Papers

Vector databases use hashing to partition data across "shards," logical units for distributed execution. This placement, however, destroys semantic locality, forcing each query into scatter-gather limited by the slowest shard. Vector-space clustering can help, but scientific evidence is often connected by factual relations that do not align with embedding distance. We present COMPASS, a framework that uses a knowledge graph (KG) to determine data placement and query-time shard selection. COMPASS detects communities, splits oversized communities, inserts embeddings by subject entity, and routes queries to a small set of shards. Across four biomedical KGs, our method searches only 13-18% of the corpus while preserving broadcast recall and recovering up to 2.6x more multi-hop evidence than an embedding-based baseline. On 15 HPC nodes, COMPASS sustains 7.9x higher throughput with lower tail latency than hash-based broadcast. These results show that KG structure provides a compact complement to embedding geometry for scalable vector search.

Public data exchange infrastructures, such as X-Road, are essential to data-intensive digital public services and the EU digital economy. Their next generation requires advanced data usage control (DUC) mechanisms that address emerging requirements, including privacy by design, to mitigate privacy threats inherent in large-scale automated data exchange. Although research within the individual disciplines already addresses many of the emerging requirements, significant practical challenges arise at the intersection of the disciplines. This paper analytically examines the practical challenges and requirements of data usage control in public data exchange infrastructures. We examine the Estonian X-Road system and emerging use cases through an interdisciplinary lens that integrates formal verification and monitoring of usage control, requirements engineering, and digital governance research. We formulate key requirements and propose a conceptual vision for developing usage control in public data-exchange infrastructures to enable future digital-governance systems to comply with privacy requirements by design.

We develop a discrete design theory for cofactor reference forms that preserve the complete finite-element elasticity tangent. A fixed continuous potential generates the coefficient, while the variation space and quadrature determine compatibility. We classify the exact local potential spaces for all-order tensor elements, anisotropic spaces, cubic serendipity and quadratic tetrahedra. At a stress-free Mooney-Rivlin state with a common positive material-coefficient sum, a convex loss expresses the attainable quadrature-point material margin. The optimum is attained; under stated space and sampling assumptions, its ideal value is attained precisely when the physical identity potential is compatible. A conforming fixed-candidate example proves a strict margin loss, linear in curvature amplitude and uniform in cell size, under displacement enrichment. Explicit positive candidates accompany this restriction. A Gram identity gives the sharp constant-coefficient defect at a prescribed margin in the stated norm. Compatible and reference-defect-corrected forms retain the original tangent through a deformation-independent external-boundary Hessian; equal potential traces give equal exact assembled gauges. Quadratic tensor geometry admits a stress-free physical coercivity bound uniform in mesh size and displacement order under explicit material, shape, quadrature and boundary assumptions. Certified curved tetrahedral shear families with zero recovered cell pressure establish finite-deformation regimes, while a loaded pressure-curvature counterexample identifies their limitation. The theory and observations distinguish local exactness, material certificates, complete-reference coercivity and unchanged equations.

This paper develops an integrated framework for estimation, prognosis, and reliability assessment of stator inter-turn faults in induction motors. The fault is characterized by its severity, defined as the fraction of short-circuited turns, and its spatial orientation. A geometric fault model shows that the resulting output signature is affine in the fault severity and exhibits a fundamental spatial periodicity. Exploiting this structure, an augmented-state particle filter jointly estimates the nonlinear electromechanical state and the unknown fault severity while quantifying posterior uncertainty. The estimated degradation is then propagated using four prognostic models: linear trend, Holt exponential smoothing, Bayesian degradation, and particle-based forecasting. Their predictions are connected to threshold-crossing remaining useful life (RUL), first-passage reliability, degradation-dependent hazard reliability, and a Weibull lifetime benchmark, thereby providing both deterministic and probabilistic health assessments. Numerical results demonstrate accurate online fault estimation and output reconstruction, characterize the effects of degradation pattern and prediction horizon on prognosis, and show consistent reliability and threshold-crossing predictions. Moreover, estimation accuracy remains comparable across the distinct fault orientations induced by the spatial periodicity. The resulting framework provides a unified connection from physics-based inter-turn fault modeling to online diagnosis, degradation prognosis, and reliability assessment.

Volume quadrature can change nonlinear finite-element constraints while preserving their reference-state derivatives. We connect an explicit determinant defect to feasible-set geometry and singular mechanical response. For affine tensor elements of coordinate degree $p\ge3$ with $n\ge p+1$ Gauss points per coordinate, determinant volume is exact precisely when $2n\ge3p$. Below that threshold we construct a boundary-fixed cubic defect at every order. The same directions yield a full-space cube-root residual–distance bound under explicit cell-support, coefficient and physical-norm assumptions, with mesh-uniform upper constants at fixed order. With all cell-pressure equations retained, the volume Jacobian gains rank at nearby feasible states despite agreement through second derivatives at rest; the cube-root exponent is sharp on each fixed mesh. A general localized-minimum theorem shows that the first reduced compatibility term contributes its weighted square to the leading energy in a joint small-load, large-bulk limit. Cubic and quadratic defects therefore produce sextic and quartic terms. The full cubic-element interior space has an exact normal form and sharp local error exponents. Curved quadratic tetrahedra supply the second-order contrast, a sparse rational witness and an exact four-Jacobian volume formula. Finite-strain tensor calculations illustrate normalized response separation, with explicit stationary-point, extreme-bulk and pressure-recovery qualifications. The constructive correction preserves exact cell volumes, so quadrature feasibility remains distinct from physical volume preservation.

Speech-to-speech LLMs like Moshi, and its derivative PersonaPlex, can listen and speak concurrently through full-duplex generation. However, they can begin speaking inappropriately during prolonged user silence: under digital-zero input, Moshi and PersonaPlex initiate speech in 30% and 27.5% of five-minute continuations, respectively. What causes this spurious speech? We investigate two hypotheses: either repeated sampling selects speech despite persistently low onset probabilities, or self-conditioning on nonspeech outputs causes an abrupt spike in onset probability. We find that, at every observed onset, speech probability spikes by over nine orders of magnitude in one 80-ms frame, supporting the latter hypothesis. Then, to suppress these onsets without blocking genuine responses, we ask a causal counterfactual question: is the model responding to user speech, or would its next-token distribution remain similar if the preceding user input were muted? Accordingly, we suppress onsets whose distributions change little under this intervention. Under realistic microphone noise, our method suppresses spurious onsets, while preserving genuine responses: one-sided 95% lower confidence bounds are 98.68% and 98.82% for Moshi, and 96.90% and 99.25% for PersonaPlex. Our inference-time method runs in real-time without retraining, with 95th-percentile decision time below 61 ms, within the 80-ms frame budget. Our code is available at https://github.com/KentoNishi/icassp27-spurious-onsets.

We demonstrate that training LLMs with RL does not improve performance equally across a dataset. RL shows large improvements on easy problems that an LLM is already good at solving, but small improvements on hard problems. We call this the Matthew Effect in RL for LLMs, after the phenomenon of cumulative advantage from economics and network science summarized as "the rich get richer". The naive explanation is that hard problems require more compute to find a solution. We argue that modern RL methods are exacerbating the issue by wasting too much compute on easy problems and instead should dynamically reallocate how they use compute. We introduce Never Give Up (NGU), a simple adaptive sampling method that keeps generating samples for a problem until one is correct. By leveraging asynchronous RL, this naturally uses fewer samples to filter out easy problems and allocates more compute to solving harder problems. We investigate the design choices that affect NGU, such as off-policy robustness, and develop a set of best practices. On the math benchmark Deepscaler, NGU improves performance per compute, especially on harder problems. On a recent coding task, Manufactoria, standard GRPO with a per-test reward fails to fully solve problems that have a range of easy and difficult tests. NGU iteratively improves, solving harder and harder tests, until it learns to fully solve coding problems.

We give a deterministic algorithm for online inverse linear optimization with regret $O(d)$, uniform in the horizon and $O(d^{2})$ time per round. A bound of this order was obtained recently by Dewasurendra, settling a question of Gollapudi et al.\ and of Oki and Sakaue, but by an improper rule that enumerates covers at every scale and costs $T^{\Theta(d)}$ a round; ours is the first efficient such bound and the first proper one. We build on the variable-metric framework of Sakaue et al., adding a self-normalized rank-one update, and we replace the $\log\det$ potential by the trace power $\tr(H^{-1/2})$, which is bounded outright and removes the $\ln T$. The bound also holds against an expert that does not optimize, and we give corruption-robust and rank-adaptive variants, and an application to convex minimization.

As compute scales, models evolve, and training algorithms advance, our ability to explain the increasingly powerful AI systems they enable is eroding. To help safeguard interpretability, we introduce a specialized training algorithm (MACCHIATO) that jointly constructs (i) an explicitly structured $\operatorname{ReLU}$-MLP from partial truth-table observations and (ii) an explicit Boolean circuit over signed literals with $\{\operatorname{AND},\operatorname{OR},\operatorname{XOR}\}$ gates certifying what its subnetworks compute and how they compose. Intuitively, we iteratively project the residuals of a Boolean function onto low-dimensional $\{\operatorname{AND},\operatorname{OR},\operatorname{XOR}\}$-circuit classes and exactly compile the resulting circuit into a $\operatorname{ReLU}$-MLP; we combine $\operatorname{ReLU}$-MLP circuit compilation, ESPRESSO logic minimization, and influence-based variable selection. Roughly speaking, our interpretability certificate is complemented by a statistical guarantee: under the theorem's influence-recovery conditions, if each of the $m$ stage-wise residuals depends on at most $\log_2(B)$ bits, a sample-splitting variant of our algorithm trained on $T$ observations returns a six-layer $\operatorname{ReLU}$-MLP (counting the input layer) of width $\mathcal{O}(mB)$ with truth-table error $\mathcal{O}\bigl(\sqrt{m(B+\log(m/\delta))/T}\bigr)$. On synthetic random-junta tasks, our networks outperform depth- and hidden-width-matched Adam-trained MLPs in several data-sparse or projection-aligned regimes, while the trained ReLU-MLPs are stronger in others. Moreover, in our explicit PyEDA truth-table implementation, the iterative procedure completes in regimes where flat ambient-dimensional ESPRESSO exceeds the three-hour computational budget.

Scientific research is a continuous process that emphasizes inheritance. Methods developed by predecessors are often expanded upon by new researchers to explore more novel and in-depth scientific questions. However, the change of lab staff, such as student graduation, leads to a lack of personnel capable of replicating methods. Methods that have been developed with significant effort and resources cannot be continued. To address these limitations, we propose LabAgent, a reproduce and discovery harness tailored for a lab's continuous work. LabAgent employs two mechanisms to guarantee that all skills can be executed and verified and to record the corrective methods and experiences, allowing for direct correction or avoidance of similar errors. We applied LabAgent to drug property prediction, biomedical problem analysis, protein variant effect prediction, and statistical genetics in life science domains. LabAgent ranks first over commercial generalist agents in every domain, and demonstrates accurate reproduction of a published figure. Overall, these results demonstrate that LabAgent can effectively integrate and reasonably expand laboratory knowledge.

Large Language Model (LLM) agents offer a promising path toward autonomously managing long-term physical tasks without human intervention. However, physical tasks require agents to continuously observe the environment, make consequential actions, and remain effective as the environment changes. Existing approaches either require substantial data and retraining, or primarily focus on agents operating in the virtual world. In this work, we explore the feasibility of building a self-adaptive physical AI agent that manages long-term physical tasks in a zero-shot manner and adapts to environmental changes without human intervention. We design a multi-agent framework that integrates planning, tool calling, observation, and verification, and evaluate it on agricultural tasks against reinforcement learning (RL) agents under different weather patterns. Our results show that zero-shot LLM agents can achieve comparable management outcomes to RL agents under the same weather pattern and adapt more effectively than RL when evaluated under a shifted environment, highlighting a promising path toward self-adaptive physical AI agents.

Given a grid-sampled 1-Lipschitz function $f:\mathbb{R}^3 \rightarrow \mathbb{R}$ (such as a signed distance function) sampled on a regular grid, we determine the sharp infimum $\sigma_\star$ such that for every $\sigma > \sigma_\star$ the true zero-level set $f^{-1}(0)$ is contained in the marching cubes mesh for $f = \sigma$. For grids with spacing $h$, the sharp infimum is $\sigma_\star = \frac{\sqrt{3}}{2} h$. This result is proven despite the planar marching cube faces potentially lying on either side of the level set of the trilinearly interpolated approximation of $f$ (which shares the same infimum). The proof instead constructs a convex lift of the corner samples and shows that the high-side portion of each marching-cubes cell lies in the projection of its $\sigma$-superlevel portion, which cannot contain a point of $f^{-1}(0)$.

Increasing conjunction frequency in low Earth orbit places growing pressure on spacecraft operators to determine not only whether an encounter requires mitigation, but whether sufficient information is available to commit to a maneuver. This work formulates this information-action tradeoff as a belief-space planning problem for conjunctions between a maneuverable spacecraft and an unmaneuverable secondary object. The planner represents the uncertain orbital states as Gaussian beliefs and uses a chance-constrained belief-space Monte Carlo tree search framework to reason over possible future tracking updates before time of closest approach (TCA). A terminal chance constraint limits the probability of reaching TCA above a prescribed collision-risk threshold, allowing the planner to wait for informative tracking while intervening when deferral becomes too risky. We evaluate the approach on eight historical conjunctions from NASA's Conjunction Assessment Risk Analysis dataset. By varying the secondary-object measurement quality and tracking cadence, we generate a total of 96 distinct evaluation scenarios. Across the evaluated conditions, the planner reaches TCA without maneuvering in approximately 40% of episodes while maintaining no terminal collision-risk violations. In contrast, fixed-time rule-based maneuver policies resolve more encounters without maneuvering when intervention is deferred closer to TCA, but at the expense of increasing terminal risk violations. The fraction of episodes reaching TCA without maneuvering depends strongly on tracking quality and measurement cadence, ranging from 76% under accurate, frequent measurements to approximately 18%-20% under the poorest tracking conditions. These results show that tracking quality and frequency are not only inputs to collision-risk estimation: they can determine when intervention becomes necessary.

Training diffusion models with multiple rewards requires distinguishing user preference from reward informativeness. User preference determines how much each reward should contribute to the overall objective; reward informativeness determines when its feedback is useful during denoising. Some rewards can meaningfully evaluate a sample as soon as global structure emerges, but others become informative only when the sample is nearly clean. To address both questions jointly, we propose ReCAST (Reward Credit ASsignment across T}imesteps), the first method, to our knowledge, for per-reward, timestep-dependent credit assignment in diffusion reward fine-tuning. ReCAST separates user preferences from temporal allocation through a reward-by-timestep weight matrix $W$, whose row sums match the user-specified reward budgets $\lambda$, while its column sums are equal, assigning the same total weight to each denoising step. Under these marginal constraints, ReCAST allocates weight according to each reward's informativeness, quantified by its Rényi discriminability gain at each step. These gains telescope to the total discriminability between the reward-induced positive policy and the current policy, providing a basis for temporal credit assignment. We evaluate ReCAST by training SD3.5-Medium under two distinct four-reward settings, each across five reward budgets $\lambda$. ReCAST improves the training rewards in one setting and matches them in the other, improves every held-out judge in both, and is preferred by an independent LLM-as-a-Judge. Together, these results show that ReCAST yields improvements that generalize beyond the training rewards and support its core principle: assigning each reward greater weight at the denoising timesteps where its feedback is most informative.

LLM judges are increasingly used to evaluate and improve AI-generated outputs, yet their reliability for complex professional work remains unclear. We study this problem through Vibe Patenting, an end-to-end patent-drafting testbed for AI-agent evaluation. A separately-invoked LLM judge evaluates generated patent drafts and provides structured feedback for iterative revision. Across multiple inventions and drafting-agent configurations, judge-guided revision consistently improves judge-assessed quality, while unguided revision tends to saturate. Notably, iterative judge feedback enables a low-reasoning agent to approach the performance of a substantially more expensive high-reasoning agent. Stronger models and increased reasoning generally improve judge-assessed drafting quality, while domain-specific agentic workflows provide further gains. We validate the judge against independent evaluation by a professional patent attorney and find meaningful but strongly metric-dependent agreement and systematic calibration differences. These results highlight both the utility and limitations of LLM judges as evaluators and optimization signals for complex professional workflows.

Real-time motion planning must execute under strict latency and energy constraints on resource-limited platforms. This paper presents Arborist, an algorithm-hardware co-design framework that accelerates Fast Marching Tree (FMT*) motion planning through synergistic algorithmic and architectural innovations. At the algorithm level, Arborist introduces safe multi-tree expansion with global competition management and cross-tree optimality checking to expose inter-tree parallelism while preserving path quality. At the architecture level, our proposed ArboristAccel integrates an Arborist Toolbox that provides hardware support for inter-tree parallelism; a KD-tree Subsystem for high-throughput near-neighbor search; a Parallel-Query Retrieval module that balances multi-bank memory accesses to alleviate system-level memory bandwidth bottlenecks; and a Look-ahead Planning Engine that exploits intra-tree parallelism while preserving path quality by in-order commit. Synthesized in 16 nm and operating at 500 MHz, ArboristAccel sustains more than 1000x speedup over CPU across all workloads, peaking at 2800x on 2D-Maze, and delivers geomean improvements of 8.8x speedup, 4.6x area efficiency, and 3.1x power efficiency over an FMT* ASIC baseline. The corresponding factors relative to a state-of-the-art RRT*-based MOPED accelerator are 8.6x, 1.0x, and 1.8x, at comparable path cost. To the best of our knowledge, ArboristAccel is the first dedicated hardware accelerator for FMT*-based motion planning.

In a 2021 paper, Jankiewicz, Norin, and Wise asked whether there exists a finite $4$-connected graph of girth at least four and nonnegative Charney–Davis curvature such that no $4$-connected ordinary subgraph admits a legal system. We construct such a graph by starting from the hexagonal prism and attaching three $K_{3,4}$-based caps along pairwise disjoint induced $4$-cycles. The key structural input is a restriction theorem showing that a legal system on an induced-$4$-cycle amalgam restricts to each side, so the obstruction carried by the negatively curved prism survives the attachments. The resulting $33$-vertex graph is $4$-regular and $4$-connected, has girth four and Charney–Davis curvature one, and, by $4$-regularity, is its own unique $4$-connected ordinary subgraph.

Differential privacy provides a mathematical framework for guaranteeing privacy for sensitive data. In quantum information processing, the interaction of privacy constraints with quantum resources such as entanglement remains a question of interest. Given that the utility of many protocols, and often the presence of a quantum advantage, relies on quantum resources such as entanglement, it is crucial to understand when a privacy requirement for a quantum channel is compatible with the channel's ability to preserve entanglement. We study this question for quantum local differential privacy (QLDP). Our main result shows that every $\varepsilon$-QLDP channel with a $d$-dimensional input is entanglement-breaking whenever $\varepsilon\leq\log\frac{d}{d-1}$. We also prove an approximate version for $(\varepsilon,\delta)$-QLDP, where channels in the same high-privacy regime are close in diamond norm to an entanglement-breaking channel. We further prove a composition result for a collection of private quantum channels having entangled inputs and global measurements in the high-privacy regime. Finally, we apply our results to private quantum learning theory. We prove that any learning protocol using arbitrary quantum memory on copies of the output of an entanglement-breaking channel can be simulated by a protocol that measures the corresponding unprocessed input copies one at a time while storing only classical information. Combining this result with our high-privacy entanglement-breaking theorem, we show that under sufficiently private local noise, a learning protocol with quantum memory for purity testing and bipartite product testing is subject to the sample complexity lower bounds for protocols with single-copy measurements on the noiseless tasks. We also obtain stronger sample complexity lower bounds when a single highly private channel acts on the entire multipartite input.

Specifications face a basic tradeoff: leave details out, and important questions go unanswered; record every detail separately, and the specification becomes large and prolix. We investigate whether a language model can serve as a compact, living specification oracle by learning facts about a target and answering questions about it directly. We compare two ways of storing the learned facts: external text notes and changes to the model's weights. Across four families of 596-fact worlds and two Qwen2.5 model sizes, weight-only oracles benefited substantially more from structure: with the 7B model, their accuracy integrated across storage capacities was 18.5 percentage points higher on structured than unstructured worlds, compared with 1.1 points for note-sheet oracles. This advantage came at a substantial storage cost, with the smallest adapter requiring approximately 175 KiB compared with a maximum note budget of 16 KiB. Adapted weights therefore exploited latent structure more successfully, while external notes required substantially less object-specific storage.

We introduce CVSS-X, a large-scale synthetic speech-to-speech translation corpus that extends CVSS by reversing the translation direction. While CVSS translates from 21 languages into English, CVSS-X enables translation from English into 28 target languages spanning 12 language families. The corpus comprises approximately 240,000 parallel speech pairs per language, totaling over 16,000 hours, eight times larger than CVSS. We provide two variants: CVSS-X-C with two canonical voices per language, and CVSS-X-T with cross-lingual voice cloning, both fully generated. Evaluation shows comparable translation quality to CVSS with consistent performance across typologically diverse languages. Combined with CVSS, this enables research on bidirectional and multilingual speech-to-speech translation. The code is available at https://github.com/ErmisAI/XVSS-X and the dataset under CC-BY-NC 4.0 license at https://huggingface.co/datasets/lgris/XVSS-X.

Phase unwrapping recovers a continuous phase map from measurements known only modulo 2{\pi}, in the presence of noise and discontinuities. We propose a tile-based method that combines local spectral reconstruction, reliability-ordered tile joining, and residual-weighted multipath averaging. Each tile is reconstructed by a discrete cosine transform (DCT) least-squares solve, and robust measurements along tile borders guide their joining along a spanning tree. We repeat this reconstruction on shifted tile grids and on rotated and reflected copies of the image, which changes tile borders and joining paths. A residual-weighted average combines the resulting reconstructions, which are computed in parallel and averaged on the fly without being stored. We also introduce the corrected circular rewrapping error, a congruence measure: after removing a global phase offset, it measures how far the rewrapped reconstruction is from the observed phase. It applies to least-squares and path-based methods, even without an unwrapped reference. Used with reference-based errors, it separates agreement with the measurement from recovery of the correct number of cycles. Experiments on radar-interferometric, magnetic-resonance, holographic, and constructed phase images compare our method with seven reference algorithms. Among the methods that complete every test image, ours has the lowest mean fraction of pixels with an error above {\pi}. On the images all methods complete, it lies on several speed-accuracy Pareto fronts. Error maps also reveal isolated cycle errors that average scores or congruence alone can hide.

Coordinated manipulation is collective: plausible accounts can jointly distort ratings, rankings, and engagement. Existing defenses primarily construct evidence from identities, graphs, content, or co-activity. We introduce an aggregate-first evidence layer that treats distortion of a context-level outcome distribution as the primary evidence object. The engine observes only a histogram, count, resolution, and reference distribution; identities are withheld until interval evidence is fixed. Because raw discrepancies have positive finite-sample expectation, we subtract a matched null expectation to obtain signed evidence and account for reference uncertainty. Participation logs then accumulate these fixed increments across accounts. We characterize matched-exposure divergence, bound self-influence, establish finite-horizon separation, and derive an exact linear reuse law for paired contexts. We evaluate the mechanism with controlled rotation experiments and paired counterfactual interventions on historical Amazon review streams. Historical reviews provide the behavioral background; synthetic identities provide known coalition membership, and exact clean twins provide counterfactual controls. In a fixed-attack sweep against historical non-donor comparison accounts, reassigning the same manipulated events across identities with increasing reuse raises account-score ROC-AUC from 0.500 to 0.797. With activity- and exposure-matched clean twins, frequency is at chance while counterfactual attribution achieves ROC-AUC 0.744. Under a mean-preserving shape intervention, Wasserstein-1 and Jensen-Shannon evidence achieve ROC-AUC 0.909 and 0.967, while frequency and mean-based attribution remain at chance. Aggregate evidence complements repeated co-activity, improving mixed-mechanism ROC-AUC from 0.750 to 0.874 with a simple untrained combination.

When we speak of recursive self-improvement (RSI), are we speaking of a phenomenon, a mechanism, or a prospect? Towards autonomous and evolving intelligence, RSI is being claimed at many scales, while no single framework that formally describes these emerging instances exists. Its counterpart in the classical realm, iterative policy improvement, is characterized by generalized policy iteration (GPI), a framework of broad applicability with well-understood theoretical properties, but only where the update principle and the evaluation base lie outside the agent. In this paper, we propose Generalized Agent Iteration (GAI), a formal framework that describes iterative policy improvement and RSI as two cases of a single learning paradigm. GAI defines the agent as a configuration of modifiable components within a system and models the learning process as a cycle of agent evaluation and agent improvement. Two pivotal dials then distinguish the instances, in particular the systems that self-identify as RSI in recent work: whether the improving mechanism is part of the agent and whether the standard it is measured against is grounded outside it. The former dial delineates the boundary between GPI and RSI, and the latter determines a system's polarity as anchored, goal drift, or fully self-referential. Moreover, we use these coordinates to place existing systems on the same two axes and make the defects of recursive self-improvement statable one condition at a time. We see this paper as a first step toward exploring a formal characterization of RSI that rests on the classical account, makes existing systems comparable, and provides a principled basis for analyzing and designing new ones.

Behavioural equivalences, such as language equivalence and probabilistic bisimilarity, are fundamental techniques for reducing the size of probabilistic models. However, these equivalences are sensitive to the precise values of transition probabilities, making them unsuitable in applications where probabilities are subject to approximation. Motivated by settings in which the support graph of a labelled Markov chain is known but the transition probabilities are uncertain, we study robust variants of these equivalences. We introduce universal (perturbation) equivalence, which captures a variant of equivalence that is resilient to all perturbations of transition probabilities: two states or distributions are universally equivalent if they remain equivalent under every assignment of transition probabilities consistent with the support graph. We also consider the dual notion of existential (perturbation) equivalence, which holds whenever there exists an assignment of transition probabilities that yields equivalence. We establish that, for states, universal language equivalence coincides with universal probabilistic bisimilarity and develop a characterisation that yields a polynomial-time partition refinement algorithm. We implement the algorithm and demonstrate experimentally that it is effective as a technique for robust model reduction. We further show that universal language equivalence for distributions is closely related to the state case, and prove NL-completeness of deciding universal equivalence for both states and distributions. We prove that, for states, existential language equivalence coincides with existential probabilistic bisimilarity and deterministic witness transition functions always suffice, leading to an NP-completeness result. In contrast, we show that existential language equivalence for distributions is complete for the existential theory of the reals.

In rendering, a camera is a sampling operator that maps each finite pixel to a bundle of rays. Different camera models change the geometry of this bundle, thus making a unified and faithful rendering formulation challenging. Consequently, Gaussian ray tracing supports generic cameras (with optical center) through their inverse ray mappings, yet typically reduces every pixel to a single center ray. This ignores the camera-dependent pixel footprint, causing aliasing under minification, while unconstrained Gaussians expose unsupported frequencies under magnification. We present ConeGaussian, a camera-model-agnostic anti-aliasing framework for Gaussian ray-based rendering. Instead of defining the pixel filter on a camera-specific image plane, ConeGaussian constructs an anisotropic footprint directly from neighboring rays produced by the camera's native inverse mapping. We derive a closed-form response under a locally linear, depth-local, moment-matched approximation of the finite pixel footprint, while the same geometry defines a per-Gaussian training-frequency floor. Notably, by construction, our filtering principle can be used unmodified across calibrated central camera models and multiple Gaussian ray-rendering backbones. Additionally, unlike in mip-splatting, our scene-space frequency floor and filtering enable trivial composition at render time, allowing us to remove excess blurring. On pinhole and strongly distorted fisheye captures, ConeGaussian consistently improves two distinct ray-based backbones, by up to 4.3 dB at 1/8 resolution, and reduces fisheye LPIPS by 30% where perspective screen-plane footprint formulations are not directly applicable.

Multi-objective Bayesian optimisation (MOBO) is a sample-efficient approach for optimising expensive black-box functions with multiple objectives. In MOBO, the goal is to adequately approximate the Pareto front; that is, to obtain a high-quality solution set with 1) good convergence (closeness to the Pareto front) and 2) good diversity (spread across the Pareto front). Existing MOBO methods typically aim to accomplish these two tasks simultaneously, i.e., driving the search towards the Pareto front while maintaining a diverse set of nondominated solutions, such that the solutions, ideally, can gradually approach the entire front. When sufficient search budgets are available, this approach is effective. However, considering both convergence and diversity throughout the search is not easy and requires careful design. Under very tight budgets, there may not be enough solutions generated to be able to simultaneously approach the entire Pareto front. To address this issue, this paper proposes a converge-then-diversify (CTD) approach that decouples convergence and diversity into two stages. In the first stage, CTD focuses on convergence, aiming to quickly drive the search toward a single point on the Pareto front. In the second stage, CTD focuses on diversity, aiming to spread solutions across the front. We present two simple instantiations of CTD by using widely adopted acquisition functions in the area. Experimental results show that, across all 446 pairwise comparisons, CTD statistically outperforms state-of-the-art methods in 72.9% of the cases, performs equivalently in 21.1%, and is statistically worse in only 6.1%, with the advantage being particularly evident in settings with very tight evaluation budgets or in high-dimensional problems.

Mixture-of-Experts (MoE) has become a widely adopted architecture for Large Language Models (LLMs), as it improves model capacity while limiting computational overhead through sparse expert activation. This property makes MoE-based LLMs particularly attractive for resource-constrained distributed environments. However, federated fine-tuning of MoE-based LLMs remains challenging under heterogeneous client data. Since clients often correspond to different task preferences, directly aggregating their local updates may weaken expert specialization and introduce conflicting update directions on shared experts. To address these challenges, we propose FedTAR, a task-aware federated fine-tuning method for MoE-based LLMs. FedTAR establishes the association between local updates and task preference via routing outputs. Specifically, we apply Singular Value Decomposition (SVD) to both routing features and local updates to extract low-dimensional task coordinates and update directions. Based on the task coordinates, FedTAR performs intra-cluster aggregation among clients with similar task preferences and inter-cluster aggregation across different task groups. The aggregated update is then reconstructed through the learned task-to-update mapping, ensuring that the final update remains aligned with task-specific optimization directions. In this way, FedTAR preserves expert specialization and mitigates destructive interference among heterogeneous clients. We evaluate FedTAR on four benchmark tasks under different non-IID settings. Experimental results demonstrate that FedTAR consistently outperforms strong federated fine-tuning baselines and achieves state-of-the-art performance.

Deep joint source-channel coding (DeepJSCC) transmits learned semantic features efficiently but can leak sensitive attributes such as gender, race, or speaker identity. We propose LEAPSC (LEACE-in-the-loop privacy for semantic communication), whose core contribution is the integration of in-loop least-squares concept erasure (LEACE) within a variational information bottleneck (VIB) encoder. By periodically refitting the projection operator during training, LEAPSC couples the encoder dynamics to the erasure mechanism, driving attribute-conditional mean differences toward zero within each task-label group on the fitting sample. Additional components, namely conditional value-at-risk (CVaR) tail-sensitive privacy, feature-wise linear modulation (FiLM) signal-to-noise ratio conditioning, and Lagrangian dual ascent, improve robustness across channel conditions and over the high-leakage tail of samples. On CelebA, FairFace, and Google Speech Commands, LEAPSC reaches task accuracy of 0.862, 0.755, and 0.925 respectively, with attacker accuracy at or below the label-only floor on CelebA (0.548 vs. floor 0.580) and within 2 percentage points (pp) of chance elsewhere, improving over an information-bottleneck adversarial baseline (IBAL) at a matched 52-epoch budget by +3.6, +2.5, and +1.3 pp (Welch's t-test, p=0.019 on CelebA).

Mapping textual specifications into formal representations is essential for ensuring the correctness of protocol designs and implementations. LLM-generated mappings, used for networking security or testing, are assumed to capture a perfect understanding of the specification, which may not hold in practice. The goal of this paper is to assess the extent to which LLMs can interpret the specification correctly. We examine the degree to which an LLM's implicit representation of a finite-state transition system-defined via natural language descriptions-aligns with a manually generated ground-truth model. We designed 4 tasks and 1482 task queries for 16 protocols. We evaluated different judge biases, observed the inherent difficulty gaps between tasks, looked into the effect of 4 context types, and the influence of protocol characteristics. Our work contributes to a step toward verifying whether LLMs can really be trusted in FSM (Finite State Machine) reasoning of protocol specifications.

Linear Optical Quantum Computing (LOQC) is emerging as a promising technology in large-scale quantum computing. Corresponding devices, including linear optical circuits, are rapidly growing in size and fabrication precision. However, even with improved quality, the circuits still exhibit imperfections leading to photon loss or phase noise that vary across the circuit. Already small deficiencies can have a vast impact on both accuracy and efficiency. For computations that do not require the entire linear optical circuit, so-called subcircuits, one can choose where to perform the computation, opening optimization opportunities that are addressed by subcircuit compilation. Current compilers spread the subcircuit's computation across the entire circuit, connecting the same input and output ports, thereby failing to fully exploit the optimization opportunities. However, given the subcircuit's reduced size compared to the entire circuit, one can keep its implementation localized and intentionally route photons to connect good input and output ports. This, however, is a highly non-trivial task, as numerous options must be weighed against one another, and the quality of the solution is influenced by a wide range of hardware parameters. In particular, the reflectivities of imperfect beam splitters can lead to photon losses during routing, which must be traded off against losses at the input and output ports due to imperfect transmission rates. Since current compilers do not account for these effects, we propose the first efficient subcircuit compilation method that optimizes for the coincidence rate, the probability of successfully measuring all desired photons, by handling hardware imperfections. Our evaluations demonstrate significant improvement over existing methods in robustness against phase noise and significantly increased coincidence rates across all benchmarks of up to plus 33.2%.