Panoptic occupancy prediction aims to jointly infer voxel-wise semantics and instance identities within a unified 3D scene representation. Nevertheless, progress in this field remains constrained by the absence of high-quality 3D mesh resources, instance-level annotations, and physically consistent occupancy datasets. Existing benchmarks typically provide incomplete and low-resolution geometry without instance-level annotations, limiting the development of models capable of achieving precise geometric reconstruction, reliable occlusion reasoning, and holistic 3D understanding. To address these challenges, this paper presents an instance-centric benchmark for the 3D panoptic occupancy prediction task. Specifically, we introduce ADMesh, the first unified 3D mesh library tailored for autonomous driving, which integrates over 15K high-quality 3D models with diverse textures and rich semantic annotations. Building upon ADMesh, we further construct CarlaOcc, a large-scale, physically consistent panoptic occupancy dataset generated using the CARLA simulator. This dataset contains over 100K frames with fine-grained, instance-level occupancy ground truth at voxel resolutions as fine as 0.05 m. Furthermore, standardized evaluation metrics are introduced to quantify the quality of existing occupancy datasets. Finally, a systematic benchmark of representative models is established on the proposed dataset, which provides a unified platform for fair comparison and reproducible research in the field of 3D panoptic perception. Code and dataset are available at https://mias.group/CarlaOcc.
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
A Closed-Form Solution for Debiasing Vision-Language Models with Utility Guarantees Across Modalities and Tasks
PDF ↗While Vision-Language Models (VLMs) have achieved remarkable performance across diverse downstream tasks, recent studies have shown that they can inherit social biases from the training data and further propagate them into downstream applications. To address this issue, various debiasing approaches have been proposed, yet most of them aim to improve fairness without having a theoretical guarantee that the utility of the model is preserved. In this paper, we introduce a debiasing method that yields a closed-form solution in the cross-modal space, achieving Pareto-optimal fairness with bounded utility losses. Our method is training-free, requires no annotated data, and can jointly debias both visual and textual modalities across downstream tasks. Extensive experiments show that our method outperforms existing methods in debiasing VLMs across diverse fairness metrics and datasets for both group and intersectional fairness in downstream tasks such as zero-shot image classification, text-to-image retrieval, and text-to-image generation while preserving task performance. Code will be made available upon acceptance.
Recent advances in 3D Gaussian Splatting (3DGS) have focused on accelerating optimization while preserving reconstruction quality. However, many proposed methods entangle implementation-level improvements with fundamental algorithmic modifications or trade performance for fidelity, leading to a fragmented research landscape that complicates fair comparison. In this work, we consolidate and evaluate the most effective and broadly applicable strategies from prior 3DGS research and augment them with several novel optimizations. We further investigate underexplored aspects of the framework, including numerical stability, Gaussian truncation, and gradient approximation. The resulting system, Faster-GS, provides a rigorously optimized algorithm that we evaluate across a comprehensive suite of benchmarks. Our experiments demonstrate that Faster-GS achieves up to 5x faster training while maintaining visual quality, establishing a new cost-effective and resource efficient baseline for 3DGS optimization. Furthermore, we demonstrate that optimizations can be applied to 4D Gaussian reconstruction, leading to efficient non-rigid scene optimization. Source code available at: https://github.com/nerficg-project/faster-gaussian-splatting
Video foundation models aim to integrate video understanding, generation, editing, and instruction following within a single framework, making them a central direction for next-generation multimodal systems. However, existing evaluation benchmarks remain fragmented and limited in scope, as they each target a single task, rely on task-specific metrics, and typically use short or simple video clips. As a result, they do not capture the unified capabilities that these models are designed to deliver. To address this gap, we introduce UniVBench, a benchmark purpose-built for evaluating video foundation models across four core abilities: video understanding, video generation, video editing, and a newly proposed task, video reconstruction, which assesses how faithfully a model can reproduce video content it has encountered. Our benchmark substantially expands the complexity of evaluation by incorporating 200 high-quality, diverse and multi-shot videos, each paired with detailed captions, multi-format editing instructions, and reference images. All videos are entirely human-created and carefully validated, offering richer cinematic information than prior benchmarks. In addition, we develop a unified agentic evaluation system (UniV-Eval) that standardizes prompting, instruction parsing, and scoring across all tasks, enabling fair, scalable, and reproducible comparisons of unified video models. By grounding evaluation in instruction-based multi-shot video tasks, UniVBench provides the first framework for measuring the integrated capabilities that video foundation models aim to achieve. Extensive human annotations ensure our evaluation aligns with human judgment, enabling rigorous assessment and accelerating progress toward robust video intelligence.
QUANTIPHY: A Quantitative Benchmark Evaluating Physical Reasoning Abilities of Vision-Language Models
PDF ↗Understanding the physical world is essential for generalist AI agents. However, it remains unclear whether state-of-the-art vision perception models (e.g., large VLMs) can perform quantitative physical reasoning tasks. Existing evaluations are predominantly VQA-based and qualitative, offering limited insight into whether these models can infer the kinematic quantities of moving objects from videos. To address this, we present QuantiPhy, the first benchmark designed to quantitatively measure a VLM's physical reasoning ability. Comprising more than 3.3K video-text instances with numerical ground truth, QuantiPhy evaluates a VLM's performance on estimating an object's size, velocity, and acceleration at a given timestamp, using one of these properties as an input prior. The benchmark standardizes prompts and scoring to assess numerical accuracy, enabling fair comparisons across models. Our experiments on state-of-the-art VLMs reveal a consistent gap between their qualitative plausibility and actual numerical correctness. We further provide an in-depth analysis of key factors like background noise, counterfactual priors, and strategic prompting and find that state-of-the-art VLMs lean heavily on pre-trained world knowledge rather than faithfully using the provided visual and textual inputs when reasoning about objects' kinematic properties. QuantiPhy offers the first rigorous, scalable testbed to move VLMs beyond mere verbal plausibility toward a numerically grounded physical understanding.
The utilization of temporal information has always been an open topic in the tracking community. However, existing trackers tend to employ more and more inputs or parameters for temporal learning, hindering their deployment in resource-constrained unmanned aerial vehicles (UAVs). More importantly, this raises ambiguity whether the performance gains come from the temporal learning itself, or come from the increased inputs and parameters. In this study, we advocate designing temporal learning components from a more balanced perspective that jointly considers performance gains and computational costs. To achieve this goal, we introduce a new evaluation metric, i.e., precision per FLOP (PPF). The PPF is introduced to quantify the tracking precision gains achieved by temporal learning components per unit of FLOP, thus enabling fair and efficiency-aware comparisons among these components and driving them toward more efficient designs. Based on this metric, we propose a low-cost yet effective temporal learning (LETL) approach to model contextual relationships. This approach continuously propagates and merges representative appearance tokens in video streams, allowing the tracker to efficiently capture the changing patterns of targets with relatively low costs. We integrate the LETL approach into existing one-stream frameworks, thereby building a simple yet effective tracker, namely LETrack, for robust UAV tracking. Extensive experimental demonstrate the superiority of our LETrack, and show that the proposed LETL approach achieves higher PPF scores. The Code is available at https://github.com/GXNU-ZhongLab/LETrack.
PPISP: Physically-Plausible Compensation and Control of Photometric Variations in Radiance Field Reconstruction
PDF ↗Multi-view 3D reconstruction methods remain highly sensitive to photometric inconsistencies arising from camera optical characteristics and variations in image signal processing (ISP). Existing mitigation strategies such as per-frame latent variables or affine color corrections lack physical grounding and generalize poorly to novel views. We propose the Physically-Plausible ISP (PPISP) correction module, which disentangles camera-intrinsic and capture-dependent effects through physically based and interpretable transformations. A dedicated PPISP controller, trained on the input views, predicts ISP parameters for novel viewpoints, analogous to auto exposure and auto white balance in real cameras. This design enables realistic and fair evaluation on novel views without access to ground-truth images. PPISP achieves SotA performance on standard benchmarks, while providing intuitive control and supporting the integration of metadata when available. The source code is available at: https://github.com/nv-tlabs/ppisp
Pretrained vision encoders are widely used as frozen, black-box feature extractors, yet they often inherit spurious correlations that disproportionately harm underrepresented groups. We introduce PLD-Debias, a fully black-box debiasing framework that requires neither access to backbone parameters nor demographic annotations. Our method integrates three components: (1) Rank-Regularized Amplification, a lightweight adapter that exaggerates latent spurious directions; (2) Unsupervised Pseudo-Bias Induction, which clusters amplified features to infer high-fidelity proxy bias labels; and (3) Bias-Guided Refinement, combining supervised contrastive alignment with cluster-aware adaptive margins to purify representations and equalize decision boundaries. We theoretically show that these components jointly tighten a worst-group risk bound under spurious correlations. Empirically, PLD-Debias achieves state-of-the-art worst-group accuracy across CelebA, Waterbirds, and CMNIST, improving performance by 3-5 points over prior black-box methods while maintaining average accuracy. Remarkably, our pseudo-bias labels align with ground-truth bias annotations at over 90% fidelity, enabling oracle-level robustness without demographic supervision. Our results demonstrate that fairness and utility can be achieved through a plug-and-play classifier adapter for any frozen foundation model.
Federated active learning (FAL) seeks to reduce annotation cost under privacy constraints, yet its effectiveness degrades in realistic settings with severe global class imbalance and highly heterogeneous clients. We conduct a systematic study of query-model selection in FAL and uncover a central insight: the model that achieves more class-balanced sampling, especially for minority classes, consistently leads to better final performance. Moreover, global-model querying is beneficial only when the global distribution is highly imbalanced and client data are relatively homogeneous; otherwise, the local model is preferable. Based on these findings, we propose FairFAL, an adaptive class-fair FAL framework. FairFAL (1) infers global imbalance and local-global divergence via lightweight prediction discrepancy, enabling adaptive selection between global and local query models; (2) performs prototype-guided pseudo-labeling using global features to promote class-aware querying; and (3) applies a two-stage uncertainty-diversity balanced sampling strategy with k-center refinement. Experiments on five benchmarks show that FairFAL consistently outperforms state-of-the-art approaches under challenging long-tailed and non-IID settings. The code is available at https://github.com/chenchenzong/FairFAL.
Text-embedding models frequently inherit societal biases, yet the influence of socio-economic markers remains largely unexplored. This paper identifies Currency Bias as a systemic representational limitation in financial AI, where models exhibit associative sensitivity to economic hierarchies. We analyze this through three dimensions: (1) the Syntax Gap, where models fail to align currency names, symbols, and acronyms; (2) Associative Sensitivity, where embeddings disproportionately link specific currency identifiers to narratives of risk or poverty; and (3) Downstream volatility, where currency substitutions induce predictive entropy, sentence misunderstanding, sentiment shifts, and credit default prediction flips in downstream tasks. Benchmarking 14 state-of-the-art architectures reveals a pervasive phenomenon of representational disparity, affecting several currencies. These findings suggest that current embedding practices inadvertently encode inequalities, posing significant risks for the fairness and reliability of global financial NLP applications.
EntroBench: Evaluating LLM Watermarking Under Multi-Entropy Scenarios and Practical User Operations
PDF ↗Large language models (LLMs) watermarking has been proposed as an active approach for content provenance verification, yet existing evaluations are largely confined to fixed entropy settings. In this paper, we introduce EntroBench, a benchmark for LLM watermarking that systematically covers three entropy levels and seven representative tasks. We conducted a fair evaluation of eight watermarking methods through hyper-parameter search based on an anchored dataset. We find that current approaches struggle to perform consistently across different entropy levels. Our analysis reveals a clear trade-off between watermark detectability and downstream output quality that varies across tasks and entropy conditions. Furthermore, we assess watermark robustness under realistic user interaction scenarios and show that common, non-adversarial user behaviors can substantially degrade watermark signals. These results indicate that practical usage-driven perturbations pose a significant challenge to current watermarking techniques. EntroBench provides a unified evaluation framework for studying these issues and supports the development of more adaptive and robust LLM watermarking methods. Dataset and codes are available at https://github.com/py-qin/EntroBench.
Multimodal Large Language Models (MLLMs) are increasingly deployed as social agents, yet their ability to integrate conflicting identity cues remains underexplored. We audit gender bias in ten recent MLLMs using a counterfactual cooperative gaming task that pairs synthetic voices with avatars of varying gender presentation and visual fidelity. Our analysis reveals distinct bias patterns that can occur independently: closed-source models (e.g., Gemini 2.5/3) exhibit a near-deterministic “voice-matching” bias that enforces binary alignment between voice and appearance, whereas open-weight models (e.g., Qwen-2.5-Omni-7B) show limited responsiveness to vocal cues and instead exhibit context-driven stereotypes, such as preferring male avatars in combat scenarios. We further find that reducing visual realism attenuates matching tendencies in some models. These findings demonstrate that multimodal fairness is not monolithic; models may appear unbiased on one dimension while enforcing strict identity congruence or role-based stereotypes on another. Code and data are available at https://github.com/halfhoon/whose-voice-whose-avatar.
LRBench and Judge-R1: Principled Evaluation and Training of LLM-Based Judges for Long-Context Reasoning
PDF ↗Large language models (LLMs) are increasingly used as judges to evaluate, rank, and supervise other models, yet their reliability in judging LLMs’ reasoning process under long-context settings remains underexplored. Existing benchmarks either overly rely on human annotators, who may miss subtle flaws in lengthy reasoning chains, or focus solely on final responses while ignoring the underlying context and reasoning process. We introduce Long-Reason Bench (LRBench), a large-scale benchmark for evaluating LLM-based judges. LRBench comprises over 100K annotated instances spanning medical, legal, and academic-review scenarios, with fine-grained labels indicating violations of six core principles: Logical Correctness, Factual Consistency, Bias and Fairness, Groundedness, Helpfulness, and Harmlessness. Experimental results reveal that state-of-the-art LLM judges struggle to identify nuanced reasoning errors in long contexts. To improve judge reliability, we further present Judge-R1, which combines reinforcement learning with multi-turn search to enable grounded and principle-aware evaluation. Across domains and principles, Judge-R1 consistently outperforms single-turn baselines, enabling scalable and trustworthy evaluation of LLM reasoning. Our dataset and code are available at https://github.com/Xinyi-0724/Judge-R1.
Generating spoken dialogue is inherently more complex than monologue text-to-speech (TTS), as it demands both realistic turn-taking and the maintenance of distinct speaker timbres. While existing autoregressive (AR) models have made progress, they often suffer from high inference latency and stability issues. To overcome these limitations, we propose ZipVoice-Dialog, a non-autoregressive (NAR) zero-shot spoken dialogue generation model based on flow-matching. Observing that applying vanilla flow-matching to dialogue generation leads to poor speech intelligibility and turn-taking precision, we introduce two simple yet effective methods to adapt flow-matching architectures for dialogue generation: (1) a curriculum learning strategy to ensure robust speech-text alignment, and (2) speaker-turn embeddings to govern precise speaker turn-taking. Additionally, we introduce dedicated strategies to support stereo dialogue generation.Recognizing the lack of training datasets in this field, we curate and release OpenDialog, the first large-scale (6.8k hours) open-source spoken dialogue dataset derived from in-the-wild speech data. Moreover, for fair and rigorous evaluations, we established a benchmark to comprehensively evaluate dialogue generation models. Experiments demonstrate the effectiveness of the proposed methods and dataset, showing that ZipVoice-Dialog achieves superior performance in inference speed, intelligibility, speaker turn-taking accuracy, and speaker similarity. Our code, model checkpoints, and the OpenDialog dataset are publicly available.
Large language models (LLMs) are increasingly entrusted with high-stakes decisions that affect human welfare. However, the principles and values that guide these models when distributing scarce societal resources remain largely unexamined. To address this, we introduce the Social Welfare Function (SWF) Benchmark, a dynamic simulation environment in which an LLM acts as a dictator, distributing tasks to heterogeneous recipients with different returns on investment (ROI). The benchmark is designed to create a dilemma between maximizing collective efficiency (i.e., overall ROI) and ensuring distributive fairness (measured by the Gini coefficient). We evaluate 20 state-of-the-art LLMs. Our findings reveal several key insights, including: (i) LLMs’ general ability, as measured by popular Arena leaderboards, misaligns with their allocation skills; (ii) Most LLMs exhibit a strong default utilitarian orientation, prioritizing overall productivity at the expense of inequality. (iii) Allocation behaviors are highly manipulated, easily perturbed by common persuasion strategies. These results highlight the risks of deploying current LLMs as societal decision-makers and underscore the need for specialized benchmarks and alignment for AI governance.
Large language models (LLMs) increasingly mediate decisions in domains where unfair treatment of demographic groups is unacceptable. Existing work probes when biased outputs appear, but gives little insight into the mechanisms that generate them, leaving existing mitigations largely fragile. In this paper, we conduct a systematic investigation of LLM unfairness and propose DiffHeads—a lightweight debiasing framework for LLMs. We first compare Direct-Answer (DA) prompting to Chain-of-Thought (CoT) prompting across eight representative open- and closed-source LLMs. DA will trigger the nature-bias component of the LLM and reduce measured unfairness by 391.9%- 534.5% in both one- and two-turn dialogues. Next, we define a token-to-head contribution score that traces each token’s influence back to individual attention heads. This reveals a small cluster of bias heads that activate under DA but stay largely dormant with CoT, providing the first causal link between prompting strategy and bias emergence. Finally, building on this insight, we propose DiffHeads, which identify bias heads through differential activation analysis between DA and CoT and selectively mask only those heads. DiffHeads reduces unfairness by 49.4%, and 40.3% under DA and CoT, respectively, without harming model utility.
Large Language Models (LLMs) are increasingly deployed in contact-center Quality Assurance (QA) to automate agent performance evaluation and coaching feedback. While LLMs offer unprecedented scalability and speed, their reliance on web-scale training data raises concerns regarding demographic and behavioral biases that may distort workforce assessment. We present a counterfactual fairness evaluation of LLM-based QA systems across 13 dimensions spanning three categories: Identity, Context, and Behavioral Style. Fairness is quantified using the Counterfactual Flip Rate (CFR), the frequency of binary judgment reversals, and the Mean Absolute Score Difference (MASD), the average shift in coaching or confidence scores across counterfactual pairs. Evaluating 18 LLMs on 3,000 real-world contact center transcripts, we find systematic disparities, with CFR ranging from 5.4% to 13.0% and consistent MASD shifts across confidence, positive, and improvement scores. Larger, more strongly aligned models show lower unfairness, though fairness does not track accuracy. Contextual priming of historical performance induces the most severe degradations (CFR up to 16.4%), while implicit linguistic identity cues remain a persistent bias source. Finally, we analyze the efficacy of fairness-aware prompting, finding that explicit instructions yield only modest improvements in evaluative consistency. Our findings underscore the need for standardized fairness auditing pipelines prior to deploying LLMs in high-stakes workforce evaluation.
Large Language Model (LLM) serving systems increasingly face strict time-to-first-token (TTFT) service-level objectives (SLOs), yet TTFT remains highly sensitive to router-side queueing effects. Prefill costs scale with prompt length, decode lengths are uncertain, and prefix locality creates strong performance skew across requests. Despite major advances in continuous batching and KV-cache management, today’s routers are often agnostic to request cost, which makes them vulnerable to head-of-line blocking and tail-latency amplification under mixed workloads. We propose QUARTZ, a quantile-aware routing and queueing layer for LLM serving that predicts conservative quantile-based request-cost proxies, rather than point estimates, using lightweight router-visible signals. QUARTZ uses these quantiles together with backlog-aware router signals to guide worker selection and admission decisions that better align with TTFT tail SLOs while preserving fairness. We implement QUARTZ as a router upgrade for SGLang and evaluate it on representative interactive and retrieval-augmented workloads. The results show reductions in TTFT tail latency and SLO violations across heterogeneous workloads.
Automatic evaluation metrics are essential for building multilingual translation systems. The common practice of evaluating these systems is averaging metric scores across languages, yet this is suspicious since metrics may suffer from cross-lingual scoring bias, where translations of equal quality receive different scores across languages. This problem has not been systematically studied because no benchmark exists that provides parallel-quality instances across languages, and expert annotation is not realistic. In this work, we propose XQ-MEval, a semi-automatically built dataset covering nine translation directions, to benchmark translation metrics. Specifically, we inject MQM-defined errors into gold translations automatically, filter them by native speakers for reliability, and merge errors to generate pseudo translations with controllable quality. These pseudo translations are then paired with corresponding sources and references to form triplets used in assessing the qualities of translation metrics. Using XQ-MEval, our experiments on nine representative metrics reveal the inconsistency between averaging and human judgment and provide the first empirical evidence of cross-lingual scoring bias. Finally, we propose a normalization strategy derived from XQ-MEval that aligns score distributions across languages, improving the fairness and reliability of multilingual metric evaluation.
Text-to-image generation models have achieved strong performance in culturally homogeneous settings, yet their ability to generate multicultural scenes—where people and landmarks originate from different cultures—remains largely unexplored. We introduce multicultural text-to-image generation as a new task and present the first benchmark designed to study this setting. Our dataset contains 9,000 images spanning five countries, three age groups, two genders, 25 historical landmarks, and five languages. Using this benchmark, we analyze the behavior of state-of-the-art text-to-image models across multiple dimensions, including alignment, image quality, aesthetics, knowledge, and fairness. As one strategy for composing cultural and demographic information, we explore MosAIG, a Multi-Agent framework that enhances multicultural image generation by leveraging large language models with distinct cultural personas. Our analysis shows that richer prompt composition can improve image quality and cultural grounding compared to simple prompts, while also revealing substantial disparities across languages and demographic groups. We release our dataset and code at https://github.com/AIM-SCU/MosAIG