arXiv subjects

Search papers

Search indexed arXiv papers on artificial intelligence, large language models, computer vision and robotics. Read source abstracts and follow links to arXiv.

At least 1,063 records · Page 59Linked to original sources

arXiv2026

Identifier-Free Code Embedding Models for Scalable Search

Function association is a useful process for binary reverse engineers. Search tools exist to perform association at scale, but they do not utilize the full range of capabilities that AI-enabled search provides. Prior work has explored the development of embedding models for association between certain reverse engineering code representations, but that work does not cover bidirectional association between source code and decompiled, stripped code with standard preprocessing requirements. To bridge this gap, we formalize this function association problem and evaluate the extent to which embedding models can bidirectionally associate between these two representations. To improve model performance at this task, we fine-tune a Qwen3-Embedding model with contrastive learning. We find that our new model outperforms other models on all function association baselines by a substantial margin and generalizes to a constant-algorithm association task it is not explicitly trained on.

cs.CR↗

arXiv2026

MOSAIC: Module Discovery via Sparse Additive Identifiable Causal Learning for Scientific Time Series

Causal representation learning (CRL) seeks to recover latent variables with identifiability guarantees, typically up to permutation and component-wise reparameterization under appropriate assumptions. However, identifiability does not imply interpretability: latent semantics are typically assigned post hoc by alignment with known ground-truth factors. This limitation is particularly acute in scientific time series, where underlying mechanisms are unknown and discovering interpretable structure is a primary goal. In contrast, scientific observations (such as residue-pair distances, climate indices, or process sensors) are inherently semantic, as they correspond to named physical quantities. This raises a key question: can the interpretability of observations be transferred to the identifiable latent space? We propose MOSAIC (Module discovery via Sparse Additive Identifiable Causal learning), a sparse temporal VAE that integrates temporal CRL identifiability with support recovery over observed variables. MOSAIC identifies latent variables via regime-conditioned temporal variation, and recovers for each latent a sparse set of associated observations through an additive decoder, yielding module-level interpretability. We show that ANOVA main-effect supports are identifiable under general smooth mixing functions, with finite-sample recovery guarantees for a group-lasso estimator and empirical evaluation of MOSAIC's entropy regularizer. Empirically, MOSAIC recovers domain-consistent variable groups across RNA molecular dynamics, solar wind, ENSO climate, the Tennessee Eastman process, and a synthetic tokamak benchmark, enabling interpretable discovery of latent mechanisms in scientific time series.

cs.LG↗

arXiv2026

Towards Emotion Consistency Analysis of Large Language Models in Emotional Conversational Contexts

In this work, we conduct an analysis to examine the consistency of Large Language Models (LLMs) with respect to their own generated responses in an emotionally-driven conversational context. Specifically, the text generated by LLM is framed as a query to the same model, and its responses are subsequently assessed. This is performed with three queries across two dimensions of extreme and moderate emotions. The three queries are, in particular, false claim queries that contain inherently wrong assumptions (false presuppositions) in increasing order of intensity. Two commercial models, Claude-3.5-haiku, GPT4o-mini, and a medium-sized model, Mistral-7B, are considered in the study. Our findings indicate that LLMs exhibit below-average performance and remain vulnerable to false beliefs embedded within queries. This susceptibility is especially pronounced for moderate emotional content. Furthermore, an extended attention-score-based analysis highlights a shift in models' priority from evaluative to generative. The results raise important considerations for LLMs' deployment in high-stakes, emotionally sensitive contexts.

cs.CL↗

arXiv2026

Self-Programmed Execution for Language-Model Agents

At the heart of a language model agent is an orchestrator program which is responsible for the state transition between consecutive turns. In existing systems, this program is implemented in the harness. This paper introduces self-programmed execution (SPE), in which the model completion is itself the orchestrator program, and the harness evaluates this program but does not impose any fixed orchestration policy. Formally, an SPE state of an agentic machine is one from which a model completion can load any successor state up to an embedding, and an instance of an SPE state is one which executes the model completion as a program. In practice, the program requires a language with unusual features, as its source code doubles as the content of the model's context window. I therefore introduce Spell, a Lisp dialect in which programs can edit and re-evaluate themselves, and effectful expressions like model invocations are structured such that re-evaluating an edited program does not replay its side effects. Experiments with existing models, not trained for SPE or Spell, show that frontier models can learn Spell in-context and accomplish challenging agentic tasks. These results demonstrate how a LM can act as an agent without any fixed orchestration policy, and they raise the question of what self-orchestration strategies might be learned by a model trained for self-programmed execution. Code is available at https://github.com/lukejoconnor/spell.

cs.AI↗

arXiv2026

Clock-transition-enabled long electron spin coherence of telecom emitters in CeO$_2$

Rare-earth-ion-doped crystals has emerged as a promising platform for quantum technologies, owing to their narrow telecom-band optical emission, long spin memory, and compatibility with silicon integrated photonic architectures. However, the realization of scalable quantum devices requires host materials with intrinsically dilute spin environments to suppress decoherence. In this context, erbium (Er$^{3+}$) doped cerium oxide (CeO$_2$) is an attractive candidate due to the ultra-low concentration of nuclear spins in the host matrix and its compatibility with silicon-based technologies. Here we perform a comprehensive investigation of the coherence properties of Er$^{3+}$ electron spin qubit in CeO$_2$ via semiclassical as well as detailed cluster correlation expansion simulations. By systematically exploring magnetic field strength, pulse sequences, erbium concentration, and spin temperature, we identify regimes where decoherence from the spin bath is strongly suppressed. Our investigations illustrate that at dilute doping concentration (of the order of 10 ppb) and sub-Kelvin temperatures, operation near clock transitions enables Hahn-echo coherence times to approach the second timescale even at natural isotopic abundance. Importantly, from a practical standpoint, coherence times on the order of $\sim 10$ ms are expected even at liquid helium temperature (about 2 K) for similar concentrations. Moreover, we demonstrate that an additional enhancement can be obtained with conventional multi-$π$-pulse dynamical decoupling protocols. Thus, our findings establish Er$^{3+}$ doped CeO$_2$ as a front-runner for realizing spin qubits, quantum memories, and integrated quantum networks.

quant-ph↗

arXiv2026

PathPainter: Transferring the Generalization Ability of Image Generation Models to Embodied Navigation

Bird's-eye-view (BEV) images have been demonstrated to provide valuable prior information for navigation. Given the global information provided by such views, two key challenges remain: how to fully exploit this information and how to reliably use it during execution. In this paper, we propose a navigation system that uses BEV images as global priors and is designed for ground and near-ground robotic platforms. The system employs an image generation model to interpret human intent from natural language, identify the target destination, and generate traversability masks. During execution, we introduce cross-view localization to align the robot's odometry with the BEV map and mitigate long-term drift in conventional odometry. We conduct extensive benchmark experiments to evaluate the proposed method and further validate its effectiveness on both UAV and quadruped robot platforms. Using only a conventional local motion planner, the quadruped robot successfully completes a 285-meter outdoor long-range navigation task. This work demonstrates how the world-understanding capabilities of foundation models can be transferred to embodied navigation, enabling robots to benefit from the strong generalization ability of existing image generation models.

cs.RO↗

arXiv2026

SimCT: Recovering Lost Supervision for Cross-Tokenizer On-Policy Distillation

On-policy distillation (OPD) is a standard tool for transferring teacher behavior to a smaller student, but it implicitly assumes that teacher and student predictions are comparable token by token, an assumption that fails whenever the two models tokenize the same text differently. Under heterogeneous tokenizers, exact shared-token matching silently discards a large fraction of the teacher signal at precisely the positions where vocabularies disagree. We propose \textbf{\underline{Sim}ple \underline{C}ross-\underline{T}okenizer OPD (SimCT)}, which restores this signal by enlarging the supervision space: alongside shared tokens, SimCT compares teacher and student over short multi-token continuations that both tokenizers can realize, leaving the OPD loss form itself unchanged. We show that these units are the finest jointly tokenizable supervision interface, and that coarser alternatives remove teacher-student distinctions that are useful for on-policy learning. Across three heterogeneous teacher-student pairs on mathematical reasoning and code-generation benchmarks, SimCT shows consistent gains over shared-vocabulary OPD and representative cross-tokenizer baselines, with ablations confirming that the improvements come from recovering supervision discarded by exact shared-token matching. Code is available at \href{https://github.com/sunjie279/SimCT-}{https://github.com/sunjie279/SimCT-}.

cs.CL↗

arXiv2026

Adjoint test modules along Cohen--Macaulay morphisms

We provide a transformation rule for adjoint test modules along Cohen--Macaulay maps between Cohen--Macaulay varieties that have $F$-rational geometric fibers. This is, in part, an effective version of Enescu's theorem on the ascent of $F$-rationality under local maps with $F$-rational geometric fibers.

math.AG↗

arXiv2026

HIMMEL: Hierarchical Interleaved Multi-stream Motion Encoding for Long Video Understanding

Long video understanding with multimodal language models suffers from three compounding bottlenecks: heavy decode cost to obtain dense RGB frames, token growth with frame count, and weak motion perception under sparse keyframe sampling. Existing remedies either prune visual tokens after decoding, which leaves the expensive RGB pipeline untouched, or they discretise codec motion vectors with a generative tokeniser, which couples motion modelling to a heavy pre-training stage. In this paper we present HIMMEL, a hierarchical video-language framework that allocates semantic and motion capacity along separate paths. A small set of sparse semantic anchor frames is routed to the expensive host ViT to ground object identity and scene layout, while the far denser inter-frame intervals are encoded by a lightweight compressed-domain tri-stream adapter that distils motion evidence from motion-vector maps, residual maps, and an I-frame context branch into aligned motion tokens. These tokens are injected into the LLM via a differentiable placeholder mechanism, after a dedicated Stage 1 contrastive alignment that places the motion representation in a geometry compatible with the frozen visual backbone. We further show that an InfoNCE alignment objective beats MSE regression by +1.5 pp. On Video-MME, HIMMEL surpasses the dense 32-frame baseline by +2.3 pp (61.2 to 63.5%) while using 3.6 times fewer context tokens and running end-to-end in 1.03 s per question instead of 2.75 s. Switching to a stronger Qwen3-VL-8B host pushes HIMMEL to 64.9%, narrowing the gap to much larger proprietary systems while keeping inference on a single NVIDIA H20 GPU. Extensive ablations across stream composition, motion-encoder family, fusion mode, alignment objective, anchor count, LoRA rank, and video duration confirm that the full tri-stream contributes to the observed gains, and that the benefit grows with video length.

cs.CV↗

arXiv2026

A finite-shot mean-squared-error boundary for Richardson zero-noise extrapolation

Zero-noise extrapolation (ZNE) reduces the bias of noisy quantum-circuit measurements, but Richardson coefficients and shot splitting increase their statistical error. This paper determines the noise strength at which Richardson ZNE starts to lower the mean-squared error relative to the unmitigated estimator at a fixed shot budget $B$. The boundary is set by the balance between the leading squared-bias improvement and the leading excess-variance penalty: depending on the variance exponent of the measurement, it shrinks with $B$, becomes a budget threshold, or disappears. For fixed Richardson extrapolation with linear leading bias, the Richardson order and the shot allocation change only constants; the boundary exponent is set by the measured observable and protocol. Deterministic stabilizer measurements and nonzero-variance Hamiltonian or variational-energy measurements have boundaries that shrink as $B^{-1}$ and $B^{-1/2}$, and for finite-outcome shot measurements these two classes are the only ones. Qiskit Aer simulations of Greenberger-Horne-Zeilinger (GHZ) and quantum approximate optimization algorithm (QAOA) circuits reproduce this separation, and on IBM Quantum hardware ZNE lowers the mean-squared error at equal budget in 75 of 101 instances.

quant-ph↗

arXiv2026

Inner Horizon Saddles and a Spectral KSW Criterion

The Bekenstein-Hawking entropy formula $ρ= e^{A/4G}$ receives significant corrections for charged black holes near extremality. Using standard results in JT gravity, the correction term can semiclassically be expressed as minus the exponential of the inner horizon area, $e^{A_{\text{inner}}/4G}$, and the cancellation between these two exponentials enforces a vanishing density of states towards extremality, when the two horizons collide. Building on arXiv:2402.10162, we argue that the correction term corresponds to a complex saddle geometry of the bulk gravitational path integral. The proposed geometry has a negative boundary length and caps off at the inner horizon; we refer to it as the inner horizon saddle. We discuss how the saddle, and its accompanying minus sign, contribute to the density of states through a Picard-Lefschetz analysis of the inverse Laplace contour, together with a stability analysis of the saddle. We also address the inner horizon saddle's violation of the Kontsevich-Segal-Witten (KSW) allowability criterion for the inclusion of complex metrics. Despite this violation, which is believed to cause unphysical divergences in path integral computations, one can describe one-loop effects on the inner horizon saddle by carefully treating wrong-sign modes. Motivated by this observation, we propose a weaker version of the KSW criterion, which we call the spectral KSW criterion. Its purpose is to characterize when one-loop corrections around complex gravitational saddles are well defined.

hep-th↗

arXiv2026

TabFORGE: A Tabular Foundation Model for Generative Modelling

Generative modelling is a demanding test of foundation models, because it requires robust, holistic representation learning for a given data modality, rather than optimisation for a supervised prediction target alone. While recent work on tabular foundation models has achieved remarkable progress in predictive modelling, generative tabular foundation models remain underexplored. Existing tabular foundation generators, in particular, have not yet consistently matched strong dataset-specific generators in synthetic data quality. A key reason is their insufficient alignment with the inter-feature dependency structure of tabular data, which can reflect underlying causal relationships. In this paper, we address this gap by introducing a novel tabular foundation model, $\textbf{TabFORGE}$, built on pretrained $\textbf{Tab}$ular $\textbf{FO}$undational $\textbf{R}$epresentations for $\textbf{GE}$neration. TabFORGE is designed to utilise transferable representations induced by a pretrained structure-aware feature encoder, which captures information useful for characterising inter-feature dependencies. It further decouples latent modelling from decoding through a two-stage design: we first pretrain a score-based diffusion transformer, and then pretrain a denoising-aligned decoder using the denoised latent embeddings. This design elegantly mitigates the distribution shifts in latent embeddings that typically arise between training and inference. We evaluate TabFORGE comprehensively against 22 benchmark methods on 45 real-world datasets. Our results show that TabFORGE effectively learns and leverages generalisable tabular representations, enabling efficient generation of high-quality synthetic tabular data, particularly with strong structural fidelity. Code is available at https://github.com/SilenceX12138/TabFORGE.

cs.LG↗

arXiv2026

Recovery Models for Linear Batch Codes

Various types of recovery algorithms for batch codes have been considered previously, including asynchronous recovery and recovery algorithms arising from Almost Affinely Disjoint (AAD) families. We initiate a systematic investigation of linear batch codes equipped with recovery algorithms. We introduce online and strongly online batch codes, investigate their relations with asynchronous batch codes, and introduce (m,L)-strongly online batch codes, which provide particularly simple recovery algorithms. We then study simple batch codes associated with arbitrary bipartite graphs, obtaining graph-theoretic conditions for such codes and generalizing several known results.

cs.IT↗

arXiv2026

Locking Pretrained Weights via Deep Low-Rank Residual Distillation

The quality of open-weight language models has dramatically improved in recent years. Sharing weights greatly facilitates model adoption by enabling their use across diverse hardware and software platforms. They also allow for more open research and testing, to the extent that users can use them as checkpoints, fine-tune them according to their needs, and potentially redistribute them. In some cases, however, concerns on modifying these weights towards unauthorized uses may outweigh the pros of giving users such freedom. Defending against such adaptation is non-trivial: since an adaptive attacker can observe all weights and architectures by definition, they can reverse simple structural defenses, and use optimization to defeat the simplest locking mechanisms. In this work, we exploit the inference-training asymmetry of automatic differentiation as a novel defense axis. We propose DLR-Lock, a method where the purveyor of the model purposely replaces each pretrained MLP in their model with a deep low-rank residual network (DLR-Net) of comparable parameter count, forcing activation memory that grows linearly with depth during backpropagation. DLR-Nets are efficiently trained via module-wise regression. We show that, beyond this memory overhead, DLR-Lock results in architectural mismatches that complicate the optimization landscape of standard fine-tuning, and a backward pass that incurs disproportionately more overhead than the forward pass. Our defense succeeds in withstanding adaptive attackers with full knowledge of the defense strategy while preserving the original model's capabilities. Experiments on LLMs validate these claims.

cs.LG↗

arXiv2026

Conditioning Strengthens Classical Entropy and Matrix Inequalities: A Submodular Approach

Dembo, Cover, and Thomas (1991) developed an elegant information-theoretic framework for proving determinantal inequalities for positive definite matrices, which relies on the structural inequalities of differential entropy. Submodular functions, which subsume entropy, inherently satisfy these structural inequalities because they obey generalized forms of the fundamental properties of entropy -- a chain rule and the property that conditioning reduces the function's value (under an appropriate definition of conditioning). Applying subadditivity, Han's inequality (1978), and partition subadditivity (i.e., subadditivity over a partition) yields Hadamard's, Szász's, and Fischer's inequalities, respectively. Furthermore, this framework recovers Ky Fan's inequality (1955), a strengthening of Hadamard's inequality. We highlight a previously unnoticed information-theoretic insight behind this improvement: conditioning yields a tighter upper bound on the joint entropy than the unconditional one. In this paper, we establish conditional strengthenings of Han's inequality and partition subadditivity in the general setting of submodular functions. We derive equality conditions for these strengthened bounds and characterize when they strictly improve their unconditional counterparts. We specialize these results to differential entropy and apply them to establish strengthened versions of Szász's and Fischer's inequalities. The strengthening of Szász's inequality recovers Ky Fan's inequality as a special case, and is strictly stronger than the classical Szász's inequality for any non-diagonal positive definite matrix. We also derive an inequality concerning eigenvalues, which generalizes and strictly strengthens a corresponding eigenvalue inequality of Ky Fan. We provide numerical examples to explicitly illustrate the tightness of our proposed matrix determinantal bounds.

cs.IT↗

arXiv2026

On set-like sunflower-free families of subspaces over finite fields

The Erdős--Rado sunflower problem admits two natural analogues in finite vector spaces, corresponding to two different ways of generalising the set-theoretic notion of a sunflower. The first, used by Ihringer and Kupavskii [FFA 110 (2026) 102746], requires the petals to be in general position over the kernel; the second, used in the subspace codes literature (cf.\ Etzion--Raviv [DAM 186 (2015) 87-97], Blokhuis--De Boeck--D'haeseleer [DCC 90 (2022) 2101-2111]), requires only that the kernel equals the pairwise intersection of distinct petals. We refer to the second version as a \emph{set-like sunflower}, following Blokhuis, De Boeck and D'haeseleer. In this note, we focus on the set-like setting. {We first observe by illustrations that general-position sunflower-freeness does not imply set-like sunflower-freeness.} We then present a construction of set-like $s$-sunflower-free families of $k$-spaces \revised{in $\mathbb{F}_q^n$ for $\lceil(n+1)/2\rceil \leq k \leq n-1$, of size $q^{n-k}$ (and $q^{n-k}+1$ for even $q$),} based on a manipulated version of the lifting construction; to our knowledge, this is the first systematic construction tailored to this setting. Finally, we compare the size of the construction with the exact optimum for small parameters, computed by exhaustive search.

math.CO↗

arXiv2026

A proximal gradient algorithm for composite log-concave sampling

We propose an algorithm to sample from composite log-concave distributions over $\mathbb{R}^d$, i.e., densities of the form $π\propto e^{-f-g}$, assuming access to gradient evaluations of $α_f$-convex $f$ and a restricted Gaussian oracle (RGO) for $α_g$-convex $g$, where $α_g \ge 0$. The latter requirement means that we can easily sample from the density $\text{RGO}_{g,h,y}(x) \propto \exp(-g(x) -\frac{1}{2h}||y-x||^2)$, which is the sampling analogue of the proximal operator for $g$. If $f + g$ is $α$-convex and $f$ is $β$-smooth, our sampler achieves $\varepsilon$ error in total variation distance using $\widetilde{\mathcal O}(κ\sqrt d \log^4(1/\varepsilon))$ oracle calls where $κ:= 1 \vee β/α$, with $0\le α_g \le β$ and $α:= α_f + α_g > 0$, which matches prior state-of-the-art results for the case $g=0$. We further extend our results to cases where (1) $π$ is non-log-concave but satisfies a Poincaré or log-Sobolev inequality, and (2) $f$ is non-smooth but Lipschitz.

math.ST↗

arXiv2026

Collision Dynamics of False-Vacuum Oscillons

We study the collision dynamics of localized oscillons in two classes of $(1+1)$-dimensional scalar field theories with metastable false vacua, a normal class with a positive quartic self-interaction term and an inverted class with a negative quartic term. We construct small-amplitude oscillon solutions around the false vacuum using the Fodor {\emph{et al.}} expansion, and show that the force between oscillons decays exponentially at large separation, with a strength modulated by their relative phase. Numerical simulations of two-oscillon collisions exhibit reflection, crossing, and formation of excited oscillons. Resonance windows occur, similar to those found in kink-antikink collisions. In the normal theory, if the oscillons have sufficient energy, the field can pass over a sphaleron barrier and evolve into a kink-antikink pair, initiating a phase transition to the true vacuum. We also simulate the collision of oscillons evolved from a slightly perturbed sphaleron.

hep-th↗