In controllable image generation, synthesizing coherent and consistent images from multiple reference inputs, i.e., Multi-Image Composition (MICo), remains a challenging problem, partly hindered by the lack of high-quality training data. To bridge this gap, we conduct a systematic study of MICo, categorizing it into 7 representative tasks and curate a large-scale collection of high-quality source images and construct diverse MICo prompts. Leveraging powerful proprietary models, we synthesize a rich amount of balanced composite images, followed by human-in-the-loop filtering and refinement, resulting in MICo-150K, a comprehensive dataset for MICo with identity consistency. We further build a Decomposition-and-Recomposition (De&Re) subset, where 11K real-world complex images are decomposed into components and recomposed, enabling both real and synthetic compositions. To enable comprehensive evaluation, we construct MICo-Bench with 100 cases per task and 300 challenging De&Re cases, and further introduce a new metric, Weighted-Ref-VIEScore, specifically tailored for MICo evaluation. Finally, we fine-tune multiple models on MICo-150K and evaluate them on MICo-Bench. The results
Large-scale image databases remain largely biased towards objects and activities encountered in a select few cultures. This absence of culturally-diverse images, which we refer to as the hidden tail, limits the applicability of pre-trained neural networks and inadvertently excludes researchers from under-represented regions. To begin remedying this issue, we curate Turath-150K, a database of images of the Arab world that reflect objects, activities, and scenarios commonly found there. In the process, we introduce three benchmark databases, Turath Standard, Art, and UNESCO, specialised subsets of the Turath dataset. After demonstrating the limitations of existing networks pre-trained on ImageNet when deployed on such benchmarks, we train and evaluate several networks on the task of image classification. As a consequence of Turath, we hope to engage machine learning researchers in under-represented regions, and to inspire the release of additional culture-focused databases. The database can be accessed here: danikiyasseh.github.io/Turath.
Video quality assessment (VQA) methods focus on particular degradation types, usually artificially induced on a small set of reference videos. Hence, most traditional VQA methods under-perform in-the-wild. Deep learning approaches have had limited success due to the small size and diversity of existing VQA datasets, either artificial or authentically distorted. We introduce a new in-the-wild VQA dataset that is substantially larger and diverse: KonVid-150k. It consists of a coarsely annotated set of 153,841 videos having five quality ratings each, and 1,596 videos with a minimum of 89 ratings each. Additionally, we propose new efficient VQA approaches (MLSP-VQA) relying on multi-level spatially pooled deep-features (MLSP). They are exceptionally well suited for training at scale, compared to deep transfer learning approaches. Our best method, MLSP-VQA-FF, improves the Spearman rank-order correlation coefficient (SRCC) performance metric on the commonly used KoNViD-1k in-the-wild benchmark dataset to 0.82. It surpasses the best existing deep-learning model (0.80 SRCC) and hand-crafted feature-based method (0.78 SRCC). We further investigate how alternative approaches perform under d
Scientific posters are one of the most common forms of scholarly communication and contain early-stage insights with potential to accelerate scientific discovery. We investigated where posters are shared, to what extent their sharing aligns with the FAIR principles, and how commonly they are reused. We identified 86 platforms hosting posters, with many not assigning persistent identifiers. A total of 150k posters are shared as of 2024 on the 43 platforms where we were able to count, which is relatively low. Looking in more detail at posters shared on Zenodo and Figshare, we found that repositories are not always supporting structured metadata critical for poster discovery, like conference information, and that researchers are not providing such metadata even if they are supported. We also observed that while there is some engagement with posters in terms of views and downloads, citing posters is not yet a common practice. Our recommendations are for the scientific community to encourage poster sharing and reuse and establish clear guidelines to make posters FAIR.
The predominant approach for training web navigation agents is to gather human demonstrations for a set of popular websites and hand-written tasks, but it is becoming clear that human data is an inefficient resource. We develop a pipeline to facilitate internet-scale training for agents without laborious human annotations. In the first stage, an LLM annotates 150k sites with agentic tasks. In the next stage, LLM agents complete tasks and produce trajectories. In the final stage, an LLM filters trajectories by judging their success. Language models are powerful data curation tools, identifying harmful content with an accuracy of 97%, judging successful trajectories with an accuracy of 82.6%, and producing effective data. We train agents based on Qwen 3 1.7B that are competitive with frontier LLMs as web agents, while being smaller and faster. Our top agent reaches a success rate of 56.9%, outperforming the data collection policy Qwen 3 235B, a 235 times larger Llama 4 Maverick, and reaching 94.7% of the performance of Gemini 2.5 Flash. We are releasing code, models and data at: https://data-for-agents.github.io.
Recent multimodal models for instruction-based face editing enable semantic manipulation but still struggle with precise attribute control and identity preservation. Structural facial representations such as landmarks are effective for intermediate supervision, yet most existing methods treat them as rigid geometric constraints, which can degrade identity when conditional landmarks deviate significantly from the source (e.g., large expression or pose changes, inaccurate landmark estimates). To address these limitations, we propose LaTo, a landmark-tokenized diffusion transformer for fine-grained, identity-preserving face editing. Our key innovations include: (1) a landmark tokenizer that directly quantizes raw landmark coordinates into discrete facial tokens, obviating the need for dense pixel-wise correspondence; (2) a location-mapped positional encoding and a landmark-aware classifier-free guidance that jointly facilitate flexible yet decoupled interactions among instruction, geometry, and appearance, enabling strong identity preservation; and (3) a landmark predictor that leverages vision-language models to infer target landmarks from instructions and source images, whose struct
Retrieval-augmented generation (RAG) methods are viable solutions for addressing the static memory limits of pre-trained language models. Nevertheless, encountering conflicting sources of information within the retrieval context is an inevitable practical challenge. In such situations, the language models are recommended to transparently inform users about the conflicts rather than autonomously deciding what to present based on their inherent biases. To analyze how current large language models (LLMs) align with our recommendation, we introduce WhoQA, a public benchmark dataset to examine model's behavior in knowledge conflict situations. We induce conflicts by asking about a common property among entities having the same name, resulting in questions with up to 8 distinctive answers. WhoQA evaluation set includes 5K questions across 13 Wikidata property types and 150K Wikipedia entities. Our experiments show that despite the simplicity of WhoQA questions, knowledge conflicts significantly degrades LLMs' performance in RAG settings.
Piano fingering -- knowing which finger to use to play each note in a musical piece, is a hard and important skill to master when learning to play the piano. While some sheet music is available with expert-annotated fingering information, most pieces lack this information, and people often resort to learning the fingering from demonstrations in online videos. We consider the AI task of automating the extraction of fingering information from videos. This is a non-trivial task as fingers are often occluded by other fingers, and it is often not clear from the video which of the keys were pressed, requiring the synchronization of hand position information and knowledge about the notes that were played. We show how to perform this task with high-accuracy using a combination of deep-learning modules, including a GAN-based approach for fine-tuning on out-of-domain data. We extract the fingering information with an f1 score of 97\%. We run the resulting system on 90 videos, resulting in high-quality piano fingering information of 150K notes, the largest available dataset of piano-fingering to date.
Towards building comprehensive real-world visual perception systems, we propose and study a new problem called panoptic scene graph generation (PVSG). PVSG relates to the existing video scene graph generation (VidSGG) problem, which focuses on temporal interactions between humans and objects grounded with bounding boxes in videos. However, the limitation of bounding boxes in detecting non-rigid objects and backgrounds often causes VidSGG to miss key details crucial for comprehensive video understanding. In contrast, PVSG requires nodes in scene graphs to be grounded by more precise, pixel-level segmentation masks, which facilitate holistic scene understanding. To advance research in this new area, we contribute the PVSG dataset, which consists of 400 videos (289 third-person + 111 egocentric videos) with a total of 150K frames labeled with panoptic segmentation masks as well as fine, temporal scene graphs. We also provide a variety of baseline methods and share useful design practices for future work.
Recently, Conformer based CTC/AED model has become a mainstream architecture for ASR. In this paper, based on our prior work, we identify and integrate several approaches to achieve further improvements for ASR tasks, which we denote as multi-loss, multi-path and multi-level, summarized as "3M" model. Specifically, multi-loss refers to the joint CTC/AED loss and multi-path denotes the Mixture-of-Experts(MoE) architecture which can effectively increase the model capacity without remarkably increasing computation cost. Multi-level means that we introduce auxiliary loss at multiple level of a deep model to help training. We evaluate our proposed method on the public WenetSpeech dataset and experimental results show that the proposed method provides 12.2%-17.6% relative CER improvement over the baseline model trained by Wenet toolkit. On our large scale dataset of 150k hours corpus, the 3M model has also shown obvious superiority over the baseline Conformer model. Code is publicly available at https://github.com/tencent-ailab/3m-asr.
We report successful growth of magnetic topological insulator (MTI) MnBi2Te4. The heating schedule basically deals with growth of the crystal from melt at 900C and very slow cooling (1C/hr) to around 600C with 24 hours hold time, followed by cooling to room temperature. Our detailed, PXRD Reitveld analysis showed that the resultant crystal is dominated mainly by MnBi2Te4 and minor phases of Bi2Te3 and MnTe. The transport measurements showed a step like behavior at around 150K followed by cusp like structure in resistivity at around 25K (TP) due reported anti-ferromagnetic ordering of Mn. Both the resistivity transitions are seen clearly in dR/dT measurements at 150K and 20K respectively. The 25K transition of the compound is also seen in magnetic susceptibility. Low temperature (5K) magnetoresistance (MR) in applied field of up to 6 Tesla exhibited negative ve MR below 3 Tesla and +ve for higher fields. Also, seen are steps in MR below one Tesla. The studied MnBi2Te4 MTI crystal could be a possible candidate for Quantum Anomalous Hall (QAH) effect.
Two B-site ordered double perovskites, La2LiMoO6 and Ba2YMoO6, based on the S = 1/2 ion, Mo5+, have been investigated in the context of geometric magnetic frustration. Powder neutron diffraction, heat capacity, susceptibility, muon spin relaxation(_SR), and 89Y NMR- including MAS NMR- data have been collected. La2LiMoO6 deviates strongly from simple Curie-Weiss paramagnetic behavior below 150K and zero-field cooled/ field cooled (ZFC/FC)irreversibility occurs below 20K with a weak, broad susceptibility maximum near 5K in the ZFC data. A Curie-Weiss fit shows a reduced mu_eff=1.42μ_B, (spin only = 1.73 muB) and a Weiss temperature, θ_c, which depends strongly on the temperature range of the fit. Powder neutron diffraction, heat capacity and 7Li NMR show no evidence for long range magnetic order to 2K. On the other hand oscillations develop below 20K in muSR indicating at least short range magnetic correlations. Susceptibility data for Ba2YMoO6 also deviate strongly from the C-W law below 150K with a similarly reduced mu_eff = 1.72μ_B and θ_c = - 219(1)K. Heat capacity, neutron powder diffraction and muSR data show no evidence for long range order to 2K but a very broad maximum appea
Pulsed-laser deposition has been used to grow epitaxial thin films of the giant-dielectric-constant material CaCu_3Ti_4O_{12} on LaAlO_3 and SrTiO_3 substrates with or without various conducting buffer layers. The latter include YBa_2Cu_3O_7, La_{1.85}Sr_{0.15}CuO_{4+δ} and LaNiO_3. Above 100K - 150K the thin films have a temperature independent dielectric constant as do single crystals. The value of the dielectric constant is of the order of 1500 over a wide temperature region, potentially making it a good candidate for many applications. The frequency dependence of its dielectric properties below 100K - 150K indicates an activated relaxation process.
Individual metallic single-wall carbon nanotubes show unsual non-Ohmic transport behaviors at high bias fields. For low resistance contact samples, the differential conductance dI/dV increases with increasing bias, reaching a maximum at $\sim$ 100mV. As the bias increases further, dI/dV drops dramatically [Yao et al., Phys. Rev. Lett. 84, 2941 (2000)]. The higher the bias, the system behaves in a more normal (Ohmic) manner. This so-called zero-bias anomaly is temperature-dependent (50--150K). We propose a new interpretation. Supercurrent runs in the graphene wall below $\sim$ 150K. The normal conduction-electron currents run outside the wall, which are subject to the scattering by phonons and impurities. The currents along the tube induce circulating magnetic fields and eventually destroy the supercurrent in the wall at high enough bias, and restore the Ohmic behavior. If the prevalent ballistic electron model is adopted, then the scattering effects cannot be discussed.
Magneto-optic studies of ZnO doped with transition metals Co, Mn, V and Ti indicate a significant magnetic circular dichroism (MCD) at the ZnO band edge at room temperature, together with an associated dispersive Faraday rotation. Similar spectra occur for each dopant which implies that the ferromagnetism is an intrinsic property of the bulk ZnO lattice. At 10 K additional paramagnetic contributions to the MCD are observed, but above about 150K, the magnitude of the MCD signal is dominated by the ferromagnetism and is almost temperature independent. The MCD at the ZnO band edge shows room temperature hysteretic behaviour.
3D recovery from multi-stereo and stereo images, as an important application of the image-based perspective geometry, serves many applications in computer vision, remote sensing and Geomatics. In this chapter, the authors utilize the imaging geometry and present approaches that perform 3D reconstruction from cross-view images that are drastically different in their viewpoints. We introduce our framework that takes ground-view images and satellite images for full 3D recovery, which includes necessary methods in satellite and ground-based point cloud generation from images, 3D data co-registration, fusion and mesh generation. We demonstrate our proposed framework on a dataset consisting of twelve satellite images and 150k video frames acquired through a vehicle-mounted Go-pro camera and demonstrate the reconstruction results. We have also compared our results with results generated from an intuitive processing pipeline that involves typical geo-registration and meshing methods.
Compton scattering provides invaluable information on the underlying Fermi surface (FS) and is a powerful tool complementary to angle-resolved photoemission spectroscopy and quantum oscillation measurements. Here we perform high-resolution Compton scattering measurements for La_{2-x}Sr_{x}CuO_{4} with x=0.08 (Tc=20K) at 300K and 150K, and image the momentum distribution function in the two-dimensional Brillouin zone. We find that the observed images cannot be reconciled with the conventional hole-like FS believed so far. Instead, our data imply that the FS is strongly deformed by the underlying nematicity in each CuO_2 plane, but the bulk FSs recover the fourfold symmetry. We also find an unusually strong temperature dependence of the momentum distribution function, which may originate from the pseudogap formation in the presence of the reconstructed FSs due to the underlying nematicity. Additional measurements for x=0.15 and 0.30 at 300K suggest similar FS deformation with weaker nematicity, which nearly vanishes at x=0.30.
We investigated the spin dynamics of two-dimensional electrons in (001) GaAs/AlGaAs heterostructure using the time resolved Kerr rotation technique under a transverse magnetic field. The in-plane spin lifetime is found to be anisotropic below 150k due to the interference of Rashba and Dresselhaus spin-orbit coupling and D'yakonov-Perel' spin relaxation. The ratio of in-plane spin lifetimes is measured directly as a function of temperature and pump power, showing that the electron density in 2DEG channel strongly affects the Rashba spin-orbit coupling.
We report experimental observation of 2s exciton radiative emission from monolayer tungsten diselenide, enabled by hexagonal boron nitride protected high-quality samples. The 2s luminescence is highly robust and persists up to 150K, offering a new quantum entity for manipulating the valley degree of freedom. Remarkably, the 2s exciton displays superior valley polarization and coherence than 1s under similar experimental conditions. This observation provides evidence that the Coulomb-exchange-interaction-driven valley-depolarization process, the Maialle-Silva-Sham mechanism, plays an important role in valley excitons of monolayer transition metal dichalcogenides.
We report a new measurement of Rb=Gamma(Z -> bb)/ Gamma(Z -> hadrons) using a double tag technique where the b selection is based on topological reconstruction of the mass of the B-decay vertex. The measurement was performed using a sample of 150k hadronic Z0 events collected with the SLD at the SLAC Linear Collider during the years 1993-1995. The method utilizes the 3-D vertexing abilities of the SLD CCD pixel vertex detector and the small stable SLC beams to obtain a high b tagging efficiency of 37% for a purity of 97.2%. The high purity reduces the systematics introduced by charm contamination and correlations with Rc. We obtain a result of Rb=0.2176+/-0.0033(stat.)+/-0.0017(syst.)+/-0.0008(Rc). (Presented at the XXXIst Rencontres de Moriond Electroweak Interactions and Unified Theories, Les Arcs, Savoie, France, March 16-23, 1996.)