Papers

Equivalent dq0 models of three-phase passive impedances are widely used for the design of three-phase inverter and rectifier controllers in the synchronous frame. However, these equivalent models assume that the impedance in each phase are the same, which is not always applicable. The fact that the impedance in each phase is different, generates second order harmonic content in the dq0 components, which hinders the advantages of the dq0 transformation. To avoid the second harmonic in these scenarios, it is possible to use the Fortescue's theorem to represent any asymmetrical set of three-phase signals as a linear combination of three symmetrical sequences. In the literature, only a few works have attempted to derive the equivalent model of unbalanced three-phase passive elements in the sequence domain, and none is able to account for the interaction between the zero sequence and the positive and negative sequences. This work presents the derivation of the equivalent dq sequence-domain model of unbalanced three-phase passive elements both analytically and with an intuitive graphical approach. Experimental results are shown to validate the proposed models.

We describe SCARAB–Swarm-Capable Autonomous Robotic Aquatic Bridging. Using distributed swarm control and multi-model sensing of agents and docking targets, agents can localize themselves, join into formations and proceed to desired target locations. Our technologies would eventually allow the Army to perform unpredictable, dispersed river crossings, enhance crew survivability, and minimize the logistics footprint compared to the current Improved Ribbon Bridge. Our methods operate without GPS or RF communications, though these can be used in non-contested environments (e.g., civilian disaster relief for flooding, etc.). We demonstrate our system via physics-engine-based simulation of several agents using unmanned surface vehicles (USVs) in the presence of currents and wind.

Active wildfire mapping from satellite imagery is challenging due to the sparse and highly imbalanced nature of fire pixels, especially in early-stage or low-density fire observations. This work investigates the use of multispectral Landsat-8 imagery for active-fire segmentation under multi-scale wildfire size conditions. We propose a data-driven protocol to characterize fire-region size distributions through connected-component analysis and an interquartile range criterion, enabling the evaluation of model robustness across different local fire-region densities. Three segmentation architectures, U-Net, DeepLabV3+, and SegFormer, are evaluated under different SWIR-based spectral configurations. Results show that U-Net achieves the strongest robustness across the evaluated conditions, SegFormer provides competitive performance, and DeepLabV3+ tends to produce conservative predictions with reduced recall. Across architectures, SWIR2 consistently achieves the strongest or near-best results, highlighting its importance for active-fire segmentation in Landsat-8 imagery. These findings suggest that both spectral band selection and architectural design are critical for robust satellite-based active wildfire mapping trained on low active fire-pixel density images.

Unmanned Aerial Vehicles (UAVs) have emerged as a promising platform for firefighting operations due to their flexibility, low operational cost, and ability to acquire high-resolution imagery in locations that may be difficult or dangerous to access using conventional methods. Recent advances in deep learning have significantly improved the capabilities of UAV-based wildfire monitoring systems. The present work investigates RGB-infrared fusion for binary wildfire segmentation on the FLAME3 dataset. In this Study, RGB and Infrared baselines are compared with three representative fusion strategies across three segmentation architectures, including U-Net, DeepLabV3+, and SegFormer. The key motivation of this work is to analyze the contribution of each modality, evaluate the impact of fusion timing, and examine how different network architectures exploit multimodal information for UAV wildfire delineation. The findings indicate that thermal information plays a dominant role in UAV segmentation and that feature-level multimodal fusion combined with transformer-based architectures offers the most promising direction for future research.

Tokenization in language models is treated by default as an input preprocessing decision. We argue that this framing is incomplete: in autoregressive models, tokenizer granularity determines what the model must resolve in a single forward pass, and therefore the supervision signal it receives. This affects both the difficulty of the learning problem and the representations that emerge inside the model. We test this in a controlled experiment on numeric reasoning with a novel decoupling of input and output tokenization. As the output supervision view predicts, differences in task performance, training dynamics, and model internals are induced by output tokenization and largely invariant to input tokenization. This may matter in practice, because models with different tokenization strategies differ not only in input representation but in the task they were trained on. Comparisons between models may thus partly reflect task definition rather than ability. A survey of 120 recent *CL papers on numeric reasoning confirms that this is rarely acknowledged: only about 10% report the numeric tokenization of the models they evaluate, while 69% compare across tokenization, and thus supervision, regimes without reporting it. While prior work documents that tokenization consistently affects model performance, there is no principled account of why. We argue that framing tokenization as output supervision provides that account.

Long-duration outdoor coverage with autonomous platforms remains challenging beyond classical planning: deployments face localization drift in open spaces, obstacles in cluttered sites, controller feasibility in turn-heavy maneuvers, and persistent autonomy with energy management. We propose a unified ROS 2 architecture for single-robot outdoor coverage: a dual-antenna RTK-GNSS fused in an EKF keeps both position and heading accurate across long missions; three controller-aware refinements extend a mature coverage planner; a Nav2-based behavior-tree mission executive coordinates multi-goal execution, layered recovery, cost-aware goal management, and autonomous docking for return-to-charge. In real-world trials across five outdoor areas with varying geometries and obstacle densities, the robot completed every coverage route, sweeping 93.1% to 96.1% of the planned coverage area.

Vision Language Models have demonstrated remarkable proficiency in interpreting static visual artifacts, but modern data analysis is inherently dynamic, requiring the active interrogation of interactive environments. Existing benchmarks are predominantly constrained to static imagery and one-shot question answering and fail to capture the epistemic demands of this domain, where evidence is frequently occluded, distributed across linked views, or conditionally revealed through user agency. In this paper, we introduce InSight, a benchmark for agentic claim verification over interactive visualizations. The dataset consists of 21,349 claims derived from human-authored analytical narratives and grounded in fully interactive web-based environments. Agents must navigate these environments to determine whether a natural language claim is supported, refuted or not verifiable given the available evidence. Unlike traditional evaluations, InSight treats interaction traces as intrinsic proxies for reasoning, enabling a rigorous audit of how models seek and synthesize visual evidence. We evaluate state-of-the-art models, revealing that interactive verification remains a non-trivial challenge. We release InSight at https://github.com/maevehutch/insight.

On the once-punctured torus, we give an exact formula for the number of curves in any given mapping class group orbit of given word length. This settles a conjecture of Chas in [Cha16].

Encoder-only Transformers remain effective for discriminative and representation-learning tasks, yet Polish encoders still largely rely on BERT/RoBERTa-style architectures. We introduce Polish ModernBERT, a family of four Polish encoders available at Base and Large scales, each with 512-token and 8K context variants. We adapt the ModernBERT pretraining recipe through staged selection experiments and release a long-context benchmark covering legal topic classification, ideological decision-direction prediction, factual-consistency assessment over literary plot summaries, and human-rights violation assessment. Across 30 tasks, Polish ModernBERT achieves the best overall performance among the evaluated Polish encoders, reaching 83.99 and 85.11 for the Base-8K and Large-8K models, respectively. On long-context tasks, the 8K variants improve over matched Polish RoBERTa-8K baselines from 67.47 to 77.15 and from 75.88 to 78.49 at the Base and Large scales, respectively. The Base-8K model achieves this gain with 22% fewer parameters (149M vs.\ 190M). Efficiency measurements in representative inference setups show lower peak memory usage and latency than matched Polish RoBERTa baselines in both 512-token and 8K settings. Polish ModernBERT-8K-Base additionally achieves the best result on a Polish retrieval benchmark among the evaluated encoders below 300M parameters.

Large Vision-Language Models (LVLMs) have achieved strong multimodal performance, yet ensuring the factual correctness of generated content remains challenging. Existing methods that provide statistical guarantees on factuality typically rely on external verifiers or generation-time confidence signals, which introduce auxiliary dependencies or often fail for confident but incorrect outputs. We argue that reliable factuality control can instead be achieved through introspective signals derived from the model itself. We introduce IntroConformal, a training-free Conformal Risk Control (CRC) framework that provides finite-sample, distribution-free factuality guarantees. We first instantiate it with layer-wise semantic stability, a conformity score derived from hidden-state representations, and then propose verification probability, a stronger score capturing the model's self-administered judgment on claim factuality. Across multiple LVLM architectures, IntroConformal satisfies the conformal risk guarantee while substantially reducing abstention and achieving competitive or superior claim-level discrimination relative to external verifier-based baselines.

Low-rank adaptation fixes the rank of the update, but it does not identify which parts of a trained write actually carry behavior. We study that question directly and show that behaviorally effective LoRA writes are sparse, structured, and far more concentrated than the raw low-rank parameterization suggests. We use Learned-Basis LoRA, a learned-basis continuation recipe, to expose that structure. The recipe warms up an unconstrained adapter, converts its learned write columns into a module-wise orthonormal basis, freezes that basis, and continues training inside the constrained parameterization. Across 14 exact switches from unconstrained to constrained form, held-out accuracy is unchanged at the conversion step and reconstructed write matrices differ by at most 0.25% relative Frobenius error. Same-state continuation then shows that the same trained checkpoint develops differently under different write subspaces, establishing write geometry as a causal state variable. A no-retraining projection test shows that useful write signal stays inside the learned write space and largely disappears from random or frozen-activation PCA controls. The concentration pattern is strong at both local and global scales. Across GSM8K, MathQA, and AQuA, per-module top-k continuation reaches its optimum at k in {2, 4} in all twelve seed-level cases we test. A stricter global ranking test shows that learned top-16 and top-32 subsets outperform matched random subsets, especially on GSM8K/Qwen and MathQA/Qwen. Single-direction ablations further reveal a sparse set of late q_proj, o_proj, and down_proj components with outsized behavioral impact.

Data is important in many deep learning-based inverse problem solvers. However, obtaining sufficient paired data in many scenarios remains highly challenging, while unpaired data is cheap. To maximize data utilization, this paper proposes LUD-DIF, a diffusion-based approach for solving inverse problems with unpaired data. Starting from the evidence lower bound (ELBO) of the joint distribution, we decouple it into two independent diffusion processes under the weak-coupling assumption. The method provides theoretical support from a variational inference perspective, derives the loss function, quantitatively analyzes the error bound introduced by the assumption, and offers a theorem-motivated heuristic for hyperparameter selection. Experimental results demonstrate that LUD-DIF achieves outstanding performance on multiple image inverse problems, validating its effectiveness and generalization capability in unpaired inverse problem settings.

Reliable evaluation of open-ended question answering remains a bottleneck for measuring answer correctness of modern LLMs. Unlike multiple-choice tasks, free-form answers may be correct in many surface forms and may fail in qualitatively different ways, including incompleteness, contradiction, overgeneration, and endorsement of false premises. Existing judgment-based and similarity-based metrics often collapse these distinctions. We address this gap with three reusable contributions. First, we introduce a semantic correctness taxonomy that assigns open-ended answers to eight ordered classes, separating verbose-but-correct answers from those contaminated by hallucinated content. Second, we release CAP-Correctness, an 8.8k-example benchmark spanning widely used QA datasets, and CAP-Statements, an 11k-example dataset for converting question-answer pairs into declarative statements for natural language inference (NLI) training and statement-based evaluation. Third, we introduce CAP (Context-Aware Precision), a reference-based metric that scores question-conditioned statements using bidirectional NLI. Under a monotonicity protocol testing whether metrics respect the taxonomy's intended ordering, CAP outperforms established baselines.

Non-destructive X-ray and computed tomography (CT) testing are essential for ensuring the dimensional accuracy of manufactured components with complex internal structures, such as the cooling channels in gas turbine blades, which directly affect thermal performance and service life. This study presents a multipart 3D-2D rigid registration approach for aligning CAD models with X-ray projections as an alternative to CT reconstruction for part inspection and measurement. A greedy registration algorithm sequentially aligns the blade's exterior before registering its internal components by maximizing the mutual information between simulated and acquired X-ray images. This stepwise approach reduces problem complexity and improves alignment accuracy. View angles are optimized using a greedy method that iteratively selects angles to minimize dimensional measurement errors. The results indicate that a small number of oblique views provides the best accuracy, although a broad range of angles yields acceptable results. The method achieves subpixel registration accuracy, with errors below one-fifth of the magnified detector-pixel pitch. Image noise and defects reduce registration precision, but direct registration in projection space mitigates these effects compared with CT reconstruction. Appropriate view selection can therefore preserve acceptable subpixel accuracy in the presence of image noise and defects.

In the Shortest Common Superstring problem (SCS), one is given a finite set of strings and is asked to find a shortest string containing every input string as a substring. Since the problem is NP-hard, various approximation algorithms have been proposed for it. The currently best known approximation ratio is $2.466$ due to Englert, Matsakis, and Veselý; a better approximation ratio $2$ has been announced by OpenAI just recently. In 1988, Tarhio and Ukkonen introduced a simple greedy algorithm that iteratively takes two strings with the longest overlap and merges them together. They showed that the greedy algorithm may produce a superstring that is about two times longer than the optimal one and conjectured that this bound is tight. This has become known as the Greedy Superstring Conjecture and has remained open for almost four decades. The best known upper bound on the approximation ratio of the greedy algorithm is $3.396$ due to Englert, Matsakis, and Veselý. In this paper, we present a series of results tightening the bounds of the greedy algorithm, with the main contributions being as follows: 1. We refute the Greedy Conjecture by constructing a family of instances for which a lower bound on the approximation ratio tends to $2.4$. 2. We prove that the greedy algorithm is $3$-approximate. 3. For input strings of fixed length $k$, we prove that the approximation guarantee is exactly $9/5$ for $k=3$ and at least $2$ for all $k \ge 6$.

Linear classifiers trained on hidden states of a large language model (LLM), linear probes, can flag factual errors from a single forward pass. Geometrically, that implies that true and false statements separate along a stable direction in hidden state space, i.e., the truth direction. Prior work disagrees on whether this generalises across input shifts, but the disagreement is hard to interpret because cross-dataset probe transfer experiments confound several kinds of input change at once. We isolate three such variables in medical question-answering (QA): writing style (register), domain (medical specialty), and corpus (dataset). We build a benchmark using 500 MedQA entries, each rewritten into four styles (textbook, patient, clinical note, colloquial), annotated with clinical specialty, and grouped with two other exam corpora, MedMCQA and MMLU-medical, for cross-dataset evaluation. Probing four open-weight LLMs (2–8B), we find that the truth direction is largely robust to writing style (mean $\Delta_\text{register} \approx 0.10$ AUROC on held-out facts) and to medical specialty ($\Delta_\text{specialty} \approx 0.03$), but degrades unevenly across corpora: by $0.12$ AUROC on MMLU-medical and by $0.21$ on MedMCQA, roughly twice the register gap. The register result replicates with a second generator and carries over to human-written patient questions. The truth direction is therefore largely stable within the medical domain but breaks under some corpus shifts, and question format does not explain the break, which suggests that the signal a linear probe recovers is partly bound to dataset structure rather than to medical knowledge alone.

Large language model (LLM) agent failures often contain multiple related errors rather than a single mistake. Existing attribution methods usually identify a responsible agent, step, or root cause, but do not explicitly model dependency between errors. We introduce EDGE, an Error Dependency Graph-guided multi-Error attribution framework. EDGE constructs an error dependency graph from observed error events and validates a reliable causal subset through counterfactual rollout. The inference graph guides a two-stage LLM-as-judge detector for error attribution, and the intervention-validated subgraph provides a more reliable basis for explanation and repair analysis. Experiments on TRAIL and MAST show that EDGE improves category-level multi-error attribution across most evaluated models and settings. Experiments with adapted Who&When-style prompts show that the graph helps across prompting strategies. These results suggest that dependency structure is a useful diagnostic prior for agent failures beyond isolated root-cause prediction.

Predicting transcriptional responses to specific perturbations is critical for understanding cellular regulatory mechanisms and accelerating drug discovery. Single-cell RNA sequencing destroys each measured cell, yielding only unpaired populations of control and perturbed cells. However, existing methods typically model perturbation prediction at the single-cell level and assume cell-to-cell correspondence, which conflicts with the unpaired nature of the observed data. To address this challenge, we propose PopPert, a framework that explicitly parameterizes population-level joint gene expression distributions for collective transcriptional state modeling. Given a control population distribution and a perturbation condition, PopPert predicts perturbation-induced changes in distribution parameters, eliminating the need for cell-level correspondence and reducing sensitivity to single-cell noise. To effectively capture gene co-expression patterns, PopPert leverages a low-rank Gaussian Copula to model cross-gene statistical dependencies and construct the joint gene expression distribution, additionally allowing sampling of synthetic perturbed single-cell profiles. Across multiple single-cell benchmarks spanning both genetic and chemical perturbations, PopPert achieves superior overall performance in differential expression recovery, perturbation effect estimation, and population-level distribution matching. These results establish population-level joint distribution learning as an effective paradigm for predicting transcriptional responses from unpaired single-cell populations. Code for PopPert is publicly available at https://github.com/whd1125/PopPert.

Multilingual large language models (mLLMs) achieve strong performance in machine translation, yet our understanding of the mechanisms by which they transform representations from one language to another remains incomplete. Prior work suggests that translation decomposes into separable processes within an mLLM, where conceptual content is first represented independently, followed by a production into language-specific form. In this work, we show that translation is even more modular than previously assumed and that the output language production in translation processes is actually further separable into a syntax and a surface language process. We construct controlled multilingual datasets that isolate cross-linguistic differences in word-order and use causal interventions and probing to track how representations are transformed during translation. We find that models first construct target-side word-order before realizing the target language surface form. We identify individual attention heads that are selectively sensitive to syntactic transformations while remaining largely invariant to language identity. These results establish the commitment to a syntactic structure as an independent stage in translation, extending prior decompositions and showing how translation is implemented by functionally different components within mLLMs.

Scenario optimization, conformal prediction, and related distribution-free certification methods use finite samples to construct decisions or prediction sets with violation-risk guarantees for fresh observations. In several classical settings, the conditional violation risk follows an exact beta law, whose tail has a beta-binomial representation and whose parameter is a support, calibration, or compression dimension. This paper identifies the deterministic boundary mechanism behind these formulas and derives the corresponding law when the observed boundary size is random. A decision rule is represented by an acceptance set for future observations, together with a boundary map selecting the sample points responsible for that set. The resulting pair is called a proper projective boundary scheme when held-out samples are accepted precisely if the full-sample boundary is retained, and accepted non-boundary samples can be deleted without changing that boundary. For every such scheme, the conditional law of the violation risk given the observed boundary size is determined by the boundary's cross-sample complexity profile. A stable profile yields the usual beta law, whereas a varying profile produces an exact profile correction. The framework covers scalar order-statistic calibration, support-reconstructive scenario programs, cascaded support-removal certificates, coordinatewise envelopes, and Pareto-frontier calibration with vector scores. It also yields conditional probabilistic certificates and a no-go result explaining why observed complexity alone is insufficient.

Reinforcement learning with verifiable rewards (RLVR) and standard benchmark evaluation both rely on an automatic verifier that turns a free text answer into a binary reward. Prior work reports that one evaluation harness accepts only about 94% of its own ground truth answers, blaming LaTeX parsing. That is an aggregate: it does not say which answer forms consume the error budget. We supply the decomposition. We apply metamorphic testing to the verifier rather than the model, generating certified equivalent answer variants, that is, rewrites that preserve mathematical meaning by construction, so that any rejection is a provable false negative needing no human adjudication. We then measure rejection per answer category across four widely used verifiers over 307,420 verdicts. We find three things. (1) Self validation ranges from 53.8% to 95.2% on identical inputs, a spread of 41.3 points. The published figure describes one implementation, not the task; two configurations of the same library disagree on 49.9% of pairs. (2) The residual is not spread across parsing categories but concentrated in whitespace and punctuation, which account for 93.0% of in contract failures for the default LaTeX configuration. A trailing period or newline dominates the budget. (3) Separating rejection from execution failure shows that verifiers with similar aggregate error fail for opposite reasons, and that a reference numeric cascade accepts off by one wrong answers as a step function of magnitude, from 0% below 10^4 to 100% at or above, because its relative tolerance is scale invariant.

Protein inverse folding aims to recover amino acid sequences for a given 3D protein structure, underpinning broad applications such as enzyme engineering and drug discovery.Current methods often follow a serial pipeline, in which a structure encoder predicts a coarse sequence, which is then refined by protein language models (PLMs). However, because PLMs only perform post-hoc sequence edits, the refinement is bounded by the quality of upstream predictions.Thanks to recent multimodal protein language models (MPLMs), we could directly encode structure to generate sequences with pretrained structural knowledge, but we observe that they are not effective for inverse folding. Therefore, we introduce a symmetric dual-path architecture that both leverages PLMs for pretrained sequence evolution knowledge and MPLMs for pretrained structural knowledge to iteratively guide protein sequence generation.Through extensive experiments across standard protein inverse folding benchmarks, our method achieves state-of-the-art performance, surpassing prior approaches, and ablation studies validate the rationale of our symmetric design, revealing a promising direction for the community.

Role-playing large language models (LLMs) are expected to adopt a character's style while also respecting that character's knowledge boundaries. Prior evaluations detect character hallucination but rarely distinguish whether errors arise from failure to recognize a boundary or from failure to comply despite recognition. We introduce CHARM, a multicultural benchmark of 40 real and fictional characters drawn from five cultural-linguistic regions, and validated by native reviewers. It probes two boundary types, Temporal (historical vs. modern) and Cross-Universe (entities outside a character's narrative or historical universe), using abstention-enabled multiple-choice questions. We propose a two-stage evaluation that separates Boundary-Awareness (explicit recognition that a query is out of scope) from Boundary-Compliance (abstention when answering concrete questions). Evaluations across six LLMs show that hallucination is driven predominantly by compliance failures. Models frequently acknowledge that a query lies outside the character's knowledge yet still provide factual, out-of-character answers. By re-posing the same questions to the target character, we confirm that a large fraction of these cases are verified parametric overrides; the model stores the relevant fact but fails to suppress it. We also observe systematic cultural variation in these failures, consistent with imbalances in how characters from different regions are represented in model knowledge.

Online planning under uncertainty remains a fundamental challenge for robotic systems operating in partially observable environments with high-dimensional state spaces. While sampling-based POMDP solvers enable approximate decision-making in large or continuous domains, their performance degrades as belief dimensionality increases due to the high variance inherent in Monte Carlo-based estimation. In this work, we extend the Rao-Blackwellized online POMDP (RB-POMDP) framework to improve its generalizability in high-dimensional settings through hybrid continuous-discrete belief representations. By analytically propagating uncertainty associated with marginalized state components during tree-based planning, the proposed approach reduces sampling-induced variance in value estimation. We demonstrate the effectiveness of this framework in a robotic search-and-rescue task by integrating it with FastSLAM 2.0. Experimental results show that the proposed planner achieves higher cumulative rewards using significantly fewer particles and planning simulations than purely sampling-based methods under equivalent computational budgets. These results suggest that structured high-dimensional robotic problems admitting tractable sufficient statistics can be effectively leveraged within the RB-POMDP framework for computationally feasible online decision-making.

We construct permutation polynomials of the form x + \gamma Tr^{q^2}_q(h(x)) over the finite field F_{q^2}. More precisely, we present two families ofpermutation polynomials whose coefficients range over all elements of the field rather than being restricted to a proper subfield. We also compute the compositional inverses of both families. Our techniques involve the evaluation of certain Kloosterman sums together with the analysis of some equations over finite fields, and we believe these can be of independent interest.

Port-Hamiltonian (PH) theory offers a passivity-based framework for grid-forming control, yet conventional dispatchable virtual oscillator control (dVOC) does not naturally admit a dissipative PH realization, since its amplitude egulation, synchronization, and power dispatch are inherently coupled without a unified energy interpretation. This paper formulates the outer-loop dynamics as a dissipative PH system, thereby unifying amplitude regulation, synchronization, and power dispatch within one energy structure. The formulation rests on the key property that logistic-type radial regulation, characterized by an inherent saturation-like nonlinearity, permits an exact gradient decomposition compatible with the quadratic energy storage. On this basis, a unified shaped Hamiltonian is constructed, which encapsulates both amplitude restoration and power dispatch. Radial gain matching derived from this Hamiltonian yields explicit closed-form arameter inequalities that guarantee almost-global asymptotic stability and local exponential convergence. Moreover, tuning the power-error weighting coefficient actively shapes the energy landscape, thereby eliminating the undesirable low-voltage power-flow solution from the stationary set and ensuring convergence to the desired high-voltage equilibrium point. The Hessian singularity condition further provides the critical weight threshold that guarantees uniqueness of the high-voltage equilibrium. Numerical simulations validate the proposed method.

Continuous cellular automata such as Lenia spontaneously produce lifelike, self-propelling patterns, including the Orbium glider, which travels in a straight line while pulsing through a fixed breathing cycle. Collision-based logic, where moving patterns compute by colliding, is established in discrete cellular automata and continuous physical media. Within continuous cellular automata, computation has so far been trained into the rule rather than emerging from collisions, and whether a fixed-rule automaton like Lenia can support general collision-based computation remains open. This paper constructs an INHIBIT gate from collisions of the Orbium glider. Of the patterns searched across four continuous-CA rule types, the Orbium glider is the only one shown to survive a collision with both copies intact. A control glider deflects a signal glider off its output line, so the output carries a signal only when no control is present. The gate blocks across all twenty-four phases of the breathing cycle and nine integer offsets of the control. Two such gates in series, with one signal line and two controls, compose into an AND-NOT chain, correct on all eight input combinations. By contrast, routing a signal beyond that single chain is undemonstrated. A deflected signal is not restored to a fixed landing position, and no reusable absorber for the surviving gliders was found. The immediate open question for collision-based computation in Lenia therefore narrows from whether a gate exists to whether a deflected signal can be delivered to a downstream gate, the next requirement for composing the gate beyond a single straight chain.

Inference cascades cut cost by answering most queries with a cheap model and escalating a hard tail to a frontier model that acts as verifier. A natural extension closes the loop: fine-tune the cheap student on the verifier's rejections so the escalation rate, and cost, fall each round. We measure this loop on real LLMs and report four findings. First, the verifier's blind spot, the fraction of the student's wrong answers it accepts, is large and moves adversarially: it grows with student capability ($\beta$ from 0.12 to 0.55 as the student scales 0.5B to 32B) and shrinks with verifier capability, so it is worst in the cheap-student, cheap-verifier regime cascades exist to create. Second, buying it away returns the saving: a frontier verifier drives $\beta$ to about 0.05 but then escalates on 46% of hard-MATH queries against a 39% true error rate, paying the frontier price on nearly half of all traffic. Third, naive corrective fine-tuning on the verifier-rejected tail does not improve the small student but degrades and ultimately collapses it, across every teacher we tried (cross-family and same-family), so at this scale the self-improving loop is self-defeating. Fourth, through all of this the cascade's own dashboard, every metric computed through the verifier, reads a flat 3% error while true delivered error swings up to 32%: the system is blind to its own degradation by construction. We then give the theory that explains the blindness, a two-population conservation law, $\epsilon_\infty \lesssim q_0 \beta_0$, under which every in-loop metric improves while true quality does not, and a synthetic study that validates the mechanism. The practical conclusion: the reliability of a self-improving cascade cannot be read from any metric computed through its own verifier.

Multimodal coding and editing systems must map a visible or semantic referent to the exact executable object that can be edited. A wrong reference may select a valid but incorrect DOM node, SVG element, graph endpoint, hierarchy member, or table cell, while final execution success alone does not reveal the source of the failure. ExBind isolates this visual-to-executable correspondence layer as a controlled diagnostic benchmark between semantic localization and action execution. It samples representation-independent latent binding instances and compiles them into SVG, DOM, canvas, tree, graph, and table cases with deterministic mappings to executable references. Models output only a strict reference; the evaluator maps predictions back to latent structure and scores structural constraints without requiring reasoning traces. The release contains a 250-case broad suite, a disjoint 240-case targeted suite, and 50 paired latent groups. Qwen2.5-VL-3B achieves 98.4% candidate validity but 76.4% exact accuracy, while Qwen3-VL-4B achieves 100.0% validity and 98.8% exact accuracy. In the targeted table suite, all Qwen2.5-VL-3B residual errors are valid correct-row/wrong-column selections. Candidate-order perturbations change case-level outcomes while preserving this error pattern. ExBind is designed for controlled diagnosis rather than population-scale ranking or end-to-end editing evaluation. Code and benchmark records are available at https://github.com/Daerwang2020/Exbind and https://huggingface.co/datasets/Ziqianwwww/ExBind.

Looped Transformers increase effective depth by iterating a shared block of layers, but most evaluations compare at fixed model size, conflating architectural advantage with extra FLOPs. We study looping on Mixture-of-Experts Transformers while closely matching per-token FLOPs, total non-embedding parameters, and KV cache. Through a series of ablations, we arrive at a recipe we call SMELT (Sparse MoE Transformer, middle layers Loop Twice), which loops the middle half of layers twice while matching the unlooped Baseline on all three budgets. We scale SMELT across four sizes up to 54B non-embedding parameters and fit a separate Chinchilla-style scaling law for each architecture. SMELT's loss drops faster with compute, saving 6.8–18.0% of training FLOPs on the compute-optimal frontier. The advantage transfers to downstream benchmarks beyond what validation loss predicts, is largest on Code, and grows with sample length and the number of in-context examples. Mechanistic analysis shows that the second visit reduces the attention sink and redirects mass toward content-relevant tokens, an inductive bias that may underlie the observed performance gains. These results show that looping can improve Transformers even under budget matching, offering a practical recipe that turns depth reuse into measurable gains.