1. Query-Kontext: An Unified Multimodal Model for Image Generation and Editing
Authors: Yuxin Song, Wenkai Dong, Shizun Wang, Qi Zhang, Song Xue, Tao Yuan, Hu Yang, Haocheng Feng, Hang Zhou, Xinyan Xiao, Jingdong Wang •
Published: 2025-09-30 •
Source: arXiv
Unified Multimodal Models (UMMs) have demonstrated remarkable performance in text-to-image generation (T2I) and editing (TI2I), whether instantiated as assembled unified frameworks which couple powerful vision-language model (VLM) with diffusion-based generator, or as naive Unified Multimodal Models with an early fusion of understanding and generation modalities. We contend that in current unified frameworks, the crucial capability of multimodal generative reasoning which encompasses instruction understanding, grounding, and image referring for identity preservation and faithful reconstruction, is intrinsically entangled with high-fidelity synthesis. In this work, we introduce Query-Kontext, a novel approach that bridges the VLM and diffusion model via a multimodal ``kontext'' composed of semantic cues and coarse-grained image conditions encoded from multimodal inputs. This design delegates the complex ability of multimodal generative reasoning to powerful VLM while reserving diffusion model's role for high-quality visual synthesis. To achieve this, we propose a three-stage progressive training strategy. First, we connect the VLM to a lightweight diffusion head via multimodal kontext tokens to unleash the VLM's generative reasoning ability. Second, we scale this head to a large, pre-trained diffusion model to enhance visual detail and realism. Finally, we introduce a low-level image encoder to improve image fidelity and perform instruction tuning on downstream tasks. Furthermore, we build a comprehensive data pipeline integrating real, synthetic, and open-source datasets, covering diverse multimodal reference-to-image scenarios, including image generation, instruction-driven editing, customized generation, and multi-subject composition. Experiments show that our approach matches strong unified baselines and even outperforms task-specific state-of-the-art methods in several cases.
2. Branching Out: Broadening AI Measurement and Evaluation with Measurement Trees
Authors: Craig Greenberg, Patrick Hall, Theodore Jensen, Kristen Greene, Razvan Amironesei •
Published: 2025-09-30 •
Source: arXiv
This paper introduces \textit{measurement trees}, a novel class of metrics designed to combine various constructs into an interpretable multi-level representation of a measurand. Unlike conventional metrics that yield single values, vectors, surfaces, or categories, measurement trees produce a hierarchical directed graph in which each node summarizes its children through user-defined aggregation methods. In response to recent calls to expand the scope of AI system evaluation, measurement trees enhance metric transparency and facilitate the integration of heterogeneous evidence, including, e.g., agentic, business, energy-efficiency, sociotechnical, or security signals. We present definitions and examples, demonstrate practical utility through a large-scale measurement exercise, and provide accompanying open-source Python code. By operationalizing a transparent approach to measurement of complex constructs, this work offers a principled foundation for broader and more interpretable AI evaluation.
3. Learning Generalizable Shape Completion with SIM(3) Equivariance
Authors: Yuqing Wang, Zhaiyu Chen, Xiao Xiang Zhu •
Published: 2025-09-30 •
Source: arXiv
3D shape completion methods typically assume scans are pre-aligned to a canonical frame. This leaks pose and scale cues that networks may exploit to memorize absolute positions rather than inferring intrinsic geometry. When such alignment is absent in real data, performance collapses. We argue that robust generalization demands architectural equivariance to the similarity group, SIM(3), so the model remains agnostic to pose and scale. Following this principle, we introduce the first SIM(3)-equivariant shape completion network, whose modular layers successively canonicalize features, reason over similarity-invariant geometry, and restore the original frame. Under a de-biased evaluation protocol that removes the hidden cues, our model outperforms both equivariant and augmentation baselines on the PCN benchmark. It also sets new cross-domain records on real driving and indoor scans, lowering minimal matching distance on KITTI by 17% and Chamfer distance $\ell1$ on OmniObject3D by 14%. Perhaps surprisingly, ours under the stricter protocol still outperforms competitors under their biased settings. These results establish full SIM(3) equivariance as an effective route to truly generalizable shape completion. Project page: https://sime-completion.github.io.
4. Learning to See Before Seeing: Demystifying LLM Visual Priors from Language Pre-training
Authors: Junlin Han, Shengbang Tong, David Fan, Yufan Ren, Koustuv Sinha, Philip Torr, Filippos Kokkinos •
Published: 2025-09-30 •
Source: arXiv
Large Language Models (LLMs), despite being trained on text alone, surprisingly develop rich visual priors. These priors allow latent visual capabilities to be unlocked for vision tasks with a relatively small amount of multimodal data, and in some cases, to perform visual tasks without ever having seen an image. Through systematic analysis, we reveal that visual priors-the implicit, emergent knowledge about the visual world acquired during language pre-training-are composed of separable perception and reasoning priors with unique scaling trends and origins. We show that an LLM's latent visual reasoning ability is predominantly developed by pre-training on reasoning-centric data (e.g., code, math, academia) and scales progressively. This reasoning prior acquired from language pre-training is transferable and universally applicable to visual reasoning. In contrast, a perception prior emerges more diffusely from broad corpora, and perception ability is more sensitive to the vision encoder and visual instruction tuning data. In parallel, text describing the visual world proves crucial, though its performance impact saturates rapidly. Leveraging these insights, we propose a data-centric recipe for pre-training vision-aware LLMs and verify it in 1T token scale pre-training. Our findings are grounded in over 100 controlled experiments consuming 500,000 GPU-hours, spanning the full MLLM construction pipeline-from LLM pre-training to visual alignment and supervised multimodal fine-tuning-across five model scales, a wide range of data categories and mixtures, and multiple adaptation setups. Along with our main findings, we propose and investigate several hypotheses, and introduce the Multi-Level Existence Bench (MLE-Bench). Together, this work provides a new way of deliberately cultivating visual priors from language pre-training, paving the way for the next generation of multimodal LLMs.
5. Singularities at the vertex of connected angular inhomogeneities under thermal and elastic loading
Authors: Yuanpeng Yang, Huiming Yin, Chunlin Wu •
Published: 2025-09-30 •
Source: arXiv
This paper investigates the singularities at the vertex of multiply connected angular inhomogeneities for heat conduction and elastic deformation. With the aid of Eshelby's equivalent inclusion method (EIM), each inhomogeneity is simulated as an equivalent inclusion, exhibiting the same material properties as the matrix but containing a continuously distributed eigen-field with potential singularities at the vertices and edge lines. Specifically, the eigen-temperature-gradient (ETG) and eigenstrain are utilized to simulate material mismatch of thermal conductivity and stiffness, respectively. Using the separation of variables, the eigen-fields can be formulated in terms of distance to vertices and opening angles, and disturbed thermal/elastic fields are evaluated by domain integrals of Green's function multiplied by eigen-fields, which form Fredholm's integral equation of the second kind. The boundary value problem is reduced to solve for eigenvalues, which are used to determine the order of singularity. The present solution is versatile - by placing two identical inhomogeneities together, it recovers the classic solutions for a single wedge in a bimaterial media or infinite domain. The general and analytical formulae take full consideration of interactions of multiple inhomogeneities and reveal the effects of opening angles and material properties on the thermal and elastic singularities.
6. Video Object Segmentation-Aware Audio Generation
Authors: Ilpo Viertola, Vladimir Iashin, Esa Rahtu •
Published: 2025-09-30 •
Source: arXiv
Existing multimodal audio generation models often lack precise user control, which limits their applicability in professional Foley workflows. In particular, these models focus on the entire video and do not provide precise methods for prioritizing a specific object within a scene, generating unnecessary background sounds, or focusing on the wrong objects. To address this gap, we introduce the novel task of video object segmentation-aware audio generation, which explicitly conditions sound synthesis on object-level segmentation maps. We present SAGANet, a new multimodal generative model that enables controllable audio generation by leveraging visual segmentation masks along with video and textual cues. Our model provides users with fine-grained and visually localized control over audio generation. To support this task and further research on segmentation-aware Foley, we propose Segmented Music Solos, a benchmark dataset of musical instrument performance videos with segmentation information. Our method demonstrates substantial improvements over current state-of-the-art methods and sets a new standard for controllable, high-fidelity Foley synthesis. Code, samples, and Segmented Music Solos are available at https://saganet.notion.site
7. DeepScientist: Advancing Frontier-Pushing Scientific Findings Progressively
Authors: Yixuan Weng, Minjun Zhu, Qiujie Xie, Qiyao Sun, Zhen Lin, Sifan Liu, Yue Zhang •
Published: 2025-09-30 •
Source: arXiv
While previous AI Scientist systems can generate novel findings, they often lack the focus to produce scientifically valuable contributions that address pressing human-defined challenges. We introduce DeepScientist, a system designed to overcome this by conducting goal-oriented, fully autonomous scientific discovery over month-long timelines. It formalizes discovery as a Bayesian Optimization problem, operationalized through a hierarchical evaluation process consisting of "hypothesize, verify, and analyze". Leveraging a cumulative Findings Memory, this loop intelligently balances the exploration of novel hypotheses with exploitation, selectively promoting the most promising findings to higher-fidelity levels of validation. Consuming over 20,000 GPU hours, the system generated about 5,000 unique scientific ideas and experimentally validated approximately 1100 of them, ultimately surpassing human-designed state-of-the-art (SOTA) methods on three frontier AI tasks by 183.7\%, 1.9\%, and 7.9\%. This work provides the first large-scale evidence of an AI achieving discoveries that progressively surpass human SOTA on scientific tasks, producing valuable findings that genuinely push the frontier of scientific discovery. To facilitate further research into this process, we will open-source all experimental logs and system code at https://github.com/ResearAI/DeepScientist/.
8. MENLO: From Preferences to Proficiency -- Evaluating and Modeling Native-like Quality Across 47 Languages
Authors: Chenxi Whitehouse, Sebastian Ruder, Tony Lin, Oksana Kurylo, Haruka Takagi, Janice Lam, Nicolò Busetto, Denise Diaz •
Published: 2025-09-30 •
Source: arXiv
Ensuring native-like quality of large language model (LLM) responses across many languages is challenging. To address this, we introduce MENLO, a framework that operationalizes the evaluation of native-like response quality based on audience design-inspired mechanisms. Using MENLO, we create a dataset of 6,423 human-annotated prompt-response preference pairs covering four quality dimensions with high inter-annotator agreement in 47 language varieties. Our evaluation reveals that zero-shot LLM judges benefit significantly from pairwise evaluation and our structured annotation rubrics, yet they still underperform human annotators on our dataset. We demonstrate substantial improvements through fine-tuning with reinforcement learning, reward shaping, and multi-task learning approaches. Additionally, we show that RL-trained judges can serve as generative reward models to enhance LLMs' multilingual proficiency, though discrepancies with human judgment remain. Our findings suggest promising directions for scalable multilingual evaluation and preference alignment. We release our dataset and evaluation framework to support further research in multilingual LLM evaluation.
9. The distorted Fourier transform for the linearized Gross-Pitaevskii equation in the Hyperbolic plane
Authors: Oussama Landoulsi, Sohrab Shahshahani •
Published: 2025-09-30 •
Source: arXiv
Motivated by the stability problem for Ginzburg-Landau vortices on the hyperbolic plane, we develop the distorted Fourier transform for a general class of radial non-self-adjoint matrix Schr\"odinger operators on the hyperbolic plane. This applies in particular to the operator obtained by linearizing the equivariant Ginzburg-Landau equation on the hyperbolic plane around the degree one vortex. We systematically construct the distorted Fourier transform by writing the Stone formula for complex energies and taking the limit as the energy tends to the spectrum of the operator on the real line. This approach entails a careful analysis of the resolvent for complex energies in a neighborhood of the real line. It is the analogue of the approaches used in \cite{KS,ES2, LSS25}, where the limiting operator as $r\to\infty$ is not self-adjoint and which we carry out for all energies. Our analysis serves as the starting point for the study of the stability of the Ginzburg-Landau vortex under equivariant perturbations.
10. Autoproof: Automated Segmentation Proofreading for Connectomics
Authors: Gary B Huang, William M Katz, Stuart Berg, Louis Scheffer •
Published: 2025-09-30 •
Source: arXiv
Producing connectomes from electron microscopy (EM) images has historically required a great deal of human proofreading effort. This manual annotation cost is the current bottleneck in scaling EM connectomics, for example, in making larger connectome reconstructions feasible, or in enabling comparative connectomics where multiple related reconstructions are produced. In this work, we propose using the available ground-truth data generated by this manual annotation effort to learn a machine learning model to automate or optimize parts of the required proofreading workflows. We validate our approach on a recent complete reconstruction of the \emph{Drosophila} male central nervous system. We first show our method would allow for obtaining 90\% of the value of a guided proofreading workflow while reducing required cost by 80\%. We then demonstrate a second application for automatically merging many segmentation fragments to proofread neurons. Our system is able to automatically attach 200 thousand fragments, equivalent to four proofreader years of manual work, and increasing the connectivity completion rate of the connectome by 1.3\% points.
11. Fairness Testing in Retrieval-Augmented Generation: How Small Perturbations Reveal Bias in Small Language Models
Authors: Matheus Vinicius da Silva de Oliveira, Jonathan de Andrade Silva, Awdren de Lima Fontao •
Published: 2025-09-30 •
Source: arXiv
Large Language Models (LLMs) are widely used across multiple domains but continue to raise concerns regarding security and fairness. Beyond known attack vectors such as data poisoning and prompt injection, LLMs are also vulnerable to fairness bugs. These refer to unintended behaviors influenced by sensitive demographic cues (e.g., race or sexual orientation) that should not affect outcomes. Another key issue is hallucination, where models generate plausible yet false information. Retrieval-Augmented Generation (RAG) has emerged as a strategy to mitigate hallucinations by combining external retrieval with text generation. However, its adoption raises new fairness concerns, as the retrieved content itself may surface or amplify bias. This study conducts fairness testing through metamorphic testing (MT), introducing controlled demographic perturbations in prompts to assess fairness in sentiment analysis performed by three Small Language Models (SLMs) hosted on HuggingFace (Llama-3.2-3B-Instruct, Mistral-7B-Instruct-v0.3, and Llama-3.1-Nemotron-8B), each integrated into a RAG pipeline. Results show that minor demographic variations can break up to one third of metamorphic relations (MRs). A detailed analysis of these failures reveals a consistent bias hierarchy, with perturbations involving racial cues being the predominant cause of the violations. In addition to offering a comparative evaluation, this work reinforces that the retrieval component in RAG must be carefully curated to prevent bias amplification. The findings serve as a practical alert for developers, testers and small organizations aiming to adopt accessible SLMs without compromising fairness or reliability.
12. The Trajectory Bundle Method: Unifying Sequential-Convex Programming and Sampling-Based Trajectory Optimization
Authors: Kevin Tracy, John Z. Zhang, Jon Arrizabalaga, Stefan Schaal, Yuval Tassa, Tom Erez, Zachary Manchester •
Published: 2025-09-30 •
Source: arXiv
We present a unified framework for solving trajectory optimization problems in a derivative-free manner through the use of sequential convex programming. Traditionally, nonconvex optimization problems are solved by forming and solving a sequence of convex optimization problems, where the cost and constraint functions are approximated locally through Taylor series expansions. This presents a challenge for functions where differentiation is expensive or unavailable. In this work, we present a derivative-free approach to form these convex approximations by computing samples of the dynamics, cost, and constraint functions and letting the solver interpolate between them. Our framework includes sample-based trajectory optimization techniques like model-predictive path integral (MPPI) control as a special case and generalizes them to enable features like multiple shooting and general equality and inequality constraints that are traditionally associated with derivative-based sequential convex programming methods. The resulting framework is simple, flexible, and capable of solving a wide variety of practical motion planning and control problems.
13. Bidding strategies for energy storage players in 100% renewable electricity market: A game-theoretical approach
Authors: Arega Getaneh Abate, Salim Hassi, Dogan Keles, Xiufeng Liu, Xiao-Bing Zhang •
Published: 2025-09-30 •
Source: arXiv
The transition to electricity systems powered entirely by renewable energy sources makes energy storage indispensable for balancing intermittency and ensuring reliability. Since RES operate at near-zero marginal cost, storage operators can strongly influence electricity prices and energy security when renewable supply alone cannot meet demand. We develop a Cournot competition model in which storage operators strategically bid quantities to maximize their profits. We propose a MILP model with the big-M method and reformulation using continuous variables, incorporating demand blocks. The strategic bidding game is solved using the diagonalization algorithm, and the social planner's problem is used for benchmarking, cast as a one-shot optimization. The proposed model is applied to Denmark's power system using current data and 2030 renewable projections, capturing both current and future market conditions. Results show that storage operators affect market performance by arbitrage between low-and high-price periods, which can smooth supply-demand imbalances, thereby improving welfare relative to the no-storage case. With limited competition, however, strategic withholding increases prices and reduces welfare, while expanding storage capacity beyond a certain point yields no further gains. As the number of firms increases, competition mitigates distortions, and outcomes converge toward the social planner's benchmark with only two to three strategic players. These findings highlight storage's dual role in both stabilizing markets and creating market power. underscoring the need for market designs that align operators' incentives with social welfare.
14. Radio-based Multi-Robot Odometry and Relative Localization
Authors: Andrés Martínez-Silva, David Alejo, Luis Merino, Fernando Caballero •
Published: 2025-09-30 •
Source: arXiv
Radio-based methods such as Ultra-Wideband (UWB) and RAdio Detection And Ranging (radar), which have traditionally seen limited adoption in robotics, are experiencing a boost in popularity thanks to their robustness to harsh environmental conditions and cluttered environments. This work proposes a multi-robot UGV-UAV localization system that leverages the two technologies with inexpensive and readily-available sensors, such as Inertial Measurement Units (IMUs) and wheel encoders, to estimate the relative position of an aerial robot with respect to a ground robot. The first stage of the system pipeline includes a nonlinear optimization framework to trilaterate the location of the aerial platform based on UWB range data, and a radar pre-processing module with loosely coupled ego-motion estimation which has been adapted for a multi-robot scenario. Then, the pre-processed radar data as well as the relative transformation are fed to a pose-graph optimization framework with odometry and inter-robot constraints. The system, implemented for the Robotic Operating System (ROS 2) with the Ceres optimizer, has been validated in Software-in-the-Loop (SITL) simulations and in a real-world dataset. The proposed relative localization module outperforms state-of-the-art closed-form methods which are less robust to noise. Our SITL environment includes a custom Gazebo plugin for generating realistic UWB measurements modeled after real data. Conveniently, the proposed factor graph formulation makes the system readily extensible to full Simultaneous Localization And Mapping (SLAM). Finally, all the code and experimental data is publicly available to support reproducibility and to serve as a common open dataset for benchmarking.
15. Automated and Scalable SEM Image Analysis of Perovskite Solar Cell Materials via a Deep Segmentation Framework
Authors: Jian Guo Pan, Lin Wang, Xia Cai •
Published: 2025-09-30 •
Source: arXiv
Scanning Electron Microscopy (SEM) is indispensable for characterizing the microstructure of thin films during perovskite solar cell fabrication. Accurate identification and quantification of lead iodide and perovskite phases are critical because residual lead iodide strongly influences crystallization pathways and defect formation, while the morphology of perovskite grains governs carrier transport and device stability. Yet current SEM image analysis is still largely manual, limiting throughput and consistency. Here, we present an automated deep learning-based framework for SEM image segmentation that enables precise and efficient identification of lead iodide, perovskite and defect domains across diverse morphologies. Built upon an improved YOLOv8x architecture, our model named PerovSegNet incorporates two novel modules: (i) Adaptive Shuffle Dilated Convolution Block, which enhances multi-scale and fine-grained feature extraction through group convolutions and channel mixing; and (ii) Separable Adaptive Downsampling module, which jointly preserves fine-scale textures and large-scale structures for more robust boundary recognition. Trained on an augmented dataset of 10,994 SEM images, PerovSegNet achieves a mean Average Precision of 87.25% with 265.4 Giga Floating Point Operations, outperforming the baseline YOLOv8x-seg by 4.08%, while reducing model size and computational load by 24.43% and 25.22%, respectively. Beyond segmentation, the framework provides quantitative grain-level metrics, such as lead iodide/perovskite area and count, which can serve as reliable indicators of crystallization efficiency and microstructural quality. These capabilities establish PerovSegNet as a scalable tool for real-time process monitoring and data-driven optimization of perovskite thin-film fabrication.The source code is available at:https://github.com/wlyyj/PerovSegNet/tree/master.
16. Ferret-UI Lite: Lessons from Building Small On-Device GUI Agents
Authors: Zhen Yang, Zi-Yi Dou, Di Feng, Forrest Huang, Anh Nguyen, Keen You, Omar Attia, Yuhao Yang, Michael Feng, Haotian Zhang, Ram Ramrakhya, Chao Jia, Jeffrey Nichols, Alexander Toshev, Yinfei Yang, Zhe Gan •
Published: 2025-09-30 •
Source: arXiv
Developing autonomous agents that effectively interact with Graphic User Interfaces (GUIs) remains a challenging open problem, especially for small on-device models. In this paper, we present Ferret-UI Lite, a compact, end-to-end GUI agent that operates across diverse platforms, including mobile, web, and desktop. Utilizing techniques optimized for developing small models, we build our 3B Ferret-UI Lite agent through curating a diverse GUI data mixture from real and synthetic sources, strengthening inference-time performance through chain-of-thought reasoning and visual tool-use, and reinforcement learning with designed rewards. Ferret-UI Lite achieves competitive performance with other small-scale GUI agents. In GUI grounding, Ferret-UI Lite attains scores of $91.6\%$, $53.3\%$, and $61.2\%$ on the ScreenSpot-V2, ScreenSpot-Pro, and OSWorld-G benchmarks, respectively. For GUI navigation, Ferret-UI Lite achieves success rates of $28.0\%$ on AndroidWorld and $19.8\%$ on OSWorld. We share our methods and lessons learned from developing compact, on-device GUI agents.
17. Rearchitecting Datacenter Lifecycle for AI: A TCO-Driven Framework
Authors: Jovan Stojkovic, Chaojie Zhang, Íñigo Goiri, Ricardo Bianchini •
Published: 2025-09-30 •
Source: arXiv
The rapid rise of large language models (LLMs) has been driving an enormous demand for AI inference infrastructure, mainly powered by high-end GPUs. While these accelerators offer immense computational power, they incur high capital and operational costs due to frequent upgrades, dense power consumption, and cooling demands, making total cost of ownership (TCO) for AI datacenters a critical concern for cloud providers. Unfortunately, traditional datacenter lifecycle management (designed for general-purpose workloads) struggles to keep pace with AI's fast-evolving models, rising resource needs, and diverse hardware profiles. In this paper, we rethink the AI datacenter lifecycle scheme across three stages: building, hardware refresh, and operation. We show how design choices in power, cooling, and networking provisioning impact long-term TCO. We also explore refresh strategies aligned with hardware trends. Finally, we use operation software optimizations to reduce cost. While these optimizations at each stage yield benefits, unlocking the full potential requires rethinking the entire lifecycle. Thus, we present a holistic lifecycle management framework that coordinates and co-optimizes decisions across all three stages, accounting for workload dynamics, hardware evolution, and system aging. Our system reduces the TCO by up to 40\% over traditional approaches. Using our framework we provide guidelines on how to manage AI datacenter lifecycle for the future.
18. Indoor/Outdoor Spectrum Sharing Enabled by GNSS-based Classifiers
Authors: Hossein Nasiri, Muhammad Iqbal Rochman, Monisha Ghosh •
Published: 2025-09-30 •
Source: arXiv
The desirability of the mid-band frequency range (1 - 10 GHz) for federal and commercial applications, combined with the growing applications for commercial indoor use-cases, such as factory automation, opens up a new approach to spectrum sharing: the same frequency bands used outdoors by federal incumbents can be reused by commercial indoor users. A recent example of such sharing, between commercial systems, is the 6 GHz band (5.925 - 7.125 GHz) where unlicensed, low-power-indoor (LPI) users share the band with outdoor incumbents, primarily fixed microwave links. However, to date, there exist no reliable, automatic means of determining whether a device is indoors or outdoors, necessitating the use of other mechanisms such as mandating indoor access points (APs) to have integrated antennas and not be battery powered, and reducing transmit power of client devices which may be outdoors. An accurate indoor/outdoor (I/O) classification addresses these challenges, enabling automatic transmit power adjustments without interfering with incumbents. To this end, we leverage the Global Navigation Satellite System (GNSS) signals for I/O classification. GNSS signals, designed inherently for outdoor reception and highly susceptible to indoor attenuation and blocking, provide a robust and distinguishing feature for environmental sensing. We develop various methodologies, including threshold-based techniques and machine learning approaches and evaluate them using an expanded dataset gathered from diverse geographical locations. Our results demonstrate that GNSS-based methods alone can achieve greater accuracy than approaches relying solely on wireless (Wi-Fi) data, particularly in unfamiliar locations. Furthermore, the integration of GNSS data with Wi-Fi information leads to improved classification accuracy, showcasing the significant benefits of multi-modal data fusion.
19. OffTopicEval: When Large Language Models Enter the Wrong Chat, Almost Always!
Authors: Jingdi Lei, Varun Gumma, Rishabh Bhardwaj, Seok Min Lim, Chuan Li, Amir Zadeh, Soujanya Poria •
Published: 2025-09-30 •
Source: arXiv
Large Language Model (LLM) safety is one of the most pressing challenges for enabling wide-scale deployment. While most studies and global discussions focus on generic harms, such as models assisting users in harming themselves or others, enterprises face a more fundamental concern: whether LLM-based agents are safe for their intended use case. To address this, we introduce operational safety, defined as an LLM's ability to appropriately accept or refuse user queries when tasked with a specific purpose. We further propose OffTopicEval, an evaluation suite and benchmark for measuring operational safety both in general and within specific agentic use cases. Our evaluations on six model families comprising 20 open-weight LLMs reveal that while performance varies across models, all of them remain highly operationally unsafe. Even the strongest models -- Qwen-3 (235B) with 77.77\% and Mistral (24B) with 79.96\% -- fall far short of reliable operational safety, while GPT models plateau in the 62--73\% range, Phi achieves only mid-level scores (48--70\%), and Gemma and Llama-3 collapse to 39.53\% and 23.84\%, respectively. While operational safety is a core model alignment issue, to suppress these failures, we propose prompt-based steering methods: query grounding (Q-ground) and system-prompt grounding (P-ground), which substantially improve OOD refusal. Q-ground provides consistent gains of up to 23\%, while P-ground delivers even larger boosts, raising Llama-3.3 (70B) by 41\% and Qwen-3 (30B) by 27\%. These results highlight both the urgent need for operational safety interventions and the promise of prompt-based steering as a first step toward more reliable LLM-based agents.
20. VitaBench: Benchmarking LLM Agents with Versatile Interactive Tasks in Real-world Applications
Authors: Wei He, Yueqing Sun, Hongyan Hao, Xueyuan Hao, Zhikang Xia, Qi Gu, Chengcheng Han, Dengchang Zhao, Hui Su, Kefeng Zhang, Man Gao, Xi Su, Xiaodong Cai, Xunliang Cai, Yu Yang, Yunke Zhao •
Published: 2025-09-30 •
Source: arXiv
As LLM-based agents are increasingly deployed in real-life scenarios, existing benchmarks fail to capture their inherent complexity of handling extensive information, leveraging diverse resources, and managing dynamic user interactions. To address this gap, we introduce VitaBench, a challenging benchmark that evaluates agents on versatile interactive tasks grounded in real-world settings. Drawing from daily applications in food delivery, in-store consumption, and online travel services, VitaBench presents agents with the most complex life-serving simulation environment to date, comprising 66 tools. Through a framework that eliminates domain-specific policies, we enable flexible composition of these scenarios and tools, yielding 100 cross-scenario tasks (main results) and 300 single-scenario tasks. Each task is derived from multiple real user requests and requires agents to reason across temporal and spatial dimensions, utilize complex tool sets, proactively clarify ambiguous instructions, and track shifting user intent throughout multi-turn conversations. Moreover, we propose a rubric-based sliding window evaluator, enabling robust assessment of diverse solution pathways in complex environments and stochastic interactions. Our comprehensive evaluation reveals that even the most advanced models achieve only 30% success rate on cross-scenario tasks, and less than 50% success rate on others. Overall, we believe VitaBench will serve as a valuable resource for advancing the development of AI agents in practical real-world applications. The code, dataset, and leaderboard are available at https://vitabench.github.io/
21. fev-bench: A Realistic Benchmark for Time Series Forecasting
Authors: Oleksandr Shchur, Abdul Fatir Ansari, Caner Turkmen, Lorenzo Stella, Nick Erickson, Pablo Guerron, Michael Bohlke-Schneider, Yuyang Wang •
Published: 2025-09-30 •
Source: arXiv
Benchmark quality is critical for meaningful evaluation and sustained progress in time series forecasting, particularly given the recent rise of pretrained models. Existing benchmarks often have narrow domain coverage or overlook important real-world settings, such as tasks with covariates. Additionally, their aggregation procedures often lack statistical rigor, making it unclear whether observed performance differences reflect true improvements or random variation. Many benchmarks also fail to provide infrastructure for consistent evaluation or are too rigid to integrate into existing pipelines. To address these gaps, we propose fev-bench, a benchmark comprising 100 forecasting tasks across seven domains, including 46 tasks with covariates. Supporting the benchmark, we introduce fev, a lightweight Python library for benchmarking forecasting models that emphasizes reproducibility and seamless integration with existing workflows. Usingfev, fev-bench employs principled aggregation methods with bootstrapped confidence intervals to report model performance along two complementary dimensions: win rates and skill scores. We report results on fev-bench for various pretrained, statistical and baseline models, and identify promising directions for future research.
22. ErrorPrism: Reconstructing Error Propagation Paths in Cloud Service Systems
Authors: Junsong Pu, Yichen Li, Zhuangbin Chen, Jinyang Liu, Zhihan Jiang, Jianjun Chen, Rui Shi, Zibin Zheng, Tieying Zhang •
Published: 2025-09-30 •
Source: arXiv
Reliability management in cloud service systems is challenging due to the cascading effect of failures. Error wrapping, a practice prevalent in modern microservice development, enriches errors with context at each layer of the function call stack, constructing an error chain that describes a failure from its technical origin to its business impact. However, this also presents a significant traceability problem when recovering the complete error propagation path from the final log message back to its source. Existing approaches are ineffective at addressing this problem. To fill this gap, we present ErrorPrism in this work for automated reconstruction of error propagation paths in production microservice systems. ErrorPrism first performs static analysis on service code repositories to build a function call graph and map log strings to relevant candidate functions. This significantly reduces the path search space for subsequent analysis. Then, ErrorPrism employs an LLM agent to perform an iterative backward search to accurately reconstruct the complete, multi-hop error path. Evaluated on 67 production microservices at ByteDance, ErrorPrism achieves 97.0% accuracy in reconstructing paths for 102 real-world errors, outperforming existing static analysis and LLM-based approaches. ErrorPrism provides an effective and practical tool for root cause analysis in industrial microservice systems.
23. CreAgentive: An Agent Workflow Driven Multi-Category Creative Generation Engine
Authors: Yuyang Cheng, Linyue Cai, Changwei Peng, Yumiao Xu, Rongfang Bie, Yong Zhao •
Published: 2025-09-30 •
Source: arXiv
We present CreAgentive, an agent workflow driven multi-category creative generation engine that addresses four key limitations of contemporary large language models in writing stories, drama and other categories of creatives: restricted genre diversity, insufficient output length, weak narrative coherence, and inability to enforce complex structural constructs. At its core, CreAgentive employs a Story Prototype, which is a genre-agnostic, knowledge graph-based narrative representation that decouples story logic from stylistic realization by encoding characters, events, and environments as semantic triples. CreAgentive engages a three-stage agent workflow that comprises: an Initialization Stage that constructs a user-specified narrative skeleton; a Generation Stage in which long- and short-term objectives guide multi-agent dialogues to instantiate the Story Prototype; a Writing Stage that leverages this prototype to produce multi-genre text with advanced structures such as retrospection and foreshadowing. This architecture reduces storage redundancy and overcomes the typical bottlenecks of long-form generation. In extensive experiments, CreAgentive generates thousands of chapters with stable quality and low cost (less than $1 per 100 chapters) using a general-purpose backbone model. To evaluate performance, we define a two-dimensional framework with 10 narrative indicators measuring both quality and length. Results show that CreAgentive consistently outperforms strong baselines and achieves robust performance across diverse genres, approaching the quality of human-authored novels.
24. Transformer Classification of Breast Lesions: The BreastDCEDL_AMBL Benchmark Dataset and 0.92 AUC Baseline
Authors: Naomi Fridman, Anat Goldstein •
Published: 2025-09-30 •
Source: arXiv
The error is caused by special characters that arXiv's system doesn't recognize. Here's the cleaned version with all problematic characters replaced: Breast magnetic resonance imaging is a critical tool for cancer detection and treatment planning, but its clinical utility is hindered by poor specificity, leading to high false-positive rates and unnecessary biopsies. This study introduces a transformer-based framework for automated classification of breast lesions in dynamic contrast-enhanced MRI, addressing the challenge of distinguishing benign from malignant findings. We implemented a SegFormer architecture that achieved an AUC of 0.92 for lesion-level classification, with 100% sensitivity and 67% specificity at the patient level - potentially eliminating one-third of unnecessary biopsies without missing malignancies. The model quantifies malignant pixel distribution via semantic segmentation, producing interpretable spatial predictions that support clinical decision-making. To establish reproducible benchmarks, we curated BreastDCEDL_AMBL by transforming The Cancer Imaging Archive's AMBL collection into a standardized deep learning dataset with 88 patients and 133 annotated lesions (89 benign, 44 malignant). This resource addresses a key infrastructure gap, as existing public datasets lack benign lesion annotations, limiting benign-malignant classification research. Training incorporated an expanded cohort of over 1,200 patients through integration with BreastDCEDL datasets, validating transfer learning approaches despite primary tumor-only annotations. Public release of the dataset, models, and evaluation protocols provides the first standardized benchmark for DCE-MRI lesion classification, enabling methodological advancement toward clinical deployment.
25. The Grammar of FAIR: A Granular Architecture of Semantic Units for FAIR Semantics, Inspired by Biology and Linguistics
Authors: Lars Vogt, Barend Mons •
Published: 2025-09-30 •
Source: arXiv
The FAIR Principles aim to make data and knowledge Findable, Accessible, Interoperable, and Reusable, yet current digital infrastructures often lack a unifying semantic framework that bridges human cognition and machine-actionability. In this paper, we introduce the Grammar of FAIR: a granular and modular architecture for FAIR semantics built on the concept of semantic units. Semantic units, comprising atomic statement units and composite compound units, implement the principle of semantic modularisation, decomposing data and knowledge into independently identifiable, semantically meaningful, and machine-actionable units. A central metaphor guiding our approach is the analogy between the hierarchy of level of organisation in biological systems and the hierarchy of levels of organisation in information systems: both are structured by granular building blocks that mediate across multiple perspectives while preserving functional unity. Drawing further inspiration from concept formation and natural language grammar, we show how these building blocks map to FAIR Digitial Objects (FDOs), enabling format-agnostic semantic transitivity from natural language token models to schema-based representations. This dual biological-linguistic analogy provides a semantics-first foundation for evolving cross-ecosystem infrastructures, paving the way for the Internet of FAIR Data and Services (IFDS) and a future of modular, AI-ready, and citation-granular scholarly communication.
26. Text-Based Approaches to Item Alignment to Content Standards in Large-Scale Reading & Writing Tests
Authors: Yanbin Fu, Hong Jiao, Tianyi Zhou, Robert W. Lissitz, Nan Zhang, Ming Li, Qingshu Xu, Sydney Peters •
Published: 2025-09-30 •
Source: arXiv
Aligning test items to content standards is a critical step in test development to collect validity evidence based on content. Item alignment has typically been conducted by human experts. This judgmental process can be subjective and time-consuming. This study investigated the performance of fine-tuned small language models (SLMs) for automated item alignment using data from a large-scale standardized reading and writing test for college admissions. Different SLMs were trained for alignment at both domain and skill levels respectively with 10 skills mapped to 4 content domains. The model performance was evaluated in multiple criteria on two testing datasets. The impact of types and sizes of the input data for training was investigated. Results showed that including more item text data led to substantially better model performance, surpassing the improvements induced by sample size increase alone. For comparison, supervised machine learning models were trained using the embeddings from the multilingual-E5-large-instruct model. The study results showed that fine-tuned SLMs consistently outperformed the embedding-based supervised machine learning models, particularly for the more fine-grained skill alignment. To better understand model misclassifications, multiple semantic similarity analysis including pairwise cosine similarity, Kullback-Leibler divergence of embedding distributions, and two-dimension projections of item embeddings were conducted. These analyses consistently showed that certain skills in SAT and PSAT were semantically too close, providing evidence for the observed misclassification.
27. Game-Time: Evaluating Temporal Dynamics in Spoken Language Models
Authors: Kai-Wei Chang, En-Pei Hu, Chun-Yi Kuan, Wenze Ren, Wei-Chih Chen, Guan-Ting Lin, Yu Tsao, Shao-Hua Sun, Hung-yi Lee, James Glass •
Published: 2025-09-30 •
Source: arXiv
Conversational Spoken Language Models (SLMs) are emerging as a promising paradigm for real-time speech interaction. However, their capacity of temporal dynamics, including the ability to manage timing, tempo and simultaneous speaking, remains a critical and unevaluated challenge for conversational fluency. To address this gap, we introduce the Game-Time Benchmark, a framework to systematically assess these temporal capabilities. Inspired by how humans learn a language through language activities, Game-Time consists of basic instruction-following tasks and advanced tasks with temporal constraints, such as tempo adherence and synchronized responses. Our evaluation of diverse SLM architectures reveals a clear performance disparity: while state-of-the-art models handle basic tasks well, many contemporary systems still struggle with fundamental instruction-following. More critically, nearly all models degrade substantially under temporal constraints, exposing persistent weaknesses in time awareness and full-duplex interaction. The Game-Time Benchmark provides a foundation for guiding future research toward more temporally-aware conversational AI. Demos and datasets are available on our project website https://ga642381.github.io/Game-Time.
28. PANDA: Towards Generalist Video Anomaly Detection via Agentic AI Engineer
Authors: Zhiwei Yang, Chen Gao, Mike Zheng Shou •
Published: 2025-09-30 •
Source: arXiv
Video anomaly detection (VAD) is a critical yet challenging task due to the complex and diverse nature of real-world scenarios. Previous methods typically rely on domain-specific training data and manual adjustments when applying to new scenarios and unseen anomaly types, suffering from high labor costs and limited generalization. Therefore, we aim to achieve generalist VAD, i.e., automatically handle any scene and any anomaly types without training data or human involvement. In this work, we propose PANDA, an agentic AI engineer based on MLLMs. Specifically, we achieve PANDA by comprehensively devising four key capabilities: (1) self-adaptive scene-aware strategy planning, (2) goal-driven heuristic reasoning, (3) tool-augmented self-reflection, and (4) self-improving chain-of-memory. Concretely, we develop a self-adaptive scene-aware RAG mechanism, enabling PANDA to retrieve anomaly-specific knowledge for anomaly detection strategy planning. Next, we introduce a latent anomaly-guided heuristic prompt strategy to enhance reasoning precision. Furthermore, PANDA employs a progressive reflection mechanism alongside a suite of context-aware tools to iteratively refine decision-making in complex scenarios. Finally, a chain-of-memory mechanism enables PANDA to leverage historical experiences for continual performance improvement. Extensive experiments demonstrate that PANDA achieves state-of-the-art performance in multi-scenario, open-set, and complex scenario settings without training and manual involvement, validating its generalizable and robust anomaly detection capability. Code is released at https://github.com/showlab/PANDA.
29. MR$^2$-Bench: Going Beyond Matching to Reasoning in Multimodal Retrieval
Authors: Junjie Zhou, Ze Liu, Lei Xiong, Jin-Ge Yao, Yueze Wang, Shitao Xiao, Fenfen Lin, Miguel Hu Chen, Zhicheng Dou, Siqi Bao, Defu Lian, Yongping Xiong, Zheng Liu •
Published: 2025-09-30 •
Source: arXiv
Multimodal retrieval is becoming a crucial component of modern AI applications, yet its evaluation lags behind the demands of more realistic and challenging scenarios. Existing benchmarks primarily probe surface-level semantic correspondence (e.g., object-text matching) while failing to assess the deeper reasoning required to capture complex relationships between visual and textual information. To address this gap, we introduce MR$^2$-Bench, a reasoning-intensive benchmark for multimodal retrieval. MR$^2$-Bench presents the following critical values: 1) all tasks are reasoning-driven, going beyond shallow matching to effectively assess models' capacity for logical, spatial, and causal inference; 2) it features diverse multimodal data, such as natural images, diagrams, and visual puzzles, enabling comprehensive evaluation across content types; 3) it supports complex queries and documents containing multiple images and covers diverse retrieval scenarios, more accurately reflecting real-world applications. Our benchmark contains 1,309 curated queries, derived either from manual collection and annotation or from selective consolidation of public datasets. Despite achieving strong results on existing benchmarks, current state-of-the-art models still struggle on MR$^2$-Bench: for example, the leading Seed1.6-Embedding model attains a Recall@1 of 77.78 on MMEB, but only 9.91 on MR$^2$-Bench. This substantial performance gap highlights both the increased challenge posed by our benchmark and the pressing need for further advances in reasoning-intensive multimodal retrieval. The dataset and evaluation code will be made publicly available at https://github.com/VectorSpaceLab/MR2-Bench.
30. SafeBehavior: Simulating Human-Like Multistage Reasoning to Mitigate Jailbreak Attacks in Large Language Models
Authors: Qinjian Zhao, Jiaqi Wang, Zhiqiang Gao, Zhihao Dou, Belal Abuhaija, Kaizhu Huang •
Published: 2025-09-30 •
Source: arXiv
Large Language Models (LLMs) have achieved impressive performance across diverse natural language processing tasks, but their growing power also amplifies potential risks such as jailbreak attacks that circumvent built-in safety mechanisms. Existing defenses including input paraphrasing, multi step evaluation, and safety expert models often suffer from high computational costs, limited generalization, or rigid workflows that fail to detect subtle malicious intent embedded in complex contexts. Inspired by cognitive science findings on human decision making, we propose SafeBehavior, a novel hierarchical jailbreak defense mechanism that simulates the adaptive multistage reasoning process of humans. SafeBehavior decomposes safety evaluation into three stages: intention inference to detect obvious input risks, self introspection to assess generated responses and assign confidence based judgments, and self revision to adaptively rewrite uncertain outputs while preserving user intent and enforcing safety constraints. We extensively evaluate SafeBehavior against five representative jailbreak attack types including optimization based, contextual manipulation, and prompt based attacks and compare it with seven state of the art defense baselines. Experimental results show that SafeBehavior significantly improves robustness and adaptability across diverse threat scenarios, offering an efficient and human inspired approach to safeguarding LLMs against jailbreak attempts.
31. Kinodynamic Motion Planning for Mobile Robot Navigation across Inconsistent World Models
Authors: Eric R. Damm, Thomas M. Howard •
Published: 2025-09-30 •
Source: arXiv
Mobile ground robots lacking prior knowledge of an environment must rely on sensor data to develop a model of their surroundings. In these scenarios, consistent identification of obstacles and terrain features can be difficult due to noise and algorithmic shortcomings, which can make it difficult for motion planning systems to generate safe motions. One particular difficulty to overcome is when regions of the cost map switch between being marked as obstacles and free space through successive planning cycles. One potential solution to this, which we refer to as Valid in Every Hypothesis (VEH), is for the planning system to plan motions that are guaranteed to be safe through a history of world models. Another approach is to track a history of world models, and adjust node costs according to the potential penalty of needing to reroute around previously hazardous areas. This work discusses three major iterations on this idea. The first iteration, called PEH, invokes a sub-search for every node expansion that crosses through a divergence point in the world models. The second and third iterations, called GEH and GEGRH respectively, defer the sub-search until after an edge expands into the goal region. GEGRH uses an additional step to revise the graph based on divergent nodes in each world. Initial results showed that, although PEH and GEH find more optimistic solutions than VEH, they are unable to generate solutions in less than one-second, which exceeds our requirements for field deployment. Analysis of results from a field experiment in an unstructured, off-road environment on a Clearpath Robotics Warthog UGV indicate that GEGRH finds lower cost trajectories and has faster average planning times than VEH. Compared to single-hypothesis (SH) search, where only the latest world model is considered, GEGRH generates more conservative plans with a small increase in average planning time.
32. TrackCore-F: Deploying Transformer-Based Subatomic Particle Tracking on FPGAs
Authors: Arjan Blankestijn, Uraz Odyurt, Amirreza Yousefzadeh •
Published: 2025-09-30 •
Source: arXiv
The Transformer Machine Learning (ML) architecture has been gaining considerable momentum in recent years. In particular, computational High-Energy Physics tasks such as jet tagging and particle track reconstruction (tracking), have either achieved proper solutions, or reached considerable milestones using Transformers. On the other hand, the use of specialised hardware accelerators, especially FPGAs, is an effective method to achieve online, or pseudo-online latencies. The development and integration of Transformer-based ML to FPGAs is still ongoing and the support from current tools is very limited to non-existent. Additionally, FPGA resources present a significant constraint. Considering the model size alone, while smaller models can be deployed directly, larger models are to be partitioned in a meaningful and ideally, automated way. We aim to develop methodologies and tools for monolithic, or partitioned Transformer synthesis, specifically targeting inference. Our primary use-case involves two machine learning model designs for tracking, derived from the TrackFormers project. We elaborate our development approach, present preliminary results, and provide comparisons.
33. AI Playing Business Games: Benchmarking Large Language Models on Managerial Decision-Making in Dynamic Simulations
Authors: Berdymyrat Ovezmyradov •
Published: 2025-09-30 •
Source: arXiv
The rapid advancement of LLMs sparked significant interest in their potential to augment or automate managerial functions. One of the most recent trends in AI benchmarking is performance of Large Language Models (LLMs) over longer time horizons. While LLMs excel at tasks involving natural language and pattern recognition, their capabilities in multi-step, strategic business decision-making remain largely unexplored. Few studies demonstrated how results can be different from benchmarks in short-term tasks, as Vending-Bench revealed. Meanwhile, there is a shortage of alternative benchmarks for long-term coherence. This research analyses a novel benchmark using a business game for the decision making in business. The research contributes to the recent literature on AI by proposing a reproducible, open-access management simulator to the research community for LLM benchmarking. This novel framework is used for evaluating the performance of five leading LLMs available in free online interface: Gemini, ChatGPT, Meta AI, Mistral AI, and Grok. LLM makes decisions for a simulated retail company. A dynamic, month-by-month management simulation provides transparently in spreadsheet model as experimental environment. In each of twelve months, the LLMs are provided with a structured prompt containing a full business report from the previous period and are tasked with making key strategic decisions: pricing, order size, marketing budget, hiring, dismissal, loans, training expense, R&D expense, sales forecast, income forecast The methodology is designed to compare the LLMs on quantitative metrics: profit, revenue, and market share, and other KPIs. LLM decisions are analyzed in their strategic coherence, adaptability to market changes, and the rationale provided for their decisions. This approach allows to move beyond simple performance metrics for assessment of the long-term decision-making.
34. Interactive Learning for LLM Reasoning
Authors: Hehai Lin, Shilei Cao, Minzhi Li, Sudong Wang, Haotian Wu, Linyi Yang, Juepeng Zheng, Chengwei Qin •
Published: 2025-09-30 •
Source: arXiv
Existing multi-agent learning approaches have developed interactive training environments to explicitly promote collaboration among multiple Large Language Models (LLMs), thereby constructing stronger multi-agent systems (MAS). However, during inference, they require re-executing the MAS to obtain final solutions, which diverges from human cognition that individuals can enhance their reasoning capabilities through interactions with others and resolve questions independently in the future. To investigate whether multi-agent interaction can enhance LLMs' independent problem-solving ability, we introduce ILR, a novel co-learning framework for MAS that integrates two key components: Dynamic Interaction and Perception Calibration. Specifically, Dynamic Interaction first adaptively selects either cooperative or competitive strategies depending on question difficulty and model ability. LLMs then exchange information through Idea3 (Idea Sharing, Idea Analysis, and Idea Fusion), an innovative interaction paradigm designed to mimic human discussion, before deriving their respective final answers. In Perception Calibration, ILR employs Group Relative Policy Optimization (GRPO) to train LLMs while integrating one LLM's reward distribution characteristics into another's reward function, thereby enhancing the cohesion of multi-agent interactions. We validate ILR on three LLMs across two model families of varying scales, evaluating performance on five mathematical benchmarks and one coding benchmark. Experimental results show that ILR consistently outperforms single-agent learning, yielding an improvement of up to 5% over the strongest baseline. We further discover that Idea3 can enhance the robustness of stronger LLMs during multi-agent inference, and dynamic interaction types can boost multi-agent learning compared to pure cooperative or competitive strategies.
35. Feedback Forensics: A Toolkit to Measure AI Personality
Authors: Arduin Findeis, Timo Kaufmann, Eyke Hüllermeier, Robert Mullins •
Published: 2025-09-30 •
Source: arXiv
Some traits making a "good" AI model are hard to describe upfront. For example, should responses be more polite or more casual? Such traits are sometimes summarized as model character or personality. Without a clear objective, conventional benchmarks based on automatic validation struggle to measure such traits. Evaluation methods using human feedback such as Chatbot Arena have emerged as a popular alternative. These methods infer "better" personality and other desirable traits implicitly by ranking multiple model responses relative to each other. Recent issues with model releases highlight limitations of these existing opaque evaluation approaches: a major model was rolled back over sycophantic personality issues, models were observed overfitting to such feedback-based leaderboards. Despite these known issues, limited public tooling exists to explicitly evaluate model personality. We introduce Feedback Forensics: an open-source toolkit to track AI personality changes, both those encouraged by human (or AI) feedback, and those exhibited across AI models trained and evaluated on such feedback. Leveraging AI annotators, our toolkit enables investigating personality via Python API and browser app. We demonstrate the toolkit's usefulness in two steps: (A) first we analyse the personality traits encouraged in popular human feedback datasets including Chatbot Arena, MultiPref and PRISM; and (B) then use our toolkit to analyse how much popular models exhibit such traits. We release (1) our Feedback Forensics toolkit alongside (2) a web app tracking AI personality in popular models and feedback datasets as well as (3) the underlying annotation data at https://github.com/rdnfn/feedback-forensics.