tinguishable from human-authored texts), and authorship entropy. We explore mechanisms such as “proof-of-interaction” authorship veri- fication and educational reforms to restore con- fidence in language. LLMs’ benefits (broader access, increased fluency, automation, etc.) are undeniable, but the upheavals they introduce to the linguistic landscape demand reckoning.

Manato Tajiri, Michimasa Inaba

Conversational Recommender Systems (CRSs) aim to elicit user preferences via natural dialogue to provide suitable item recommendations. However, current CRSs often deviate from realistic human interactions by rapidly recommending items in brief sessions. This work addresses this gap by leveraging Large Language Models (LLMs) to generate dialogue summaries from dialogue history and item recommendation information from item description. This approach enables the extraction of both explicit user statements and implicit preferences inferred from the dialogue context. We introduce a method using Direct Preference Optimization (DPO) to ensure dialogue summary and item recommendation information are rich in information crucial for effective recommendations. Experiments on two public datasets validate our method’s effectiveness in fostering more natural and realistic conversational recommendation processes. Our implementation is publicly available at: https://github.com/UEC-InabaLab/Refining-LLM-Text

Yevheniia Kryklyvets, Mohammed Irfan Kurpath, Sahal Shaji Mullappilly, Jinxing Zhou, Fahad Shahbaz Khan, Rao Muhammad Anwer, Salman Khan, Hisham Cholakkal

Fine-grained understanding and species-specific, multimodal question answering are vital for advancing biodiversity conservation and ecological monitoring. However, existing multimodal large language models (MM-LLMs) face challenges when it comes to specialized topics like avian species, making it harder to provide accurate and contextually relevant information in these areas. To address this limitation, we introduce the **MAviS-Dataset**, a large-scale multimodal avian species dataset that integrates image, audio, and text modalities for over 1,000 bird species, comprising both pretraining and instruction-tuning subsets enriched with structured question–answer pairs. Building on the MAviS-Dataset, we introduce **MAviS-Chat**, a multimodal LLM that supports audio, vision, and text designed for fine-grained species understanding, multimodal question answering, and scene-specific description generation. Finally, for quantitative evaluation, we present **MAviS-Bench**, a benchmark of over 25,000 Q&A pairs designed to assess avian species-specific perceptual and reasoning abilities across modalities. Experimental results show that MAviS-Chat outperforms the baseline MiniCPM-o-2.6 by a large margin, achieving state-of-the-art open-source results and demonstrating the effectiveness of our instruction-tuned MAviS-Dataset. Our findings highlight the necessity of domain-adaptive MM-LLMs for ecological applications. Our code, training data, evaluation benchmark, and models are available at https://github.com/yevheniia-uv/MAviS.