arXiv subjects

Krzysztof Czarnecki

Publications and source records attributed to Krzysztof Czarnecki.

At least 19 recordsLinked to original sources

arXiv2026

LitTraceQA: A Benchmark for Multi-Stage Grounding and Verification in Scientific Question Answering

Scientific literature is increasingly used as a knowledge source for language models, retrieval-augmented generation systems, and research assistants, but answering research questions from papers requires more than fluent generation. A reliable system must identify the relevant papers, locate the concrete evidence that supports the answer, and produce a response that is faithful to that evidence. We present LitTraceQA, a benchmark for literature-grounded question answering over scientific papers. Given a research question and a metadata pool of papers, a system must return three connected outputs: canonical paper identifiers, supporting evidence locations, and answers in one or more requested formats, including free-form text, multiple-choice answers, and structured tables. LitTraceQA targets evidence types common in scientific reading: tables, figures, text spans, equations or algorithms, and citation contexts. The public development split contains 55 examples, including 26 hidden-source single-paper questions and 29 multi-paper questions, and provides gold papers, evidence annotations, and answers for local validation. We also analyze a larger final annotation collection with 4,978 unique-question records over 4,859 unique gold papers. By evaluating paper retrieval, evidence grounding, and answer accuracy separately, LitTraceQA provides a testbed for scientific QA systems that produce verifiable answers rather than unsupported summaries.

cs.CL↗

arXiv2026

Where Does the Answer Come From? Benchmarking View-Level Visual Evidence Identification in Multi-View MLLMs for Autonomous Driving

Multimodal large language models (MLLMs) achieve strong results on visual reasoning benchmarks, but answer accuracy alone does not indicate whether a model relied on the correct visual evidence. This gap is particularly important in multi-view driving scenes used for autonomous driving, where a model can produce a plausible answer while grounding it in the wrong camera view. We introduce a multi-view visual question answering benchmark for evaluating evidence-source identification: given six synchronized NuScenes views and a question, the model must identify the supporting camera view and answer the question. The benchmark contains 122 conflict-centric question-answer pairs from 73 scenes, spanning causality, counterfactual reasoning, and intent prediction. View labels are proposed by an automatic conflict-mining pipeline and manually verified by annotators. We evaluate three settings: camera-view selection, oracle QA given the golden view, and joint prediction in which the model selects a view and answers in one pass. Answers are evaluated in both multiple-choice and free-form formats, using exact match for structured predictions and an LLM judge for free-form responses. By explicitly separating visual-source identification from answer correctness, the benchmark exposes grounding failures that answer-only evaluation misses.

cs.CL↗

arXiv2026

VISTAQA: Benchmarking Joint Visual Question Answering and Pixel-Level Evidence

Establishing a clear link between model predictions and the visual evidence that supports them is critical for transparency and reliability in multimodal reasoning, yet current multimodal large language model (MLLM) evaluations do not explicitly enforce this alignment. Existing benchmarks assess either textual answer correctness or pixel-level localization in isolation, leaving the coupling of reasoning and grounding an open challenge. We introduce VISTAQA, a comprehensive benchmark for joint evaluation of free-form answer correctness and pixel-level evidence grounding in visual question answering. VISTAQA comprises 1,157 expert-curated samples spanning six task types and six visual domains, ranging from direct perception to compositional and relational reasoning. VISTAQA requires models to not only answer correctly, but to also provide precise segmentation masks that support their answers. It also includes hallucination-aware examples where no valid visual evidence exists. To support this enhanced evaluation, we introduce GROVE, a unified evaluation metric that enforces joint correctness by combining textual accuracy and grounding quality via a per-sample geometric mean, ensuring neither dimension can compensate for deficiencies in the other. Comprehensive experiments across grounding-aware models and hybrid pipelines with general-purpose MLLMs reveal that even the strongest systems achieve limited performance under GROVE, highlighting a substantial gap between answer accuracy and visual evidence alignment.

cs.CV↗

arXiv2026

Rethinking the Mixture of Vision Encoders Paradigm for Enhanced Visual Understanding in Multimodal LLMs

Mixture of Vision Encoders (MoVE) has emerged as a powerful approach to enhance the fine-grained visual understanding of multimodal large language models (MLLMs), improving their ability to handle tasks such as complex optical character recognition and scene understanding. Despite these advances, effectively combining diverse encoders and their visual tokens, while also scaling to high-resolution inputs, remains an open challenge. In this work, we conduct a systematic study of fusion designs for MoVE-based MLLMs, highlighting principles for token-level integration across complementary encoders. Our study shows that a lightweight recipe consisting of post-adaptation fusion with independent projectors, tile-level sequence interleaving, and dynamic tiling with global context delivers strong performance on diverse benchmarks. We integrate these principles into a simple and effective architecture that we call LEO. Extensive evaluation on 11 vision-language benchmarks demonstrates that LEO achieves better results on the majority of tasks compared to existing MoVE-based approaches. Furthermore, LEO adapts effectively to the specialized domain of autonomous driving without altering its architecture or training recipe, achieving competitive performance against established baselines and thereby highlighting its ability to generalize. The code is available at https://github.com/Mozhgan91/LEO.

cs.CV↗

arXiv2026

Stalled, Biased, and Confused: Uncovering Reasoning Failures in LLMs for Cloud-Based Root Cause Analysis

Root cause analysis (RCA) is essential for diagnosing failures within complex software systems to ensure system reliability. The highly distributed and interdependent nature of modern cloud-based systems often complicates RCA efforts, particularly for multi-hop fault propagation, where symptoms appear far from their true causes. Recent advancements in Large Language Models (LLMs) present new opportunities to enhance automated RCA. However, their practical value for RCA depends on the fidelity of reasoning and decision-making. Existing work relies on historical incident corpora, operates directly on high-volume telemetry beyond current LLM capacity, or embeds reasoning inside complex multi-agent pipelines -- conditions that obscure whether failures arise from reasoning itself or from peripheral design choices. We present a focused empirical evaluation that isolates an LLM's reasoning behavior. We design a controlled experimental framework that foregrounds the LLM by using a simplified experimental setting. We evaluate six LLMs under two agentic workflows (ReAct and Plan-and-Execute) and a non-agentic baseline on two real-world case studies (GAIA and OpenRCA). In total, we executed 48,000 simulated failure scenarios, totaling 228 days of execution time. We measure both root-cause accuracy and the quality of intermediate reasoning traces. We produce a labeled taxonomy of 16 common RCA reasoning failures and use an LLM-as-a-Judge for annotation. Our results clarify where current open-source LLMs succeed and fail in multi-hop RCA, quantify sensitivity to input data modalities, and identify reasoning failures that predict final correctness. Together, these contributions provide transparent and reproducible empirical results and a failure taxonomy to guide future work on reasoning-driven system diagnosis.

cs.SE↗

arXiv2026

Mitigating the Modality Gap: Few-Shot Out-of-Distribution Detection with Multi-modal Prototypes and Image Bias Estimation

Existing vision-language model (VLM)-based methods for out-of-distribution (OOD) detection typically rely on similarity scores between input images and in-distribution (ID) text prototypes. However, the modality gap between image and text often results in high false positive rates, as OOD samples can exhibit high similarity to ID text prototypes. To mitigate the impact of this modality gap, we propose incorporating ID image prototypes along with ID text prototypes. We present theoretical analysis and empirical evidence indicating that this approach enhances VLM-based OOD detection performance without any additional training. To further reduce the gap between image and text, we introduce a novel few-shot tuning framework, SUPREME, comprising biased prompts generation (BPG) and image-text consistency (ITC) modules. BPG enhances image-text fusion and improves generalization by conditioning ID text prototypes on the Gaussian-based estimated image domain bias; ITC reduces the modality gap by minimizing intra- and inter-modal distances. Moreover, inspired by our theoretical and empirical findings, we introduce a novel OOD score $S_{\textit{GMP}}$, leveraging uni- and cross-modal similarities. Finally, we present extensive experiments to demonstrate that SUPREME consistently outperforms existing VLM-based OOD detection methods.

cs.CV↗

arXiv2025

HAWAII: Hierarchical Visual Knowledge Transfer for Efficient Vision-Language Models

Improving the visual understanding ability of vision-language models (VLMs) is crucial for enhancing their performance across various tasks. While using multiple pretrained visual experts has shown great promise, it often incurs significant computational costs during training and inference. To address this challenge, we propose HAWAII, a novel framework that distills knowledge from multiple visual experts into a single vision encoder, enabling it to inherit the complementary strengths of several experts with minimal computational overhead. To mitigate conflicts among different teachers and switch between different teacher-specific knowledge, instead of using a fixed set of adapters for multiple teachers, we propose to use teacher-specific Low-Rank Adaptation (LoRA) adapters with a corresponding router. Each adapter is aligned with a specific teacher, avoiding noisy guidance during distillation. To enable efficient knowledge distillation, we propose fine-grained and coarse-grained distillation. At the fine-grained level, token importance scores are employed to emphasize the most informative tokens from each teacher adaptively. At the coarse-grained level, we summarize the knowledge from multiple teachers and transfer it to the student using a set of general-knowledge LoRA adapters with a router. Extensive experiments on various vision-language tasks demonstrate the superiority of HAWAII compared to popular open-source VLMs. The code is available at https://github.com/yimuwangcs/wise-hawaii.

cs.CV↗

arXiv2025

LEO-MINI: An Efficient Multimodal Large Language Model using Conditional Token Reduction and Mixture of Multi-Modal Experts

Redundancy of visual tokens in multi-modal large language models (MLLMs) significantly reduces their computational efficiency. Recent approaches, such as resamplers and summarizers, have sought to reduce the number of visual tokens, but at the cost of visual reasoning ability. To address this, we propose LEO-MINI, a novel MLLM that significantly reduces the number of visual tokens and simultaneously boosts visual reasoning capabilities. For efficiency, LEO-MINI incorporates CoTR, a novel token reduction module to consolidate a large number of visual tokens into a smaller set of tokens, using the similarity between visual tokens, text tokens, and a compact learnable query. For effectiveness, to scale up the model's ability with minimal computational overhead, LEO-MINI employs MMoE, a novel mixture of multi-modal experts module. MMOE employs a set of LoRA experts with a novel router to switch between them based on the input text and visual tokens instead of only using the input hidden state. MMoE also includes a general LoRA expert that is always activated to learn general knowledge for LLM reasoning. For extracting richer visual features, MMOE employs a set of vision experts trained on diverse domain-specific data. To demonstrate LEO-MINI's improved efficiency and performance, we evaluate it against existing efficient MLLMs on various benchmark vision-language tasks.

cs.CV↗

arXiv2025

How Hard Is Snow? A Paired Domain Adaptation Dataset for Clear and Snowy Weather: CADC+

The impact of snowfall on 3D object detection performance remains underexplored. Conducting such an evaluation requires a dataset with sufficient labelled data from both weather conditions, ideally captured in the same driving environment. Current driving datasets with LiDAR point clouds either do not provide enough labelled data in both snowy and clear weather conditions, or rely on de-snowing methods to generate synthetic clear weather. Synthetic data often lacks realism and introduces an additional domain shift that confounds accurate evaluations. To address these challenges, we present CADC+, the first paired weather domain adaptation dataset for autonomous driving in winter conditions. CADC+ extends the Canadian Adverse Driving Conditions dataset (CADC) using clear weather data that was recorded on the same roads and in the same period as CADC. To create CADC+, we pair each CADC sequence with a clear weather sequence that matches the snowy sequence as closely as possible. CADC+ thus minimizes the domain shift resulting from factors unrelated to the presence of snow. We also present some preliminary results using CADC+ to evaluate the effect of snow on 3D object detection performance. We observe that snow introduces a combination of aleatoric and epistemic uncertainties, acting as both noise and a distinct data domain.

cs.CV↗

arXiv2025

MFSeg: Efficient Multi-frame 3D Semantic Segmentation

We propose MFSeg, an efficient multi-frame 3D semantic segmentation framework. By aggregating point cloud sequences at the feature level and regularizing the feature extraction and aggregation process, MFSeg reduces computational overhead while maintaining high accuracy. Moreover, by employing a lightweight MLP-based point decoder, our method eliminates the need to upsample redundant points from past frames. Experiments on the nuScenes and Waymo datasets show that MFSeg outperforms existing methods, demonstrating its effectiveness and efficiency.

cs.CV↗

arXiv2025

OV-SCAN: Semantically Consistent Alignment for Novel Object Discovery in Open-Vocabulary 3D Object Detection

Open-vocabulary 3D object detection for autonomous driving aims to detect novel objects beyond the predefined training label sets in point cloud scenes. Existing approaches achieve this by connecting traditional 3D object detectors with vision-language models (VLMs) to regress 3D bounding boxes for novel objects and perform open-vocabulary classification through cross-modal alignment between 3D and 2D features. However, achieving robust cross-modal alignment remains a challenge due to semantic inconsistencies when generating corresponding 3D and 2D feature pairs. To overcome this challenge, we present OV-SCAN, an Open-Vocabulary 3D framework that enforces Semantically Consistent Alignment for Novel object discovery. OV-SCAN employs two core strategies: discovering precise 3D annotations and filtering out low-quality or corrupted alignment pairs (arising from 3D annotation, occlusion-induced, or resolution-induced noise). Extensive experiments on the nuScenes dataset demonstrate that OV-SCAN achieves state-of-the-art performance.

cs.CV↗

arXiv2024

VADet: Multi-frame LiDAR 3D Object Detection using Variable Aggregation

Input aggregation is a simple technique used by state-of-the-art LiDAR 3D object detectors to improve detection. However, increasing aggregation is known to have diminishing returns and even performance degradation, due to objects responding differently to the number of aggregated frames. To address this limitation, we propose an efficient adaptive method, which we call Variable Aggregation Detection (VADet). Instead of aggregating the entire scene using a fixed number of frames, VADet performs aggregation per object, with the number of frames determined by an object's observed properties, such as speed and point density. VADet thus reduces the inherent trade-offs of fixed aggregation and is not architecture specific. To demonstrate its benefits, we apply VADet to three popular single-stage detectors and achieve state-of-the-art performance on the Waymo dataset.

cs.CV↗

arXiv2024

SSL-Interactions: Pretext Tasks for Interactive Trajectory Prediction

This paper addresses motion forecasting in multi-agent environments, pivotal for ensuring safety of autonomous vehicles. Traditional as well as recent data-driven marginal trajectory prediction methods struggle to properly learn non-linear agent-to-agent interactions. We present SSL-Interactions that proposes pretext tasks to enhance interaction modeling for trajectory prediction. We introduce four interaction-aware pretext tasks to encapsulate various aspects of agent interactions: range gap prediction, closest distance prediction, direction of movement prediction, and type of interaction prediction. We further propose an approach to curate interaction-heavy scenarios from datasets. This curated data has two advantages: it provides a stronger learning signal to the interaction model, and facilitates generation of pseudo-labels for interaction-centric pretext tasks. We also propose three new metrics specifically designed to evaluate predictions in interactive scenes. Our empirical evaluations indicate SSL-Interactions outperforms state-of-the-art motion forecasting methods quantitatively with up to 8% improvement, and qualitatively, for interaction-heavy scenarios.

cs.CV↗

arXiv2024

Solution Concepts in Hierarchical Games under Bounded Rationality with Applications to Autonomous Driving

With autonomous vehicles (AV) set to integrate further into regular human traffic, there is an increasing consensus on treating AV motion planning as a multi-agent problem. However, the traditional game-theoretic assumption of complete rationality is too strong for human driving, and there is a need for understanding human driving as a \emph{bounded rational} activity through a behavioural game-theoretic lens. To that end, we adapt four metamodels of bounded rational behaviour: three based on Quantal level-k and one based on Nash equilibrium with quantal errors. We formalize the different solution concepts that can be applied in the context of hierarchical games, a framework used in multi-agent motion planning, for the purpose of creating game theoretic models of driving behaviour. Furthermore, based on a contributed dataset of human driving at a busy urban intersection with a total of approximately 4k agents and 44k decision points, we evaluate the behaviour models on the basis of model fit to naturalistic data, as well as their predictive capacity. Our results suggest that among the behaviour models evaluated, at the level of maneuvers, modeling driving behaviour as an adaptation of the Quantal level-k model with level-0 behaviour modelled as pure rule-following provides the best fit to naturalistic driving behaviour. At the level of trajectories, bounds sampling of actions and a maxmax non-strategic models is the most accurate within the set of models in comparison. We also find a significant impact of situational factors on the performance of behaviour models.

cs.AI↗

arXiv2024

A Driver-Vehicle Model for ADS Scenario-based Testing

Scenario-based testing for automated driving systems (ADS) must be able to simulate traffic scenarios that rely on interactions with other vehicles. Although many languages for high-level scenario modelling have been proposed, they lack the features to precisely and reliably control the required micro-simulation, while also supporting behavior reuse and test reproducibility for a wide range of interactive scenarios. To fill this gap between scenario design and execution, we propose the Simulated Driver-Vehicle (SDV) model to represent and simulate vehicles as dynamic entities with their behavior being constrained by scenario design and goals set by testers. The model combines driver and vehicle as a single entity. It is based on human-like driving and the mechanical limitations of real vehicles for realistic simulation. The model leverages behavior trees to express high-level behaviors in terms of lower-level maneuvers, affording multiple driving styles and reuse. Furthermore, optimization-based maneuver planners guide the simulated vehicles towards the desired behavior. Our extensive evaluation shows the model's design effectiveness using NHTSA pre-crash scenarios, its motion realism in comparison to naturalistic urban traffic, and its scalability with traffic density. Finally, we show the applicability of our SDV model to test a real ADS and to identify crash scenarios, which are impractical to represent using predefined vehicle trajectories. The SDV model instances can be injected into existing simulation environments via co-simulation.

cs.RO↗

arXiv2024

Enhancing Safety in Mixed Traffic: Learning-Based Modeling and Efficient Control of Autonomous and Human-Driven Vehicles

With the increasing presence of autonomous vehicles (AVs) on public roads, developing robust control strategies to navigate the uncertainty of human-driven vehicles (HVs) is crucial. This paper introduces an advanced method for modeling HV behavior, combining a first-principles model with Gaussian process (GP) learning to enhance velocity prediction accuracy and provide a measurable uncertainty. We validated this innovative HV model using real-world data from field experiments and applied it to develop a GP-enhanced model predictive control (GP-MPC) strategy. This strategy aims to improve safety in mixed vehicle platoons by integrating uncertainty assessment into distance constraints. Comparative simulation studies with a conventional model predictive control (MPC) approach demonstrated that our GP-MPC strategy ensures more reliable safe distancing and fosters efficient vehicular dynamics, achieving notably higher speeds within the platoon. By incorporating a sparse GP technique in HV modeling and adopting a dynamic GP prediction within the MPC framework, we significantly reduced the computation time of GP-MPC, marking it only 4.6% higher than that of the conventional MPC. This represents a substantial improvement, making the process about 100 times faster than our preliminary work without these approximations. Our findings underscore the effectiveness of learning-based HV modeling in enhancing both safety and operational efficiency in mixed-traffic environments, paving the way for more harmonious AV-HV interactions.

cs.RO↗

arXiv2024

Assessing Visually-Continuous Corruption Robustness of Neural Networks Relative to Human Performance

While Neural Networks (NNs) have surpassed human accuracy in image classification on ImageNet, they often lack robustness against image corruption, i.e., corruption robustness. Yet such robustness is seemingly effortless for human perception. In this paper, we propose visually-continuous corruption robustness (VCR) -- an extension of corruption robustness to allow assessing it over the wide and continuous range of changes that correspond to the human perceptive quality (i.e., from the original image to the full distortion of all perceived visual information), along with two novel human-aware metrics for NN evaluation. To compare VCR of NNs with human perception, we conducted extensive experiments on 14 commonly used image corruptions with 7,718 human participants and state-of-the-art robust NN models with different training objectives (e.g., standard, adversarial, corruption robustness), different architectures (e.g., convolution NNs, vision transformers), and different amounts of training data augmentation. Our study showed that: 1) assessing robustness against continuous corruption can reveal insufficient robustness undetected by existing benchmarks; as a result, 2) the gap between NN and human robustness is larger than previously known; and finally, 3) some image corruptions have a similar impact on human perception, offering opportunities for more cost-effective robustness assessments. Our validation set with 14 image corruptions, human robustness data, and the evaluation code is provided as a toolbox and a benchmark.

cs.CV↗

arXiv2024

STEAM & MoSAFE: SOTIF Error-and-Failure Model & Analysis for AI-Enabled Driving Automation

Driving Automation Systems (DAS) are subject to complex road environments and vehicle behaviors and increasingly rely on sophisticated sensors and Artificial Intelligence (AI). These properties give rise to unique safety faults stemming from specification insufficiencies and technological performance limitations, where sensors and AI introduce errors that vary in magnitude and temporal patterns, posing potential safety risks. The Safety of the Intended Functionality (SOTIF) standard emerges as a promising framework for addressing these concerns, focusing on scenario-based analysis to identify hazardous behaviors and their causes. Although the current standard provides a basic cause-and-effect model and high-level process guidance, it lacks concepts required to identify and evaluate hazardous errors, especially within the context of AI. This paper introduces two key contributions to bridge this gap. First, it defines the SOTIF Temporal Error and Failure Model (STEAM) as a refinement of the SOTIF cause-and-effect model, offering a comprehensive system-design perspective. STEAM refines error definitions, introduces error sequences, and classifies them as error sequence patterns, providing particular relevance to systems employing advanced sensors and AI. Second, this paper proposes the Model-based SOTIF Analysis of Failures and Errors (MoSAFE) method, which allows instantiating STEAM based on system-design models by deriving hazardous error sequence patterns at module level from hazardous behaviors at vehicle level via weakest precondition reasoning. Finally, the paper presents a case study centered on an automated speed-control feature, illustrating the practical applicability of the refined model and the MoSAFE method in addressing complex safety challenges in DAS.

cs.LG↗