In this paper we describe Ninja Codes, neurally generated fiducial markers that can be made to naturally blend into various real-world environments. An encoder network converts arbitrary images into Ninja Codes by applying visually modest alterations; the resulting codes, printed and pasted onto surfaces, can provide stealthy 6-DoF location tracking for a wide range of applications including robotics and augmented reality. Ninja Codes can be printed using standard color printers on regular printing paper, and can be detected using any device equipped with a modern RGB camera and capable of running inference. Through experiments, we demonstrate Ninja Codes' ability to provide reliable location tracking under common indoor lighting conditions, while successfully concealing themselves within diverse environmental textures. We expect Ninja Codes to offer particular value in scenarios where the conspicuous appearance of conventional fiducial markers makes them undesirable for aesthetic and other reasons.
Exo-NINJA will realize nearIR R$\sim$4000 diffraction-limited narrow-field spectro-imaging for characterization of exoplanets and circumstellar disk structures. It uniquely combines mid-R spectroscopy, high throughput, and spatial resolution, in contrast to CHARIS, which does spectro-imaging, and REACH, which is single-point (no spatial resolution). Exo-NINJA's spectro-imaging at the telescope diffraction limit will characterize exoplanet atmospheres, detect and map (spatially and spectrally) gas accretion on protoplanets, and also detect exoplanets at small angular separation ($λ$/D) from their host star by spectro-astrometry. Exo-NINJA will link two instruments at the Subaru Telescope using a high-throughput hexagonal multi-mode fiber bundle (hexabundle). The fiber coupling resides between the high contrast imaging system SCExAO, which combines ExAO and coronagraph, and the medium-resolution spectrograph NINJA (R$=$4000 at JHK bands). Exo-NINJA will provide an end-to-end throughput of 20% compared to the 1.5% obtained with REACH. Exo-NINJA is scheduled for implementation on the Subaru Telescope's NasIR platform in 2025; we will present a concise overview of its future installatio
The NINJA experiment aims to precisely measure neutrino-nucleus interactions using a nuclear emulsion detector to reduce systematic errors in neutrino oscillation experiments. The nuclear emulsion has a sub-micron positional resolution, enabling the detection of low-momentum charged particles such as protons with a threshold of 200 MeV/c. In the NINJA experiment, a muon detector placed downstream of the emulsion detector is used to identify muons from $ν_μ$ charged-current interactions. The majority of the tracks accumulated in the nuclear emulsion are from cosmic rays. Although the emulsion detector provides highly accurate positional information, it lacks timing information. Therefore, the positional resolution of the muon detector is not enough to identify neutrino interaction tracks that match between the muon detector and the emulsion detector from the enormous background of cosmic rays recorded in the emulsion detector. To address this, a scintillation tracker is used to provide both timing and positional information for the tracks. The NINJA experiment is planning a third physics run with about 130 kg water target from the autumn of 2025 to the spring of 2026. Since the targ
We present the NINJA suite of cosmological hydrodynamical simulations developed to investigate galaxy formation and evolution at $z \gtrsim 5$ in the era of JWST. Using our fiducial simulation, we explore a range of spectral synthesis prescriptions and dust attenuation models, demonstrating that suitably chosen parameters can reproduce the observed UV luminosity functions (UVLFs) over $5 \leq z \leq 10$. In all cases, the inferred dust-to-metal ratio evolves with redshift, although its normalization at fixed redshift varies by a factor of $\sim 7$, depending on the adopted dust--metallicity scaling and attenuation curve. These model variations introduce substantial scatter in predictions for the $B$-band luminosity function, the H$α$ luminosity function, the UV slope--UV magnitude relation, the stellar mass--Balmer ratio relation, and the relation between stellar and nebular colour excesses. Simultaneously reproducing these observables across multiple redshifts will therefore be essential for constraining dust models at high redshift with forthcoming observations. Observations of galaxies spanning a broad range of stellar masses with the Atacama Large Millimeter/submillimeter Array
Recent research shows that visualizing linguistic bias mitigates its negative effects. However, reliable automatic detection methods to generate such visualizations require costly, knowledge-intensive training data. To facilitate data collection for media bias datasets, we present News Ninja, a game employing data-collecting game mechanics to generate a crowdsourced dataset. Before annotating sentences, players are educated on media bias via a tutorial. Our findings show that datasets gathered with crowdsourced workers trained on News Ninja can reach significantly higher inter-annotator agreements than expert and crowdsourced datasets with similar data quality. As News Ninja encourages continuous play, it allows datasets to adapt to the reception and contextualization of news over time, presenting a promising strategy to reduce data collection expenses, educate players, and promote long-term bias mitigation.
Population protocols are a model of distributed computation in which an arbitrary number of indistinguishable finite-state agents interact in pairs to decide some property of their initial configuration. We investigate the behaviour of population protocols under adversarial faults that cause agents to silently crash and no longer interact with other agents. As a starting point, we consider the property ``the number of agents exceeds a given threshold $t$'', represented by the predicate $x \geq t$, and show that the standard protocol for $x \geq t$ is very fragile: one single crash in a computation with $x:=2t-1$ agents can already cause the protocol to answer incorrectly that $x \geq t$ does not hold. However, a slightly less known protocol is robust: for any number $t' \geq t$ of agents, at least $t' - t+1$ crashes must occur for the protocol to answer that the property does not hold. We formally define robustness for arbitrary population protocols, and investigate the question whether every predicate computable by population protocols has a robust protocol. Angluin et al. proved in 2007 that population protocols decide exactly the Presburger predicates, which can be represented a
The Numerical INJection Analysis (NINJA) project is a collaborative effort between members of the numerical relativity and gravitational wave data analysis communities. The purpose of NINJA is to study the sensitivity of existing gravitational-wave search and parameter-estimation algorithms using numerically generated waveforms, and to foster closer collaboration between the numerical relativity and data analysis communities. The first NINJA project used only a small number of injections of short numerical-relativity waveforms, which limited its ability to draw quantitative conclusions. The goal of the NINJA-2 project is to overcome these limitations with long post-Newtonian - numerical relativity hybrid waveforms, large numbers of injections, and the use of real detector data. We report on the submission requirements for the NINJA-2 project and the construction of the waveform catalog. Eight numerical relativity groups have contributed 63 hybrid waveforms consisting of a numerical portion modelling the late inspiral, merger, and ringdown stitched to a post-Newtonian portion modelling the early inspiral. We summarize the techniques used by each group in constructing their submissio
The Numerical INJection Analysis (NINJA) project is a collaborative effort between members of the numerical relativity and gravitational-wave astrophysics communities. The purpose of NINJA is to study the ability to detect gravitational waves emitted from merging binary black holes and recover their parameters with next-generation gravitational-wave observatories. We report here on the results of the second NINJA project, NINJA-2, which employs 60 complete binary black hole hybrid waveforms consisting of a numerical portion modelling the late inspiral, merger, and ringdown stitched to a post-Newtonian portion modelling the early inspiral. In a "blind injection challenge" similar to that conducted in recent LIGO and Virgo science runs, we added 7 hybrid waveforms to two months of data recolored to predictions of Advanced LIGO and Advanced Virgo sensitivity curves during their first observing runs. The resulting data was analyzed by gravitational-wave detection algorithms and 6 of the waveforms were recovered with false alarm rates smaller than 1 in a thousand years. Parameter estimation algorithms were run on each of these waveforms to explore the ability to constrain the masses, co
The Numerical INJection Analysis (NINJA) project is a collaborative effort between members of the numerical relativity and gravitational-wave data analysis communities. The purpose of NINJA is to study the sensitivity of existing gravitational-wave search algorithms using numerically generated waveforms and to foster closer collaboration between the numerical relativity and data analysis communities. We describe the results of the first NINJA analysis which focused on gravitational waveforms from binary black hole coalescence. Ten numerical relativity groups contributed numerical data which were used to generate a set of gravitational-wave signals. These signals were injected into a simulated data set, designed to mimic the response of the Initial LIGO and Virgo gravitational-wave detectors. Nine groups analysed this data using search and parameter-estimation pipelines. Matched filter algorithms, un-modelled-burst searches and Bayesian parameter-estimation and model-selection algorithms were applied to the data. We report the efficiency of these search methods in detecting the numerical waveforms and measuring their parameters. We describe preliminary comparisons between the differ
Our society is increasingly fond of computational tools. This phenomenon has greatly increased over the past decade following, among other factors, the emergence of a new Artificial Intelligence paradigm. Specifically, the coupling of two algorithmic techniques, Deep Neural Networks and Stochastic Gradient Descent, thrusted by an exponentially increasing computing capacity, has and is continuing to become a major asset in many modern technologies. However, as progress takes its course, some still wonder whether other methods could similarly or even more greatly benefit from these various hardware advances. In order to further this study, we delve in this thesis into Evolutionary Algorithms and their application to Dynamic Neural Networks, two techniques which despite enjoying many advantageous properties have yet to find their niche in contemporary Artificial Intelligence. We find that by elaborating new methods while exploiting strong computational resources, it becomes possible to develop strongly performing agents on a variety of benchmarks but also some other agents behaving very similarly to human subjects on the video game Shinobi III : Return of The Ninja Master, typical com
The gravitational wave signature from binary black hole coalescences is an important target for LIGO and VIRGO. The Numerical INJection Analysis (NINJA) project brought together the numerical relativity and gravitational wave data analysis communities, with the goal to optimize the detectability of these events. In its first instantiation, the NINJA project produced a simulated data set with numerical waveforms from binary black hole coalescences of various morphologies (spin, mass ratio, initial conditions), superimposed to Gaussian colored noise at the design sensitivity for initial LIGO and VIRGO. We analyzed this simulated data set with the Q-pipeline burst algorithm. This code, designed for the all-sky detection of gravitational wave bursts with minimal assumptions on the shape of the waveform, filters the data with a bank of sine-Gaussians, or sinusoids with Gaussian envelope. The algorithm's performance was compared to matched filtering with ring-down templates. The results are qualitatively consistent; however due to the low simulation statistics in the first NINJA project, it is premature to draw quantitative conclusions at this stage.
The 2008 NRDA conference introduced the Numerical INJection Analysis project (NINJA), a new collaborative effort between the numerical relativity community and the data analysis community. NINJA focuses on modeling and searching for gravitational wave signatures from the coalescence of binary system of compact objects. We review the scope of this collaboration and the components of the first NINJA project, where numerical relativity groups shared waveforms and data analysis teams applied various techniques to detect them when embedded in colored Gaussian noise.
The Ninja data analysis challenge allowed the study of the sensitivity of data analysis pipelines to binary black hole numerical relativity waveforms in simulated Gaussian noise at the design level of the LIGO observatory and the VIRGO observatory. We analyzed NINJA data with a pipeline based on the Hilbert Huang Transform, utilizing a detection stage and a characterization stage: detection is performed by triggering on excess instantaneous power, characterization is performed by displaying the kernel density enhanced (KD) time-frequency trace of the signal. Using the simulated data based on the two LIGO detectors, we were able to detect 77 signals out of 126 above SNR 5 in coincidence, with 43 missed events characterized by signal to noise ratio SNR less than 10. Characterization of the detected signals revealed the merger part of the waveform in high time and frequency resolution, free from time-frequency uncertainty. We estimated the timelag of the signals between the detectors based on the optimal overlap of the individual KD time-frequency maps, yielding estimates accurate within a fraction of a millisecond for half of the events. A coherent addition of the data sets according
Can drivers' situation awareness during automated driving be maintained using haptic cues that provide information about road and traffic scenarios while the drivers are engaged in a secondary task? And can this be done without disengaging them from the secondary task? Multiple Resource Theory predicts that using different sensory channels can improve multiple-task performance. Using haptics to provide information avoids the audio-visual channels likely occupied by the secondary task. An experiment was conducted to assess whether drivers' situation awareness could be maintained using haptic cues. Drivers played Fruit Ninja as the secondary task while seated in a driving simulator with a Level 4 autonomous system driving. A mixed design was used for the experiment with the presence of haptic cues and the presentation time of situation awareness questions as the between-subjects conditions. Five road and traffic scenarios comprised the within-subjects part of the design. Subjects who received haptic cues had a higher number of correct responses to the situation awareness questions and looked up at the simulator screen fewer times than those who were not provided cues. Subjects did no
This paper presents SPIROS (Streamlined, Precise, Intuitive, and Rapid Optical Simulator), a dedicated optical simulation tool developed for the design and analysis of particle physics detectors. Unlike general-purpose frameworks such as GEANT4, SPIROS offers a lightweight simulation engine and a user-friendly interface optimized for optical processes, including scintillation, Cherenkov emission, and photon transport with reflection, refraction, scattering, absorption, and detection. Detector geometries can be directly imported from 3D CAD models, and all configurations including materials, surfaces, sources, and sensors are specified via a single human-readable input file. Validation against GEANT4 shows excellent agreement in photon generation and propagation behaviors, while benchmark tests demonstrate that SPIROS runs more than two times faster for typical detector configurations. The software has already been applied to multiple neutrino experiments, including T2K, NINJA, and AXEL, for detector design, performance studies, and optimization. SPIROS is open-source and freely available at https://github.com/tkikawa/spiros.
Recent advances in long-context language models (LMs) have enabled million-token inputs, expanding their capabilities across complex tasks like computer-use agents. Yet, the safety implications of these extended contexts remain unclear. To bridge this gap, we introduce NINJA (short for Needle-in-haystack jailbreak attack), a method that jailbreaks aligned LMs by appending benign, model-generated content to harmful user goals. Critical to our method is the observation that the position of harmful goals play an important role in safety. Experiments on standard safety benchmark, HarmBench, show that NINJA significantly increases attack success rates across state-of-the-art open and proprietary models, including LLaMA, Qwen, Mistral, and Gemini. Unlike prior jailbreaking methods, our approach is low-resource, transferable, and less detectable. Moreover, we show that NINJA is compute-optimal -- under a fixed compute budget, increasing context length can outperform increasing the number of trials in best-of-N jailbreak. These findings reveal that even benign long contexts -- when crafted with careful goal positioning -- introduce fundamental vulnerabilities in modern LMs.
Gravitational waves (GWs) from massive black hole (MBH) mergers will provide a novel way to probe the high-redshift universe and are key to understanding galactic dynamics and evolution. In this work, we analyze MBH mergers, their GW signals and detectability, as well as their population properties, using the cosmological hydrodynamical simulation - NINJA Simulation Suite. We discuss the effect of resolution and finite volume on the black hole mass function (BHMF), which in turn limits the mergers associated with low mass black holes, $M_{BH} \lesssim 10^{6.5} M_\odot$. We find the upper limit on the total mass of the MBH binaries detectable by LISA to be $\sim 10^{8.4} M_\odot$. We also find that adding time delays pertaining to dissipative processes like dynamical friction and stellar hardening during the final stages of the inspiral for which the simulation lacks sufficient resolution to model, considerably shifts the peak of redshift distribution of detectable binaries from $z\sim0.5$ to $z\sim0.1$. Time delays reduce the number of detectable GW events but on the other hand their signal-to-noise is increased. From the observational point of view, we find a strong correlation be
In response to the critical need for effective reconnaissance in disaster scenarios, this research article presents the design and implementation of a complete autonomous robot system using the Turtlebot3 with Robotic Operating System (ROS) Noetic. Upon deployment in closed, initially unknown environments, the system aims to generate a comprehensive map and identify any present 'victims' using AprilTags as stand-ins. We discuss our solution for search and rescue missions, while additionally exploring more advanced algorithms to improve search and rescue functionalities. We introduce a Cubature Kalman Filter to help reduce the mean squared error [m] for AprilTag localization and an information-theoretic exploration algorithm to expedite exploration in unknown environments. Just like turtles, our system takes it slow and steady, but when it's time to save the day, it moves at ninja-like speed! Despite Donatello's shell, he's no slowpoke - he zips through obstacles with the agility of a teenage mutant ninja turtle. So, hang on tight to your shells and get ready for a whirlwind of reconnaissance! Full pipeline code https://github.com/rzhao5659/MRProject/tree/main Exploration code https
This work proposes a compilation flow using open-source compiler passes to build a framework to achieve ninja performance from a generic linear algebra high-level abstraction. We demonstrate this flow with a proof-of-concept MLIR project that uses input IR in Linalg-on-Tensor from TensorFlow and PyTorch, performs cache-level optimizations and lowering to micro-kernels for efficient vectorization, achieving over 90% of the performance of ninja-written equivalent programs. The contributions of this work include: (1) Packing primitives on the tensor dialect and passes for cache-aware distribution of tensors (single and multi-core) and type-aware instructions (VNNI, BFDOT, BFMMLA), including propagation of shapes across the entire function; (2) A linear algebra pipeline, including tile, fuse and bufferization strategies to get model-level IR into hardware friendly tile calls; (3) A mechanism for micro-kernel lowering to an open source library that supports various CPUs.
Zero-shot optimization involves optimizing a target task that was not seen during training, aiming to provide the optimal solution without or with minimal adjustments to the optimizer. It is crucial to ensure reliable and robust performance in various applications. Current optimizers often struggle with zero-shot optimization and require intricate hyperparameter tuning to adapt to new tasks. To address this, we propose a Pretrained Optimization Model (POM) that leverages knowledge gained from optimizing diverse tasks, offering efficient solutions to zero-shot optimization through direct application or fine-tuning with few-shot samples. Evaluation on the BBOB benchmark and two robot control tasks demonstrates that POM outperforms state-of-the-art black-box optimization methods, especially for high-dimensional tasks. Fine-tuning POM with a small number of samples and budget yields significant performance improvements. Moreover, POM demonstrates robust generalization across diverse task distributions, dimensions, population sizes, and optimization horizons. For code implementation, see https://github.com/ninja-wm/POM/.