Robustness, the ability of a system to maintain performance under significant and unanticipated environmental changes, is a critical property for robotic systems. While biological systems naturally exhibit robustness, there is no comprehensive understanding of how to achieve similar robustness in robotic systems. In this work, we draw inspirations from biological systems and propose a design principle that advocates active interconnections among system components to enhance robustness to environmental variations. We evaluate this design principle in a challenging long-horizon manipulation task: solving lockboxes. Our extensive simulated and real-world experiments demonstrate that we could enhance robustness against environmental changes by establishing active interconnections among system components without substantial changes in individual components. Our findings suggest that a systematic investigation of design principles in system building is necessary. It also advocates for interdisciplinary collaborations to explore and evaluate additional principles of biological robustness to advance the development of intelligent and adaptable robotic systems.

Knowledge probing assesses to which degree a language model (LM) has successfully learned relational knowledge during pre-training. Probing is an inexpensive way to compare LMs of different sizes and training configurations. However, previous approaches rely on the objective function used in pre-training LMs and are thus applicable only to masked or causal LMs. As a result, comparing different types of LMs becomes impossible. To address this, we propose an approach that uses an LM's inherent ability to estimate the log-likelihood of any given textual statement. We carefully design an evaluation dataset of 7,731 instances (40,916 in a larger variant) from which we produce alternative statements for each relational fact, one of which is correct. We then evaluate whether an LM correctly assigns the highest log-likelihood to the correct statement. Our experimental evaluation of 22 common LMs shows that our proposed framework, BEAR, can effectively probe for knowledge across different LM types. We release the BEAR datasets and an open-source framework that implements the probing approach to the research community to facilitate the evaluation and development of LMs.

Classification tasks in NLP are typically addressed by selecting a pre-trained language model (PLM) from a model hub, and fine-tuning it for the task at hand. However, given the very large number of PLMs that are currently available, a practical challenge is to determine which of them will perform best for a specific downstream task. With this paper, we introduce TransformerRanker, a lightweight library that efficiently ranks PLMs for classification tasks without the need for computationally costly fine-tuning. Our library implements current approaches for transferability estimation (LogME, H-Score, kNN), in combination with layer aggregation options, which we empirically showed to yield state-of-the-art rankings of PLMs (Garbas et al., 2024). We designed the interface to be lightweight and easy to use, allowing users to directly connect to the HuggingFace Transformers and Dataset libraries. Users need only select a downstream classification task and a list of PLMs to create a ranking of likely best-suited PLMs for their task. We make TransformerRanker available as a pip-installable open-source library this https URL

Navigation is controlled by at least two partially dissociable, concurrently developed systems in the brain. The cognitive map informs an organism of its location and bearing, updated by distance-based prediction and vestibular integration. Response-based systems, on the other hand, directly evaluate movement decisions from immediate percepts. Here we demonstrate the sufficiency of visual response-based decision-making in a classic open field navigation task often assumed to require a cognitive map. Three distinct strategies emerge to robustly navigate to a hidden goal, each conferring contextual tradeoffs, as well as aligning with behavior observed with rodents, insects, fish, and sperm cells. We propose reframing navigation from the bottom-up, without assuming online access to computationally expensive top-down representations, to better explain behavior under energetic or attentional constraints.

Relation classification models are conventionally evaluated using only a single measure, e.g., micro-F1, macro-F1 or AUC. In this work, we analyze weighting schemes, such as micro and macro, for imbalanced datasets. We introduce a framework for weighting schemes, where existing schemes are extremes, and two new intermediate schemes. We show that reporting results of different weighting schemes better highlights strengths and weaknesses of a model.

The objects we perceive guide our eye movements when observing real-world dynamic scenes. Yet, gaze shifts and selective attention are critical for perceiving details and refining object boundaries. Object segmentation and gaze behavior are, however, typically treated as two independent processes. Here, we present a computational model that simulates these processes in an interconnected manner and allows for hypothesis-driven investigations of distinct attentional mechanisms. Drawing on an information processing pattern from robotics, we use a Bayesian filter to recursively segment the scene, which also provides an uncertainty estimate for the object boundaries that we use to guide active scene exploration. We demonstrate that this model closely resembles observers' free viewing behavior on a dataset of dynamic real-world scenes, measured by scanpath statistics, including foveation duration and saccade amplitude distributions used for parameter fitting and higher-level statistics not used for fitting. These include how object detections, inspections, and returns are balanced and a delay of returning saccades without an explicit implementation of such temporal inhibition of return. Extensive simulations and ablation studies show that uncertainty promotes balanced exploration and that semantic object cues are crucial to forming the perceptual units used in object-based attention. Moreover, we show how our model's modular design allows for extensions, such as incorporating saccadic momentum or pre-saccadic attention, to further align its output with human scanpaths.

Knowledge probing evaluates the extent to which a language model (LM) has acquired relational knowledge during its pre-training phase. It provides a cost-effective means of comparing LMs of different sizes and training setups and is useful for monitoring knowledge gained or lost during continual learning (CL). In prior work, we presented an improved knowledge probe called BEAR (Wiland et al., 2024), which enables the comparison of LMs trained with different pre-training objectives (causal and masked LMs) and addresses issues of skewed distributions in previous probes to deliver a more unbiased reading of LM knowledge. With this paper, we present LM-PUB- QUIZ, a Python framework and leaderboard built around the BEAR probing mechanism that enables researchers and practitioners to apply it in their work. It provides options for standalone evaluation and direct integration into the widely-used training pipeline of the Hugging Face TRANSFORMERS library. Further, it provides a fine-grained analysis of different knowledge types to assist users in better understanding the knowledge in each evaluated LM. We publicly release LM-PUB-QUIZ as an open-source project.

Available training data for named entity recognition (NER) often contains a significant percentage of incorrect labels for entity types and entity boundaries. Such label noise poses challenges for supervised learning and may significantly deteriorate model quality. To address this, prior work proposed various noise-robust learning approaches capable of learning from data with partially incorrect labels. These approaches are typically evaluated using simulated noise where the labels in a clean dataset are automatically corrupted. However, as we show in this paper, this leads to unrealistic noise that is far easier to handle than real noise caused by human error or semi-automatic annotation. To enable the study of the impact of various types of real noise, we introduce NoiseBench, an NER benchmark consisting of clean training data corrupted with 6 types of real noise, including expert errors, crowdsourcing errors, automatic annotation errors and LLM errors. We present an analysis that shows that real noise is significantly more challenging than simulated noise, and show that current state-of-the-art models for noise-robust learning fall far short of their theoretically achievable upper bound. We release NoiseBench to the research community.

The ability of groups to make accurate collective decisions depends on a complex interplay of various factors, such as prior information, biases, social influence, and the structure of the interaction network. Here, we investigate a spin model that accounts for heterogeneous preferences and enables control over the non-linearity of social interactions. Building on previous results for complete graphs and regular 2D lattices, we investigate how the modification of network topology towards (sparse) random graphs can affect collective decision-making. We use two different measures of susceptibility to assess the responsiveness of the system to internal and external perturbations. In particular, we investigate how the maximum of susceptibility depends on network connectivity. Based on our findings, we discuss how collective systems might adapt to changes in environmental fluctuations by adjusting their network structure or the nature of their social interactions in order to remain in the region of maximal susceptibility.

Given a max-plus linear system and a semimodule, the problem of computing the maximal controlled invariant subsemimodule is still open to this day. In this paper, we consider this problem for the specific class of fully actuated systems and constraints in the form of precedence semimodules. The assumption of full actuation corresponds to the existence of an input for each component of the system state. A precedence semimodule is the set of solutions of inequalities typically used to represent time-window constraints. We prove that, in this setting, it is possible to (i) compute the maximal controlled invariant subsemimodule and (ii) decide the convergence of a fixed-point algorithm introduced by R.D. Katz in strongly polynomial time.

Precedence constraints are inequalities used to model time dependencies. In 1958, Gallai proved that a finite system of precedence constraints admits solutions if and only if the corresponding precedence graph does not contain positive-weight circuits. We show that this result extends naturally to the case of infinitely many constraints. We then analyze two specific classes of infinite precedence graphs -- $\mathbb{N}$-periodic and ultimately periodic graphs -- and prove that the existence of solutions of their related constraints can be verified in strongly polynomial time. The obtained algorithms find applications in P-time event graphs, which are a subclass of P-time Petri nets able to model production systems under cyclic schedules where tasks need to be performed within given time windows.

Pre-trained language models (PLM) are effective components of few-shot named entity recognition (NER) approaches when augmented with continued pre-training on task-specific out-of-domain data or fine-tuning on in-domain data. However, their performance in low-resource scenarios, where such data is not available, remains an open question. We introduce an encoder evaluation framework, and use it to systematically compare the performance of state-of-the-art pre-trained representations on the task of low-resource NER. We analyze a wide range of encoders pre-trained with different strategies, model architectures, intermediate-task fine-tuning, and contrastive learning. Our experimental results across ten benchmark NER datasets in English and German show that encoder performance varies significantly, suggesting that the choice of encoder for a specific low-resource scenario needs to be carefully evaluated.

Sample efficiency is a crucial property of language models with practical implications for training efficiency. In real-world text, information follows a long-tailed distribution. Yet, we expect models to learn and recall frequent and infrequent facts. Sample-efficient models are better equipped to handle this challenge of learning and retaining rare information without requiring excessive exposure. This study analyzes multiple models of varying architectures and sizes, all trained on the same pre-training data. By annotating relational facts with their frequencies in the training corpus, we examine how model performance varies with fact frequency. Our findings show that most models perform similarly on high-frequency facts but differ notably on low-frequency facts. This analysis provides new insights into the relationship between model architecture, size, and factual learning efficiency.

A gaze-fixating robot perceives distance to the fixated object and relative positions of surrounding objects immediately, accurately, and robustly. We show how fixation, which is the act of looking at one object while moving, exploits regularities in the geometry of 3D space to obtain this information. These regularities introduce rotation-translation couplings that are not commonly used in structure from motion. To validate, we use a Franka Emika Robot with an RGB camera. We a) find that error in distance estimate is less than 5 mm at a distance of 15 cm, and b) show how relative position can be used to find obstacles under challenging scenarios. We combine accurate distance estimates and obstacle information into a reactive robot behavior that is able to pick up objects of unknown size, while impeded by unforeseen obstacles. Project page: this https URL .

We study a model of self-propelled particles interacting with their $k$ nearest neighbors through polar alignment. By exploring its phase space as a function of two nondimensional parameters (alignment strength $g$ and Peclet number $\mathrm{Pe}$), we identify two distinct order-disorder transitions. One is continuous, occurs at a low critical $g$ value independent of Pe, and resembles a mean-field transition with no density-order coupling. The other is discontinuous, depends on a combined control parameter involving $g$ and Pe, and results from the formation of small, dense, highly persistent clusters of particles that follow metric-like dynamics. These dense clusters form at a critical value of the combined control parameter $\mathrm{Pe}/g^{\alpha}$, with $\alpha \approx 1.5$, which appears to be valid for different alignment-based models. Our study shows that models of active particles with metric-free interactions can produce characteristic length-scales and self-organize into metric-like collective states that undergo metric-like transitions.

We propose a sensorization method for soft pneumatic actuators that uses an embedded microphone and speaker to measure different actuator properties. The physical state of the actuator determines the specific modulation of sound as it travels through the structure. Using simple machine learning, we create a computational sensor that infers the corresponding state from sound recordings. We demonstrate the acoustic sensor on a soft pneumatic continuum actuator and use it to measure contact locations, contact forces, object materials, actuator inflation, and actuator temperature. We show that the sensor is reliable (average classification rate for six contact locations of 93%), precise (mean spatial accuracy of 3.7 mm), and robust against common disturbances like background noise. Finally, we compare different sounds and learning methods and achieve best results with 20 ms of white noise and a support vector classifier as the sensor model.

In this paper, an approach for active power control of individual wind turbines is presented. State-of-the-art controllers typically employ separate control loops for torque and pitch control. In contrast, we use a multivariable control approach. In detail, active power control is achieved by using reference trajectories for generator speed, generator torque, and pitch angle such that a desired power demand is met if weather conditions allow. Then, a linear quadratic (LQ) optimal controller is used for reference tracking. In an OpenFAST simulation environment, the controller is compared to a state-of-the-art approach. The simulations show a similar active power tracking performance, while the LQ optimal controller results in lower mechanical wear. Moreover, the presented approach exhibits good reference tracking and by improving the reference trajectory generation further performance increases can be expected.

Natural and artificial collectives exhibit heterogeneities across different dimensions, contributing to the complexity of their behavior. We investigate the effect of two such heterogeneities on collective opinion dynamics: heterogeneity of the quality of agents' prior information and of centrality in the network, i.e., the number of immediate neighbors. To study these heterogeneities, we not only consider them in our model, proposing a novel network generator with heterogeneous centrality, but also introduce uncertainty as an additional dimension. By quantifying the uncertainty of each agent, we provide a mechanism for agents to adaptively weigh their individual against social information. As uncertainties develop according to the interactions between agents, they capture information on heterogeneities. Therefore, uncertainty is a relevant additional observable in the study of complex collective opinion dynamics that we use to show the bidirectional relationship of heterogeneous centrality and information. Furthermore, we demonstrate that uncertainty-driven adaptive weighting leads to increased accuracy and speed of consensus, especially under heterogeneity, and provide guidelines for avoiding performance-decreasing errors in uncertainty modeling. These opportunities for improved performance and observability suggest the importance of uncertainty both for the study of natural and the design of artificial heterogeneous systems.