共找到 20 条结果
In this work, we propose a simple and computationally efficient framework for evaluating whether machine learning models align with the structure of the data they learn from; that is, whether the model says what the data says. Unlike existing interpretability methods that focus exclusively on explaining model behavior, our approach establishes a baseline derived directly from the data itself. Drawing inspiration from Rubin's Potential Outcomes Framework, we quantify how strongly each feature separates the two outcome groups in a binary classification task, moving beyond traditional descriptive statistics to estimate each feature's effect on the outcome. By comparing these data-derived feature rankings with model-based explanations, we provide practitioners with an interpretable and model-agnostic method for assessing model-data alignment.
I expound and defend the ``bare probabilism'' reading of Gibbsian (i.e. mainstream) statistical mechanics, responding to Frigg and Werndl's recent (BJPS 72 (2021), 105-129) plea: ``can somebody please say what Gibbsian statistical mechanics says?''
Logic can define how agents are provided or denied access to resources, how to interlink resources using mining processes and provide users with choices for possible next steps in a workflow. These decisions are for the most part hidden, internal to machines processing data. In order to exchange this internal logic a portable Web logic is required which the Semantic Web could provide. Combining logic and data provides insights into the reasoning process and creates a new level of trust on the Semantic Web. Current Web logics carries only a fragment of first-order logic (FOL) to keep exchange languages decidable or easily processable. But, this is at a cost: the portability of logic. Machines require implicit agreements to know which fragment of logic is being exchanged and need a strategy for how to cope with the different fragments. These choices could obscure insights into the reasoning process. We created RDF Surfaces in order to express the full expressivity of FOL including saying explicitly `no'. This vision paper provides basic principles and compares existing work. Even though support for FOL is semi-decidable, we argue these problems are surmountable. RDF Surfaces span man
Gibbsian statistical mechanics (GSM) is the most widely used version of statistical mechanics among working physicists. Yet a closer look at GSM reveals that it is unclear what the theory actually says and how it bears on experimental practice. The root cause of the difficulties is the status of the Averaging Principle, the proposition that what we observe in an experiment is the ensemble average of a phase function. We review different stances toward this principle, and eventually present a coherent interpretation of GSM that provides an account of the status and scope of the principle.
During 2015 and early 2016, the cultural application of Computational Creativity research and practice took a big leap forward, with a project where multiple computational systems were used to provide advice and material for a new musical theatre production. Billed as the world's first 'computer musical... conceived by computer and substantially crafted by computer', Beyond The Fence was staged in the Arts Theatre in London's West End during February and March of 2016. Various computational approaches to analytical and generative sub-projects were used to bring about the musical, and these efforts were recorded in two 1-hour documentary films made by Wingspan Productions, which were aired on SkyArts under the title Computer Says Show. We provide details here of the project conception and execution, including details of the systems which took on some of the creative responsibility in writing the musical, and the contributions they made. We also provide details of the impact of the project, including a perspective from the two (human) writers with overall control of the creative aspects the musical.
Its WeatherNext model, which will be open-sourced, can accurately predict both a storm’s track and intensity using lower-resolution weather data。 Researchers don't yet fully understand how it does this
Language models can state that "the Earth orbits the Sun" and, when role-playing Aristotle, assert the opposite. Recent work argues that persona adoption is fundamental to how language models behave, with models selecting the most appropriate persona for a given context. Does such role-playing merely change the model's outputs, or does it also affect what the model internally represents as truthful? We study this question using the role-play of characters whose beliefs differ from the modern consensus, and induce personas with a number of different methods: prompting, in-context learning (ICL), supervised fine-tuning (SFT), and Open Character Training (OCT), and Emergent Misalignment (EM). We measure belief internalization across these approaches with truth probes and with behavioral tests, finding a broad spectrum of belief internalization. Prompting, ICL, and SFT change what the model says with little representational change. EM creates a large, broad shift in the model's truth representation, and OCT a smaller shift that is clearest on the larger model. Understanding when training changes a model's worldview rather than merely its behavior may become increasingly important as AI
Generative models are increasingly used to improve the quality of medical imaging, such as reconstruction of magnetic resonance images and computed tomography. However, it is well-known that such models are susceptible to hallucinations: they may insert features into the reconstructed image which are not actually present in the original image. In a medical setting, such hallucinations may endanger patient health as they can lead to incorrect diagnoses. In this work, we aim to quantify the extent to which state-of-the-art generative models suffer from hallucinations in the context of magnetic resonance image reconstruction. Specifically, we craft adversarial perturbations resembling random noise for the unprocessed input images which induce hallucinations when reconstructed using a generative model. We perform this evaluation on the brain and knee images from the fastMRI data set using UNet and end-to-end VarNet architectures to reconstruct the images. Our results show that these models are highly susceptible to small perturbations and can be easily coaxed into producing hallucinations. This fragility may partially explain why hallucinations occur in the first place and suggests tha
Stratospheric aerosol injection (SAI) is a solar radiation modification technique, proposed as an interim measure to offset warming while greenhouse gas (GHG) emissions are reduced. This paper discusses a possible SAI implementation route - an alternative to sulfate aerosols formed in situ - based on engineered solid particles having dedicated properties such as size, composition, surface chemistry, and traceable origin, supporting safety, controllability, and functionality needed for SAI systems. These engineered properties also open up options for any future multi-state coordination of SAI through two technical building blocks: (1) the SAI-induced radiative forcing (SRF) - the magnitude of the cooling effect attributable specifically to the SAI layer - as an operator-independent quantity, derivable from direct aerosol-layer measurements; and (2) particle traceability through identifying signatures embedded at production. Both could feed into a shared, publicly accessible monitoring database open to independent interrogation, addressing several governance challenges by anchoring compliance assessments in measurable parameters. Drawing on precedents from the Montreal Protocol, IAEA
Autonomous drone delivery systems are rapidly advancing, but ensuring safe and reliable package drop-offs remains highly challenging in cluttered urban and suburban environments where accurately identifying suitable package drop zones is critical. Existing approaches typically rely on either geometry-based analysis or semantic segmentation alone, but these methods lack the integrated semantic reasoning required for robust decision-making. To address this gap, we propose See&Say, a novel framework that combines geometric safety cues with semantic perception, guided by a Vision-Language Model (VLM) for iterative refinement. The system fuses monocular depth gradients with open-vocabulary detection masks to produce safety maps, while the VLM dynamically adjusts object category prompts and refines hazard detection across time, enabling reliable reasoning under dynamic conditions during the final delivery phase. When the primary drop-pad is occupied or unsafe, the proposed See&Say also identifies alternative candidate zones for package delivery. We curated a dataset of urban delivery scenarios with moving objects and human activities to evaluate the approach. Experimental results
This study comprehensively analyzes three open star clusters: SAI 16, SAI 81, and SAI 86 using Gaia DR3 data. Based on the ASteCA code, we determined the most probable star candidates (P >= 50%) and estimated the number of star members of each cluster as 125, 158, and 138, respectively. We estimated the internal structural parameters by fitting the King model to the observed RDPs, including the core, limited, and tidal radii. The isochrone fitting to the color-magnitude diagram provided log(age) values of 9.13 +/- 0.04, 8.10 +/- 0.04, and 8.65 +/- 0.04 and distances of 3790 +/- 94 pc, 3900 +/- 200 pc, and 3120 +/- 30 pc for SAI 16, SAI 81, and SAI 86, respectively. We also calculated their projected distances from the Galactic plane (X_sun, Y_sun) as well as their vertical distances (Z_sun), Galactocentric distances (R_gc), and total masses (M_C) in solar units, which are about 142 +/- 12, 302 +/- 17, and 192 +/- 14 for SAI 16, SAI 81, and SAI 86, respectively. Examining the dynamical relaxation state indicates that all three clusters are dynamically relaxed. By undertaking a kinematic analysis of the cluster data, the space velocity was determined. We calculated the coordinates
This study investigates the open clusters SAI 72 and SAI 75 using Gaia DR3 data, employing the Automated Stellar Cluster Analysis (ASteCA) tool to determine their structural and fundamental properties, including center coordinates, size, age, distance, mass, luminosity, and kinematics. Based on membership probabilities (P >= 50%), we identified 112 and 115 stars as probable members of SAI 72 and SAI 75, respectively. Radial density profile (RDP) analysis yielded cluster radii of 2.35 arcmin for SAI 72 and 2.19 arcmin for SAI 75. The spectral energy distribution (SED) fitting was performed to refine metallicity, distance, and color excess parameters, ensuring consistency within 1 sigma of isochrone-based estimates. Isochrone fitting of the color-magnitude diagram (CMD) suggests ages of 316 Myr and 302 Myr, with corresponding distances of 3160 +/- 80 pc and 3200 +/- 200 pc. We derived their Galactic positions, projected distances (X_sun, Y_sun), and vertical displacements (Z_sun). Mass function analysis estimates cluster masses of 612 +/- 174 solar masses for SAI 72 and 465 +/- 90 solar masses for SAI 75. Kinematic studies indicate that both clusters have reached dynamical equilib
This paper introduces feature subset weighting using monotone measures for distance-based supervised learning. The Choquet integral is used to define a distance metric that incorporates these weights. This integration enables the proposed distances to effectively capture non-linear relationships and account for interactions both between conditional and decision attributes and among conditional attributes themselves, resulting in a more flexible distance measure. In particular, we show how this approach ensures that the distances remain unaffected by the addition of duplicate and strongly correlated features. Another key point of this approach is that it makes feature subset weighting computationally feasible, since only $m$ feature subset weights should be calculated each time instead of calculating all feature subset weights ($2^m$), where $m$ is the number of attributes. Next, we also examine how the use of the Choquet integral for measuring similarity leads to a non-equivalent definition of distance. The relationship between distance and similarity is further explored through dual measures. Additionally, symmetric Choquet distances and similarities are proposed, preserving the c
Ensuring the safety alignment of Large Language Models (LLMs) is critical for generating responses consistent with human values. However, LLMs remain vulnerable to jailbreaking attacks, where carefully crafted prompts manipulate them into producing toxic content. One category of such attacks reformulates the task as an optimization problem, aiming to elicit affirmative responses from the LLM. However, these methods heavily rely on predefined objectionable behaviors, limiting their effectiveness and adaptability to diverse harmful queries. In this study, we first identify why the vanilla target loss is suboptimal and then propose enhancements to the loss objective. We introduce DSN (Don't Say No) attack, which combines a cosine decay schedule method with refusal suppression to achieve higher success rates. Extensive experiments demonstrate that DSN outperforms baseline attacks and achieves state-of-the-art attack success rates (ASR). DSN also shows strong universality and transferability to unseen datasets and black-box models.
This paper presents TRACE, a tool to analyze the quality of 2D embeddings generated through dimensionality reduction techniques. Dimensionality reduction methods often prioritize preserving either local neighborhoods or global distances, but insights from visual structures can be misleading if the objective has not been achieved uniformly. TRACE addresses this challenge by providing a scalable and extensible pipeline for computing both local and global quality measures. The interactive browser-based interface allows users to explore various embeddings while visually assessing the pointwise embedding quality. The interface also facilitates in-depth analysis by highlighting high-dimensional nearest neighbors for any group of points and displaying high-dimensional distances between points. TRACE enables analysts to make informed decisions regarding the most suitable dimensionality reduction method for their specific use case, by showing the degree and location where structure is preserved in the reduced space.
We propose a novel type of Artificial Immune System (AIS): Symbiotic Artificial Immune Systems (SAIS), drawing inspiration from symbiotic relationships in biology. SAIS parallels the three key stages (i.e., mutualism, commensalism and parasitism) of population updating from the Symbiotic Organisms Search (SOS) algorithm. This parallel approach effectively addresses the challenges of large population size and enhances population diversity in AIS, which traditional AIS and SOS struggle to resolve efficiently. We conducted a series of experiments, which demonstrated that our SAIS achieved comparable performance to the state-of-the-art approach SOS and outperformed other popular AIS approaches and evolutionary algorithms across 26 benchmark problems. Furthermore, we investigated the problem of parameter selection and found that SAIS performs better in handling larger population sizes while requiring fewer generations. Finally, we believe SAIS, as a novel bio-inspired and immune-inspired algorithm, paves the way for innovation in bio-inspired computing with the symbiotic paradigm.
Generating stylized talking head with diverse head motions is crucial for achieving natural-looking videos but still remains challenging. Previous works either adopt a regressive method to capture the speaking style, resulting in a coarse style that is averaged across all training data, or employ a universal network to synthesize videos with different styles which causes suboptimal performance. To address these, we propose a novel dynamic-weight method, namely Say Anything withAny Style (SAAS), which queries the discrete style representation via a generative model with a learned style codebook. Specifically, we develop a multi-task VQ-VAE that incorporates three closely related tasks to learn a style codebook as a prior for style extraction. This discrete prior, along with the generative model, enhances the precision and robustness when extracting the speaking styles of the given style clips. By utilizing the extracted style, a residual architecture comprising a canonical branch and style-specific branch is employed to predict the mouth shapes conditioned on any driving audio while transferring the speaking style from the source to any desired one. To adapt to different speaking st