Quality Estimation (QE) aims to assess machine translation quality without reference translations, but recent studies have shown that existing QE models exhibit systematic gender bias. In particular, they tend to favor masculine realizations in gender-ambiguous contexts and may assign higher scores to gender-misaligned translations even when gender is explicitly specified. To address these issues, we propose FairQE, a multi-agent-based, fairness-aware QE framework that mitigates gender bias in both gender-ambiguous and gender-explicit scenarios. FairQE detects gender cues, generates gender-flipped translation variants, and combines conventional QE scores with LLM-based unbiased reasoning through a dynamic bias-aware aggregation mechanism. This design preserves the strengths of existing QE models while calibrating their gender-related biases in a plug-and-play manner. Extensive experiments across multiple gender bias evaluation settings demonstrate that FairQE consistently improves gender fairness over strong QE baselines. Moreover, under MQM-based meta-evaluation following the WMT 2023 Metrics Shared Task, FairQE achieves competitive or improved general QE performance. These results show that gender bias in QE can be effectively mitigated without sacrificing evaluation accuracy, enabling fairer and more reliable translation evaluation.
论文检索
输入标题、作者或关键词，从 2,854 篇学术成果中精准定位
Confident, Calibrated, or Complicit: Safety Alignment and Ideological Bias in LLM Hate Speech Detection
PDF ↗We investigate the efficacy of Large Language Models (LLMs) in detecting implicit and explicit hate speech, examining how models with minimal safety alignment (uncensored) compare with more heavily aligned (censored) counterparts in a deployed-model setting when deployed using political personas. While uncensored models are often framed as offering a less constrained perspective, our results reveal a trade-off: censored models outperform their uncensored counterparts in both accuracy and robustness, achieving 69.0% versus 64.1% strict accuracy. However, this higher performance is also associated with greater resistance to persona-based influence, while uncensored models are more malleable to ideological framing. Furthermore, we identify critical failures across all models in understanding nuanced language such as irony. We also find alarming fairness disparities in performance across different targeted groups and systemic overconfidence that renders self-reported certainty unreliable. These findings challenge the notion of LLMs as objective arbiters and highlight the need for more sophisticated auditing frameworks that account for fairness, calibration, and ideological consistency. Taken together, these results point to censorship-as-deployed rather than safety alignment in isolation as the more appropriate frame for interpreting model differences.
Language identification (LID) is a fundamental step in curating multilingual corpora. However, LID models still perform poorly for many languages, especially on the noisy and heterogeneous web data often used to train multilingual language models. In this paper, we introduce CommonLID, a community-driven, human-annotated LID benchmark for the web domain, covering 109 languages. Many of the included languages have been previously under-served, making CommonLID a key resource for developing more representative high-quality text corpora. We show CommonLID’s value by using it, alongside five other common evaluation sets, to test eight popular LID models. We analyse our results to situate our contribution and to provide an overview of the state of the art. In particular, we highlight that existing evaluations overestimate LID accuracy for many languages in the web domain. We make CommonLID and the code used to create it available under an open, permissive license.
Ensuring fairness in social survey simulation is critical, as biased outputs can misrepresent underrepresented groups. This issue is growing as large language models (LLMs) are increasingly used for this task. However, standard fine-tuning based on Empirical Risk Minimization (ERM) often under-optimizes minority groups, causing substantial subgroup disparities. Distributionally robust Optimization (DRO) methods reduce worst-case errors, but their strict worst-case selection can lead to noisy and unstable optimization under demographic sparsity. These issues create intertwined challenges for fairness, convergence and stability. We propose SAFO, a dynamic utility–fairness optimization framework for LLM-based survey simulation that explicitly targets both fairness and training stability. SAFO combines (i) an Optimizer that preserves mean-loss utility, (ii) an Adversary that performs temperature-controlled, EMA-smoothed and loss-driven group reweighting, and (iii) a Nash-inspired Regulator that adaptively adjusts the utility–fairness trade-off by tracking weak-group gains and collateral utility damages. Experiments on three large-scale survey datasets from China, the U.S., and Europe show that SAFO consistently improves minority performance and social-welfare metrics. It reduces worst-group gaps by up to 12.7%, maintains overall accuracy with a mean change of less than 0.3% and lowers variance across random seeds. Our code is available at https://github.com/PiLab-ZJU/SAFO.
Recent advances in large language models (LLMs) and text-aware graph learning have increased interest in reasoning over text-attributed graphs(TAGs). In many real-world settings, such graphs are inherently heterogeneous, with most existing benchmarks remaining largely homogeneous in structure. As a result, the lack of large-scale benchmarks for heterogeneous text-attributed graphs has hindered systematic evaluation and fair comparison of existing methods. In this work, we introduce CITE - **C**atalytic **I**nformation **T**extual **E**ntities Graph, the first and largest heterogeneous text-attributed citation graph benchmark for catalytic materials. CITE contains over 438K nodes and 1.2M edges spanning four node types and four relation types, with rich node-level textual information. We establish standardized evaluation protocols for node classification and link prediction, and conduct ablation studies to assess the impact of graph heterogeneity and textual attributes. Using CITE, we benchmark four classes of learning paradigms, including homogeneous graph models, heterogeneous graph models, LLM-centric models, and hybrid LLM–graph models. By providing a large-scale heterogeneous text-attributed benchmark together with standardized evaluation protocols and comprehensive baselines, CITE enables systematic assessment across diverse modeling paradigms and offers new insights into text-aware and LLM-enhanced graph learning. The dataset, codebase and evaluation suite are publicly available.
Understanding and Mitigating Bias Inheritance in LLM-based Data Augmentation on Downstream Tasks
PDF ↗Generating synthetic datasets via large language models (LLMs) has emerged as a promising approach to improve LLM performance.However, LLMs inherently reflect biases in their training data, leading to a critical challenge: when models are trained on synthetic data, they may propagate and amplify the inherent biases that can significantly impact fairness and robustness on downstream tasks—a phenomenon we term bias inheritance. This work presents the first systematic investigation in understanding, analyzing, and mitigating bias inheritance. We fine-tune LLMs with a combined dataset of real and LLM-augmented data with varied bias ratio as the proportion of augmented data. Through systematic experiments across 10 classification and generation tasks, we analyze how 6 different types of biases manifest. Our results indicate that bias inheritance harms downstream task performance in bias directly-related classification and generation tasks. Then, our analysis identifies three key misalignment factors: misalignment of values, group data, and data distributions. Based on these insights, we propose three mitigation strategies: token-based, mask-based, and loss-based approaches, which can work differently on various tasks and bias, indicating the substantial challenges to mitigate bias inheritance. We hope this work can provide insights to the research of LLM data augmentation.
Human infants, with only a few hundred hours of speech exposure, acquire basic units of new languages, highlighting a striking efficiency gap compared to the data-hungry self-supervised speech models. To address this gap, this paper introduces SpidR-Adapt for rapid adaptation of speech units to new languages using minimal unlabeled data. We cast such low-resource speech representation learning as a meta-learning problem and construct a multi-task adaptive pre-training (MAdaPT) protocol which formulates the adaptation process as a bi-level optimization framework. To enable scalable meta-training under this framework, we propose a novel heuristic solution, first-order bi-level optimization (FOBLO), avoiding heavy computation costs. Finally, we stabilize meta-training by using a robust initialization through interleaved supervision which alternates self-supervised and supervised objectives. Empirically, SpidR-Adapt achieves rapid gains in phonemic discriminability (ABX) and downstream spoken language modeling scores (sWUGGY, sBLIMP, tSC), surpassing in-domain toplines after training on less than 1h of target-language audio and delivering 100\times greater data efficiency than standard multi-task training.. These findings highlight a practical, architecture-agnostic path toward biologically inspired, data-efficient representations. We open-source the training code and model checkpoints at https://github.com/facebookresearch/spidr-adapt.
Vision language models (VLMs) demonstrate strong zero-shot performance, but often perpetuate social stereotypes in person-centric queries, yielding skewed demographic distributions. Current debiasing methods apply uniform bias corrections across all input queries regardless of their bias sensitivity, creating a fundamental fairness–utility trade-off. Strong debiasing distorts semantically meaningful information in bias-insensitive queries, while weak debiasing fails to mitigate stereotypes in bias-sensitive ones. This one-size-fits-all approach hampers simultaneously achieving high utility on bias-insensitive queries and fairness on bias-sensitive queries. We introduce Reward-Gated Test-Time Adaptation (RG-TTA), a reinforcement learning-based test-time adaptation framework that selectively applies debiasing based on input sensitivity. RG-TTA adaptively triggers fairness regularization based on the bias sensitivity of each input during test-time policy adaptation, while focusing exclusively on optimizing cross-modal alignment for bias-insensitive inputs. Experiments on fairness benchmarks (e.g., FairFace, UTKFace) demonstrate substantial bias reduction while simultaneously improving zero-shot utility, resolving the trade-off of uniform debiasing.
Fair-CCD: Mitigating Bias in Large Language Models for Tabular Classification Through Context-Contrastive Decoding
PDF ↗While recent studies show the effectiveness of in-context learning (ICL) for tabular data prediction, they also reveal significant fairness issues in large language models (LLMs). Prior work to mitigate fairness issues often employs interventions relying on subjective demonstration selection. Its effectiveness varies significantly with the specific demonstration content, leading to low controllability. Moreover, the improvement of fairness is highly unstable across different models and tasks. To address the challenges of low controllability and limited stability in fairness interventions, we propose Fairness-Aware Context-Contrastive Decoding (Fair-CCD). Fair-CCD first constructs Structural Bias Templates (SBTs), motivated by behavioral patterns observed in demonstrations, to encode the relationship between sensitive attributes and predicted labels in a structured and controllable form. During inference, Fair-CCD injects multiple SBTs and contrasts the model’s responses, generating two differential signals that guide fairness adjustment and preserve task performance. By leveraging attention signals to scale decoding adjustments guided by the difference signals, Fair-CCD achieves stable and adaptive bias mitigation across models and tasks. Extensive experimental results demonstrate that Fair-CCD consistently improves fairness metrics without degrading task accuracy.
Large Language Models (LLMs) behave non-deterministically, and prompting has become a common method for steering their outputs.A popular strategy is to assign a persona to the model to produce more varied, context-sensitive responses, similar to how responses vary across human individuals.Against the expectation that persona prompting yields a wide range of opinions, our experiments show that LLMs keep consistent value orientations.We observe a persistent inertia in their responses, where certain moral and value dimensions (especially harm avoidance and fairness) stay skewed in one direction across persona settings.To study this, we use role-play at scale, which pairs randomized persona prompts with a macro-level analysis of model outputs.Our results point to strong internal biases and value preferences in LLMs, which we call value orientation and inertia. These models warrant scrutiny and adjustment before use in applications where balanced outputs matter.
Stable-RAG: Mitigating Retrieval-Permutation-Induced Hallucinations in Retrieval-Augmented Generation
PDF ↗Retrieval-Augmented Generation (RAG) has become a key paradigm for reducing factual hallucinations in Large Language Models (LLMs), yet little is known about how the order of retrieved documents affects model behavior. We empirically show that under a Top-5 retrieval setting with the gold document included, LLM answers vary substantially across permutations of the retrieved set, even when the gold document is fixed in the first position. This reveals a previously underexplored sensitivity to retrieval permutations. Although existing robust RAG methods focus primarily on enhancing LLM robustness to low-quality retrieval and mitigating positional bias to distribute attention fairly over long contexts, neither approach directly addresses permutation sensitivity. In this paper, we propose Stable-RAG, which exploits permutation sensitivity estimation to mitigate permutation-induced hallucinations. Stable-RAG runs the generator under multiple retrieval orders, clusters hidden states, and decodes from a cluster-center representation that captures the dominant reasoning pattern. It then uses these reasoning results to align hallucinated outputs toward the correct answer, encouraging the model to produce consistent and accurate predictions across document permutations. Experiments on three QA datasets show that Stable-RAG improves answer accuracy, reasoning consistency, and generalization across datasets, retrievers, and input lengths compared with strong baselines.
Unlocking Multilingual Reasoning Capability of LLMs and LVLMs through Representation Engineering
PDF ↗Large Language Models (LLMs) and Large Vision-Language Models (LVLMs) demonstrate strong reasoning capabilities, yet their performance in English significantly outperforms that in low-resource languages, raising fairness concerns in multilingual applications. Existing approaches either rely on costly multilingual training or employ prompting with external translation tools, both of which are resource-intensive and sensitive to translation quality. To address these limitations, we propose a training-free inference-time method to enhance Multilingual Reasoning capabilities via Representation Engineering (MRRE) without using any additional training data or tools. MRRE sequentially injects two precomputed vectors at specific layers during inference processing: cross-lingual reasoning enhancement vectors, which steer non-English reasoning representations toward English space to unlock multilingual reasoning, and target-language output anchoring vectors, which restore the distribution of the target language to preserve input–output language consistency. Comprehensive experiments across six advanced LLMs and LVLMs on four reasoning benchmarks demonstrate that MRRE consistently enhances non-English reasoning by an average gain of 5.48% and up to 7.54% in low-resource languages (e.g., Thai and Swahili), while improving input-output language consistency by 3.78%.
Can Persona-Prompted LLMs Emulate Subgroup Values? An Empirical Analysis of Generalisability and Fairness in Cultural Alignment
PDF ↗Despite their global prevalence, many Large Language Models (LLMs) are aligned to a monolithic, often Western-centric set of values. This paper investigates the more challenging task of fine-grained value alignment: examining whether LLMs can emulate the distinct cultural values of demographic subgroups. Using Singapore as a case study and the World Values Survey (WVS), we examine the value landscape and show that even state-of-the-art models like GPT-4.1 achieve only 57.4% accuracy in predicting subgroup modal preferences. We construct a dataset of over 20,000 samples to train and evaluate a range of models. We demonstrate that simple fine-tuning on structured numerical preferences yields substantial gains, improving accuracy on unseen, out-of-distribution subgroups by an average of 17.4%. These gains partially transfer to open-ended generation. However, we find significant pre-existing performance biases, where models better emulate young, male, Chinese, and Christian personas. Furthermore, while fine-tuning improves average performance, it widens the disparity between subgroups when measured by distance-aware metrics. Our work offers insights into the limits and fairness implications of subgroup-level cultural alignment.
SecureVibeBench: Benchmarking Secure Vibe Coding of AI Agents via Reconstructing Vulnerability-Introducing Scenarios
PDF ↗Large language model-powered code agents are rapidly transforming software engineering, yet the security risks of their generated code have become a critical concern. Existing benchmarks have provided valuable insights, but they fail to capture scenarios in which vulnerabilities are actually introduced by human developers, making fair comparisons between humans and agents infeasible. We therefore introduce SecureVibeBench, a benchmark of 105 C/C++ secure coding tasks sourced from 41 projects in OSS-Fuzz for code agents. SecureVibeBench has the following features: (i) realistic task settings that require multi-file edits in large repositories, (ii) aligned contexts based on real-world open-source vulnerabilities with precisely identified vulnerability introduction points, and (iii) comprehensive evaluation that combines functionality testing and security checking with both static and dynamic oracles. We evaluate 5 popular code agents like OpenHands, supported by 5 LLMs (e.g., Claude sonnet 4.5) on SecureVibeBench. Results show that current agents struggle to produce both correct and secure code, as even the best-performing one, produces merely 23.8% correct and secure solutions on SecureVibeBench.
Advancements in Large language models (LLMs) have enabled a variety of downstream applications like story and interview script generation.However, recent research raised concerns about culture-related fairness issues in LLM-generated content.In this work, we identify and systematically investigate LLMs’ **insider-outsider bias**, a phenomenon where models position themselves as "insiders" of mainstream cultures during generation while externalizing less dominant cultures.We propose the ***InsideOut*** benchmark with 4,000 generation prompts and three evaluation metrics to quantify this bias through a *culturally situated interview script generation* task, in which an LLM is positioned as a reporter interviewing local people across 10 diverse cultures.Empirical evaluation on 5 state-of-the-art LLMs reveals that while models adopt insider tones in over 88% US-contexted scripts on average, they disproportionately default to "outsider" stances for non-Western cultures.To mitigate these biases, we propose *2 inference-time methods*: a baseline prompt-based **Fairness Intervention Pillars (FIP)** method, and a structured **Mitigation via Fairness Agents (MFA)** framework consisting of a Single-Agent (MFA-SA), a Hierarchical-Agent (MFA-HA), and an autonomous Agentic Planning (MFA-Plan) pipeline.Empirical results demonstrate that agent-based MFA methods achieve outstanding and robust performance in mitigating the insider-outsider bias:For instance, on the Cultural Alignment Gap (CAG) metric, *MFA-SA reduces bias in Llama model by 89.70 % and MFA-HA mitigates bias in Qwen by 82.54%*.These findings showcase the effectiveness of agent-based methods as a promising direction for mitigating biases in generative LLMs.
The rapid advancement of large language models (LLMs) has not been matched by their evaluation in low-resource languages, especially Southeast Asian languages like Lao. To fill this gap, we introduce LaoBench, the first large-scale, high-quality, and multidimensional benchmark for assessing LLM language understanding and reasoning in Lao. LaoBench contains 17,000+ expert-curated samples across three dimensions: culturally grounded knowledge application, curriculum-aligned K12 education, and bilingual translation among Lao, Chinese, and English. It includes open-source and held-out subsets, where the held-out portion enables secure black-box evaluation via a controlled service to improve fairness and data security. We construct LaoBench with a hybrid pipeline that combines expert authoring with agent-assisted verification, ensuring linguistic accuracy, cultural relevance, and educational validity. We evaluate diverse state-of-the-art open-source and closed-source LLMs, and find that even strong multilingual models lag behind human experts, particularly in culturally grounded reasoning and translation fidelity. We hope LaoBench will catalyze research on Lao and other underrepresented Southeast Asian languages for more inclusive multilingual evaluation.
Deep search agents that combine large language models with retrieval tools excel at complex, multi-hop queries. Yet, existing benchmarks such as BrowseComp rely on black-box web search APIs, facing key limitations. (1) Fairness: for agents, dynamic and opaque web APIs hinder reproducibility and fair comparisons across agents. (2) Disentanglement: for retrieval, the lack of a fixed document corpus makes it impossible to isolate retriever contributions from end-to-end search agent accuracy. We introduce BrowseComp-Plus, a benchmark derived from BrowseComp that employs a fixed, human-verified corpus, enabling controlled retrieval for deep search agents. BrowseComp-Plus clearly distinguishes agent performance: with a BM25 retriever, the open-source Search-R1 achieves 3.86% accuracy, while GPT-5 achieves 55.9%. Additionally, BrowseComp-Plus makes retrieval gains explicit: pairing GPT-5 with Qwen3-Embedding-8B retriever further improves accuracy to 70.1% while reducing search calls. Overall, BrowseComp-Plus provides a fair and disentangled testbed, advancing both deep search agent evaluation and retrieval research for agentic search. Code and data can be found at: https://texttron.github.io/BrowseComp-Plus/
LLMEval-Fair: A Large-Scale Longitudinal Study on Robust and Fair Evaluation of Large Language Models
PDF ↗Existing evaluation of Large Language Models (LLMs) on static benchmarks is vulnerable to data contamination and leaderboard overfitting, critical issues that obscure true model capabilities. To address this, we introduce LLMEval-Fair, a framework for dynamic evaluation of LLMs. LLMEval-Fair is built on a proprietary bank of 220k graduate-level questions, from which it dynamically samples unseen test sets for each evaluation run. Its automated pipeline ensures integrity via contamination-resistant data curation, a novel anti-cheating architecture, and a calibrated LLM-as-a-judge process achieving 90% agreement with human experts, complemented by a relative ranking system for fair comparison. An 30-month longitudinal study of nearly 60 leading models reveals a performance ceiling on knowledge memorization and exposes data contamination vulnerabilities undetectable by static benchmarks. The framework demonstrates exceptional robustness in ranking stability and consistency, providing strong empirical validation for the dynamic evaluation paradigm. LLMEval-Fair offers a robust and credible methodology for assessing the true capabilities of LLMs beyond leaderboard scores, promoting the development of more trustworthy evaluation standards.
When Bigger Isn’t Better: A Comprehensive Fairness Evaluation of Political Bias in Multi-News Summarisation
PDF ↗Multi-document news summarisation systems are increasingly adopted for their convenience in processing vast daily news content, making fairness across diverse political perspectives critical. However, these systems can exhibit political bias through unequal representation of viewpoints, disproportionate emphasis on certain perspectives, and systematic underrepresentation of minority voices. This study presents a comprehensive evaluation of such bias in multi-document news summarisation using FairNews, a dataset of complete news articles with political orientation labels, examining how large language models (LLMs) handle sources with varying political leanings across 13 models and five fairness metrics. We investigate both baseline model performance and effectiveness of various debiasing interventions, including prompt-based and judge-based approaches. Our findings challenge the assumption that larger models yield fairer outputs, as mid-sized variants consistently outperform their larger counterparts, offering the best balance of fairness and efficiency. Prompt-based debiasing proves highly model dependent, while entity sentiment emerges as the most stubborn fairness dimension, resisting all intervention strategies tested. These results demonstrate that fairness in multi-document news summarisation requires multi-dimensional evaluation frameworks and targeted, architecture-aware debiasing rather than simply scaling up.
Despite the recent success of instruction-tuned language models and their ubiquitous usage, very little is known of how models process instructions internally. In this work, we address this gap from a mechanistic point of view by investigating how instruction-specific representations are constructed and utilized in different stages of post-training: Supervised Fine-Tuning (SFT) and Direct Preference Optimization (DPO). Via causal mediation, we identify that instruction representation is fairly localized in models. These representations, which we call Instruction Vectors (IVs), demonstrate a curious juxtaposition of linear separability along with non-linear causal interaction, broadly questioning the scope of the linear representation hypothesis commonplace in mechanistic interpretability. To disentangle the non-linear causal interaction, we propose a novel method to localize information processing in language models that is free from the implicit linear assumptions of patching-based techniques. We find that, conditioned on the task representations formed in the early layers, different information pathways are selected in the later layers to solve that task, i.e., IVs act as circuit selectors.