Negative / Null Result ReportOpen accessComputer Science
Masahito Hayashi · 2025 · arXiv
We consider estimation of unknown unitary operation when the set of possible unitary operations is given by a projective unitary representation of a compact group. We show that neither indefinite causal order strategy nor adaptive strategy improves the performance of this estimation when error function satisfies group covariance. That is, the optimal parallel strategy gives the optimal performance even under indefinite causal order strategy and adaptive strategy. To study this problem, we newly introduce the concept of generalized positive operator valued measure (GPOVM), and its convariance c
View details →Negative / Null Result ReportOpen accessComputer Science
Ushak Rahaman, Soebur Razzaque · 2021 · arXiv
The $ν_μ\to ν_e$ oscillation probability over short baseline ($\lesssim 1$~km) would be negligible in case the mixing matrix for three active neutrinos is unitary. However, in case of non-unitary mixing of three neutrinos, this probability would be non-negligible due to the so-called "zero distance" effect. Hence, the near detector of the accelerator experiments such as NO$ν$A can provide strong constraints on the parameters of the non-unitary mixing with very large statistics. By analyzing the NO$ν$A near detector data we find that the non-unitary mixing does not improve fits to the $ν_e$ or
View details →Negative / Null Result ReportOpen accessComputer Science
Xirui Li, Ming Li, Tianyi Zhou · 2026 · arXiv
Reinforcement learning (RL) with verifiable rewards has become a standard post-training stage for boosting visual reasoning in vision-language models, yet it remains unclear what capabilities RL actually improves compared with supervised fine-tuning as cold-start initialization (IN). End-to-end benchmark gains conflate multiple factors, making it difficult to attribute improvements to specific skills. To bridge the gap, we propose a Frankenstein-style analysis framework including: (i) functional localization via causal probing; (ii) update characterization via parameter comparison; and (iii) t
View details →Negative / Null Result ReportOpen accessComputer Science
Yulin Chen, He He, Chen Zhao · 2026 · arXiv
Reinforcement Learning with Verifiable Reward (RLVR) has proven effective in improving Large Language Model's (LLM) reasoning ability. However, the learning dynamics of RLVR remain underexplored. In this paper, we reveal a counterintuitive phenomenon: among hard examples that the model initially struggles with, a substantial subset remains unlearnable even when correct rollouts are present. To understand the phenomenon, we first demonstrate that existing optimization and sampling techniques fail to resolve unlearnability. With cross-example gradient analysis, we show that unlearnable examples
View details →Negative / Null Result ReportOpen accessComputer Science
Hyeong Kyu Choi, Xiaojin Zhu, Sharon Li · 2025 · arXiv
Multi-Agent Debate~(MAD) has emerged as a promising paradigm for improving the performance of large language models through collaborative reasoning. Despite recent advances, the key factors driving MAD's effectiveness remain unclear. In this work, we disentangle MAD into two key components--Majority Voting and inter-agent Debate--and assess their respective contributions. Through extensive experiments across seven NLP benchmarks, we find that Majority Voting alone accounts for most of the performance gains typically attributed to MAD. To explain this, we propose a theoretical framework that mo
View details →Negative / Null Result ReportOpen accessComputer Science
Mitchell A. Gordon, Kevin Duh, Nicholas Andrews · 2020 · arXiv
Pre-trained universal feature extractors, such as BERT for natural language processing and VGG for computer vision, have become effective methods for improving deep learning models without requiring more labeled data. While effective, feature extractors like BERT may be prohibitively large for some deployment scenarios. We explore weight pruning for BERT and ask: how does compression during pre-training affect transfer learning? We find that pruning affects transfer learning in three broad regimes. Low levels of pruning (30-40%) do not affect pre-training loss or transfer to downstream tasks a
View details →Negative / Null Result ReportOpen accessComputer Science
Damian Ruck, R. Alexander Bentley, Alberto Acerbi et al. · 2017 · arXiv
Here we test Neutral models against the evolution of English word frequency and vocabulary at the population scale, as recorded in annual word frequencies from three centuries of English language books. Against these data, we test both static and dynamic predictions of two neutral models, including the relation between corpus size and vocabulary size, frequency distributions, and turnover within those frequency distributions. Although a commonly used Neutral model fails to replicate all these emergent properties at once, we find that modified two-stage Neutral model does replicate the static a
View details →Negative / Null Result ReportOpen accessComputer Science
C. Fernandez-Ramirez, A. M. Bernstein · 2012 · arXiv
With the availability of the new neutral pion photoproduction from the proton data from the A2 and CB-TAPS Collaborations at Mainz it is mandatory to revisit Heavy Baryon Chiral Perturbation Theory (HBChPT) and address the extraction of the partial waves as well as other issues such as the value of the low-energy constants, the energy range where the calculation provides a good agreement with the data and the impact of unitarity. We find that, within the current experimental status, HBChPT with the fitted LECs gives a good agreement with the existing neutral pion photoproduction data up to $\s
View details →Negative / Null Result ReportOpen accessComputer Science
Zichao Wang, Alexa Siu · 2026 · arXiv
Large language models (LLMs) have shown strong performance on standardized social science instruments, but their value for product discovery remains unclear. We investigate whether interview-informed generative agents can simulate user responses in concept testing scenarios. Using in-depth workflow interviews with knowledge workers, we created personalized agents and compared their evaluations of novel AI concepts against the same participants' responses. Our results show that agents are distribution-calibrated but identity-imprecise: they fail to replicate the specific individual they are gro
View details →Negative / Null Result ReportOpen accessComputer Science
Lawrence Clegg, John Cartlidge · 2025 · arXiv
Intransitive player dominance, where player A beats B, B beats C, but C beats A, is common in competitive tennis. Yet, there are few known attempts to incorporate it within forecasting methods. We address this problem with a graph neural network approach that explicitly models these intransitive relationships through temporal directed graphs, with players as nodes and their historical match outcomes as directed edges. Our model (65.7% accuracy, 0.214 Brier score) forecasts competitively with established rating systems such as Weighted Elo. Although it does not improve on the baseline in uncond
View details →Negative / Null Result ReportOpen accessComputer Science
Roland S. Zimmermann, Thomas Klein, Wieland Brendel · 2023 · arXiv
In light of the recent widespread adoption of AI systems, understanding the internal information processing of neural networks has become increasingly critical. Most recently, machine vision has seen remarkable progress by scaling neural networks to unprecedented levels in dataset and model size. We here ask whether this extraordinary increase in scale also positively impacts the field of mechanistic interpretability. In other words, has our understanding of the inner workings of scaled neural networks improved as well? We use a psychophysical paradigm to quantify one form of mechanistic inter
View details →Negative / Null Result ReportOpen accessComputer Science
Wataru Masaka, Mitsuki Sakamoto, Kenshi Abe et al. · 2025 · arXiv
We investigate how perturbation does and does not improve the Follow-the-Regularized-Leader (FTRL) algorithm in solving imperfect-information extensive-form games under sampling, where payoffs are estimated from sampled trajectories. While optimistic algorithms are effective under full feedback, they often become unstable in the presence of sampling noise. Payoff perturbation offers a promising alternative for stabilizing learning and achieving \textit{last-iterate convergence}. We present a unified framework for \textit{Perturbed FTRL} algorithms and study two variants: PFTRL-KL (standard KL
View details →Negative / Null Result ReportOpen accessComputer Science
Hamed R. Tavakoli, Rakshith Shetty, Ali Borji et al. · 2017 · arXiv
To bridge the gap between humans and machines in image understanding and describing, we need further insight into how people describe a perceived scene. In this paper, we study the agreement between bottom-up saliency-based visual attention and object referrals in scene description constructs. We investigate the properties of human-written descriptions and machine-generated ones. We then propose a saliency-boosted image captioning model in order to investigate benefits from low-level cues in language models. We learn that (1) humans mention more salient objects earlier than less salient ones i
View details →Negative / Null Result ReportOpen accessComputer Science
Arpit Agarwal · 2021 · arXiv
Predicting the outcome of liquid droplet collisions is an extensively studied phenomenon but the current physics based models for predicting the outcomes are poor (accuracy $\approx 43\%$). The key weakness of these models is their limited complexity. They only account for 3 features while there are many more relevant features that go unaccounted for. This limitation of traditional models can be easily overcome through machine learning modeling of the problem. In an ML setting this problem directly translates to a classification problem with 4 classes. Here we compile a large labelled dataset
View details →Negative / Null Result ReportOpen accessComputer Science
Thomas Vogel, Chinh Tran, Lars Grunske · 2019 · arXiv
In search-based software engineering we often use popular heuristics with default configurations, which typically lead to suboptimal results, or we perform experiments to identify configurations on a trial-and-error basis, which may lead to better results for a specific problem. To obtain better results while avoiding trial-and-error experiments, a fitness landscape analysis is helpful in understanding the search problem, and making an informed decision about the heuristics. In this paper, we investigate the search problem of test suite generation for mobile applications (apps) using SAPIENZ w
View details →Negative / Null Result ReportOpen accessComputer Science
Shintaro Sakai, Jisun An, Migyeong Kang et al. · 2025 · arXiv
Prior clinical psychology research shows that Western individuals with depression tend to report psychological symptoms, while Eastern individuals report somatic ones. We test whether Large Language Models (LLMs), which are increasingly used in mental health, reproduce these cultural patterns by prompting them with Western or Eastern personas. Results show that LLMs largely fail to replicate the patterns when prompted in English, though prompting in major Eastern languages (i.e., Chinese, Japanese, and Hindi) improves alignment in several configurations. Our analysis pinpoints two key reasons
View details →Negative / Null Result ReportOpen accessComputer Science
Shufan Wang, Yixiao Song, Andrew Drozdov et al. · 2023 · arXiv
In this paper, we study the generation quality of interpolation-based retrieval-augmented language models (LMs). These methods, best exemplified by the KNN-LM, interpolate the LM's predicted distribution of the next word with a distribution formed from the most relevant retrievals for a given prefix. While the KNN-LM and related methods yield impressive decreases in perplexity, we discover that they do not exhibit corresponding improvements in open-ended generation quality, as measured by both automatic evaluation metrics (e.g., MAUVE) and human evaluations. Digging deeper, we find that interp
View details →Negative / Null Result ReportOpen accessComputer Science
Niek Beckers, Edwin van Asseldonk, Herman van der Kooij · 2020 · arXiv
Haptic interaction between two humans, for example, a physiotherapist assisting a patient regaining the ability to grasp a cup, likely facilitates motor skill acquisition. Haptic human-human interaction has been shown to enhance individual performance improvement in a tracking task with a visuomotor rotation perturbation. These results are remarkable given that haptically assisting or guiding an individual rarely benefits their individual improvement when the assistance is removed. We, therefore, replicated a study that reported that haptic interaction between humans was beneficial for individ
View details →Negative / Null Result ReportOpen accessComputer Science
Matthew Rueben, Frank J. Bernieri, Cindy M. Grimm et al. · 2019 · arXiv
Privacy-sensitive robotics is an emerging area of HRI research. Judgments about privacy would seem to be context-dependent, but none of the promising work on contextual "frames" has focused on privacy concerns. This work studies the impact of contextual "frames" on local users' privacy judgments in a home telepresence setting. Our methodology consists of using an online questionnaire to collect responses to animated videos of a telepresence robot after framing people with an introductory paragraph. The results of four studies indicate a large effect of manipulating the robot operator's identit
View details →Negative / Null Result ReportOpen accessComputer Science
Runtian Zhai, Chen Dan, Zico Kolter et al. · 2022 · arXiv
Empirical risk minimization (ERM) is known in practice to be non-robust to distributional shift where the training and the test distributions are different. A suite of approaches, such as importance weighting, and variants of distributionally robust optimization (DRO), have been proposed to solve this problem. But a line of recent work has empirically shown that these approaches do not significantly improve over ERM in real applications with distribution shift. The goal of this work is to obtain a comprehensive theoretical understanding of this intriguing phenomenon. We first posit the class o
View details →Negative / Null Result ReportOpen accessComputer Science
Alex Ayoub, Samuel Robertson, Dawen Liang et al. · 2025 · arXiv
Matrix factorization is a widely used approach for top-N recommendation and collaborative filtering. When implemented on implicit feedback data (such as clicks), a common heuristic is to upweight the observed interactions. This strategy has been shown to improve performance for certain algorithms. In this paper, we conduct a systematic study of various weighting schemes and matrix factorization algorithms. Somewhat surprisingly, we find that training with unweighted data can perform comparably to, and sometimes outperform, training with weighted data, especially for large models. This observat
View details →Negative / Null Result ReportOpen accessComputer Science
Paul K. Mandal · 2025 · arXiv
In this paper, I investigate the effectiveness of dataset cartography for extractive question answering on the SQuAD dataset. I begin by analyzing annotation artifacts in SQuAD and evaluate the impact of two adversarial datasets, AddSent and AddOneSent, on an ELECTRA-small model. Using training dynamics, I partition SQuAD into easy-to-learn, ambiguous, and hard-to-learn subsets. I then compare the performance of models trained on these subsets to those trained on randomly selected samples of equal size. Results show that training on cartography-based subsets does not improve generalization to
View details →Negative / Null Result ReportOpen accessComputer Science
Ab Mosca, Alvitta Ottley, Remco Chang · 2021 · arXiv
Interaction enables users to navigate large amounts of data effectively, supports cognitive processing, and increases data representation methods. However, there have been few attempts to empirically demonstrate whether adding interaction to a static visualization improves its function beyond popular beliefs. In this paper, we address this gap. We use a classic Bayesian reasoning task as a testbed for evaluating whether allowing users to interact with a static visualization can improve their reasoning. Through two crowdsourced studies, we show that adding interaction to a static Bayesian reaso
View details →Negative / Null Result ReportOpen accessComputer Science
Richard G. Clegg · 2006 · arXiv
The aim of this paper is to use a very simple queuing model to compare a number of models from the literature which have been used to replicate the statistical nature of internet traffic and, in particular, the long-range dependence of this traffic. The four models all have the form of discrete time Markov-modulated processes (two other models are introduced for comparison purposes). While it is often stated that long-range dependence has a critical effect on queuing performance, it appears that the models used here do not well replicated the queuing performance of real internet traffic. In pa
View details →Negative / Null Result ReportOpen accessComputer Science
Dillon Niederhut · 2023 · arXiv
Over the last decade, deep neural networks have achieved state of the art in computer vision tasks. These models, however, are susceptible to unusual inputs, known as adversarial examples, that cause them to misclassify or otherwise fail to detect objects. Here, we provide evidence that the increasing success of adversarial attacks is primarily due to increasing their size. We then demonstrate a method for generating the largest possible adversarial patch by building a adversarial pattern out of repeatable elements. This approach achieves a new state of the art in evading detection by YOLOv2 a
View details →Negative / Null Result ReportOpen accessComputer Science
Joris Dannemann, Gero Junike · 2025 · arXiv
Batch normalization is one of the most important regularization techniques for neural networks, significantly improving training by centering the layers of the neural network. There have been several attempts to provide a theoretical justification for batch ormalization. Santurkar and Tsipras (2018) [How does batch normalization help optimization? Advances in neural information rocessing systems, 31] claim that batch normalization improves initialization. We provide a counterexample showing that this claim s not true, i.e., batch normalization does not improve initialization.
View details →Negative / Null Result ReportOpen accessComputer Science
Cheng Tang, Andrew Arnold · 2020 · arXiv
Recently, Nogueira et al. [2019] proposed a new approach to document expansion based on a neural Seq2Seq model, showing significant improvement on short text retrieval task. However, this approach needs a large amount of in-domain training data. In this paper, we show that this neural document expansion approach can be effectively adapted to standard IR tasks, where labels are scarce and many long documents are present.
View details →Negative / Null Result ReportOpen accessComputer Science
Mohan Raj Chanthran, Lay-Ki Soon, Huey Fang Ong et al. · 2024 · arXiv
Malaysian English is a low resource creole language, where it carries the elements of Malay, Chinese, and Tamil languages, in addition to Standard English. Named Entity Recognition (NER) models underperform when capturing entities from Malaysian English text due to its distinctive morphosyntactic adaptations, semantic features and code-switching (mixing English and Malay). Considering these gaps, we introduce MENmBERT and MENBERT, a pre-trained language model with contextual understanding, specifically tailored for Malaysian English. We have fine-tuned MENmBERT and MENBERT using manually annot
View details →Negative / Null Result ReportOpen accessComputer Science
Nima Akhlaghi, Ananya Dhawan, Amir A. Khan et al. · 2018 · arXiv
Objective: The objectives of this paper are to determine the optimal location for ultrasound transducer placement on the anterior forearm for imaging maximum muscle deformations during different hand motions and to investigate the effect of using a sparse set of ultrasound scanlines for motion classification for ultrasound-based muscle computer interfaces (MCIs). Methods: The optimal placement of the ultrasound transducer along the forearm is identified using freehand 3D reconstructions of the muscle thickness during rest and motion completion. From the ultrasound images acquired from the opti
View details →Negative / Null Result ReportOpen accessComputer Science
Adam Brandenburger, Pierfrancesco La Mura · 2025 · arXiv
Any quasi-probability representation of a no-signaling system -- including quantum systems -- can be simulated via a purely classical scheme by allowing signed events and a cancellation procedure. This raises a fundamental question: What properties of the non-classical system does such a classical simulation fail to replicate? We answer by using large deviation theory to show that the probability of a large fluctuation under the classical simulation can be strictly greater than under the actual non-classical system. The key finding driving our result is that negativity in probability relaxes t
View details →