Skip to Content

Symmetry

Symmetry is an international, peer-reviewed, open access journal covering research on symmetry/asymmetry phenomena wherever they occur in all aspects of natural sciences, and is published monthly online by MDPI.  

Get Alerted

Add your email address to receive forthcoming issues of this journal.

All Articles (18,154)

  • Article
  • Open Access

Hidden-state probes are widely used to study safety-related behavior in language and multimodal models, yet their results can change with the data sources, behavior labels, extracted layer, and probe family. We introduce TriSAIL, an evaluation protocol that records these choices and tests how they affect the resulting claim. TriSAIL uses three response-derived states—benign, non-refusal jailbreak, and borderline/refusal—and assigns train, validation, and test data by source. We extract hidden states from the final input position in the first generation forward pass, before a generated token is returned. The operating layer is selected using validation data only. We evaluate the protocol on five text-only LLMs and six MLLMs with logistic, kNN, and SVM probes. The LLM results vary substantially across held-out attack families, probe families, and source assignments. The MLLM probes achieve high label separability, while source-ID decoding, prediction–source association, and residualization tests show strong source alignment in the same representations. Because the main MLLM matrix couples sources and labels, these results describe source-sensitive separability. Source-transfer conclusions use source-held-out columns; mixed accuracy is a descriptive overall summary. TriSAIL reports these results with probe sensitivity and source diagnostics so that each score is interpreted under the conditions that produced it.

Symmetry

30 September 2026

Overview of the TriSAIL protocol. Source-aware data construction separates benign/helpful, response-labeled non-refusal jailbreak, and refusal/borderline examples and assigns them to train, validation, and test splits. A frozen target LLM or MLLM then yields one hidden-state vector per layer. The vector is taken at the final input position in the first generation forward pass, before a generated token is returned. Candidate probes are trained independently at each layer, while layer selection is performed only on validation data. The selected layer is finally evaluated with source-aware macro-F1 reporting, separating per-source behavior from aggregate performance.
  • Article
  • Open Access

Robotic platoons must maintain geometric formation integrity while navigating complex environments, yet existing performance metrics primarily evaluate tracking accuracy or inter-vehicle spacing without directly quantifying the evolution of formation symmetry. This paper presents a real-time symmetry monitoring framework based on the proposed Turn-Induced Lateral Dispersion (TILD) metric and the Symmetry Stability Index (SSI), a normalized indicator that continuously evaluates geometric symmetry degradation during platoon operation. The methodology estimates the lateral geometric error of each follower, accumulates it through the TILD metric, and transforms the resulting dispersion into the bounded SSI. Using predefined thresholds, the framework classifies platoon operation into NORMAL, DEGRADED, and CRITICAL states, enabling continuous online supervision without modifying the underlying controller. The framework was implemented in a ROS 2 and Gazebo environment and validated with a robotic platoon navigating representative BARN benchmark worlds of increasing geometric complexity. The results reveal a consistent inverse relationship between accumulated lateral dispersion and symmetry stability, allowing progressive formation degradation to be quantitatively characterized across operating conditions. The framework thus offers a lightweight, interpretable, and continuously operating symmetry assessment that complements conventional platoon metrics, establishing a foundation for future symmetry-aware supervisory, diagnostic, and cooperative recovery control strategies.

Symmetry

30 September 2026

Simulated robotic platoon in the Gazebo environment, composed of one leader and two followers arranged in a predecessor-following formation on an obstacle-free demonstration plane.
  • Article
  • Open Access

Does square-lattice symmetry determine the search parameters selected by a fractional random walk with resetting? We compare a walk that updates one coordinate per movement with a walk generated by a fractional power of the full square-torus Laplacian. Both have the same lattice symmetry and nearest-neighbor limit. For the coordinate-updated walk, we prove that coordinate exchange maps the complete first-passage distribution to that of the parameter-swapped target problem. This identity pairs asymmetric local minima and separates the Hessian at a symmetric stationary point into even and odd sectors. For the full-torus walk, we derive the probability of simultaneous two-coordinate motion and relate it exactly to the ratio of the two lazy-chain spectral gaps. Matched searches on a 31-by-31 torus resolve one equal-index candidate for the coordinate-updated walk and two for the full-torus walk at axial target distances 3–5. One-sided derivatives support the reported boundary candidates as constrained solutions, not estimates of an unconstrained optimal index. A fixed-relative-geometry comparison at side lengths 31, 62, and 93 shows that these candidates remain sensitive to lattice resolution and the index cutoff. Thus exchange symmetry constrains relations among search problems, while the generator, encounter rule, and reset timing determine the candidates recovered in a specified finite system.

Symmetry

30 September 2026

Model and transition family. (a) A periodic square lattice with a source/reset site, a five-site digital disk of radius 
  
    ρ
    =
    1
  
, one-coordinate updates, and reset-to-source events. The displayed walker position and arrows are schematic. (b) One-axis displacement probabilities for representative fractional indices on the production cycle 
  
    L
    =
    31
  
. The index-two endpoint places all mass at the two nearest neighbors, whereas smaller indices allocate increasing mass to longer displacements.
  • Article
  • Open Access

The study of hybrid multiterminal high-voltage direct current (HVDC) systems revealed different outcomes regarding their performance under fault conditions. Previous research showed that under the fault conditions, the high-voltage direct current (HVDC) voltage experienced a drop to zero, accompanied by a reverse overshoot. In this paper, a model that integrates hybrid multiterminal line commutated converters (LCCs) and a voltage source converter (VSC) with an HVDC network is presented. The model has been mathematically formulated and implemented using Matlab/Simulink (R2018b) software to examine its fault behaviors and locations, particularly on the DC line to ground fault, line to line fault, and single line to ground fault across various fault resistance levels. The system consists of a wind energy conversion system, a photovoltaic array, protection scheme, LCC rectifier station, VSC inverter station, inverter control, AC filters, a distributed parameter transmission line for the HVDC transmission line, a three-phase inductor-capacitor (LC) filter, and a three-phase transformer. The findings indicated that during the scenario of an 8 ohms of resistance under a single line to ground fault, phase A of the inverter AC grid voltage decreased from its operational level. Meanwhile, the voltage across the DC line increased, accompanied by a rise in DC line current. Additionally, calculations of the fault location were determined to be 1.1561 km from the point of reference. This study contributes to the understanding of fault dynamics in HVDC systems, providing essential insights that enhance operational reliability and efficiency in power transmission networks.

Symmetry

30 September 2026

Proposed diagram of photovoltaic/wind energy conversion grid, connected via HVD.

Featured Articles of Last Quarter

Energy of positive- and negative-parity states in the 
  
    Ca
    
    
    
    
    40
  
 nucleus as a function of J. The states identified with small green circles correspond to well-established spin–parity assignments, while the small red circles indicate states with uncertain spin–parity assignments. The brown lines connecting the circles represent electromagnetic transitions. All the states shown in the figure are taken from the compilation [59,60]. The black lines indicate rotational bands. In (a), the bands built on the 0+, 8+, and 3+ states, as well as the SD band, were previously proposed [9,59,60]. In (b), the 0− band was previously suggested [11,59,60]. The arranged bands are labeled according to the spin–parity (
  
    J
    π
  
) of their band head, with subscripts used to distinguish between bands with the same bandhead spin–parity.
Graphical representations of geometrical parameter used for calculations. The d distance is the distance from halogen atom (X) and centre of double or triple bond.
(a) Isometric view of the tundish–stopper system computational domain. (b) Meshing of the tundish drain system composed of the nozzle and the stopper rod. (c) Stopper mesh. (d) Upper view of the tundish with dimensions. (e) Lateral view of the tundish. (f) Frontal view of the tundish, including a representation of the stopper rod.

Highly Accessed Articles

News & Conferences

Latest Issues

Open for Submission

Journal Sections

Applications of Machine Learning in Large-Scale Optimization and High-Dimensional Learning
Reprint

Applications of Machine Learning in Large-Scale Optimization and High-Dimensional Learning

Editors: Jeng-Shyang Pan, Junzo Watada, Vaclav Snasel, Pei Hu
Intelligent Optimization Algorithm
Reprint

Intelligent Optimization Algorithm

Theory and Applications
Editors: Shi Cheng, Chaomin Luo, Shangce Gao
XFacebookLinkedIn
Symmetry - ISSN 2073-8994