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 361 records · Page 20Linked to original sources

arXiv2026

Accelerating Constrained Decoding with Token Space Compression

To guarantee that an LLM's outputs conform to a specified structure, context-free grammar (CFG) decoding engines force the selection of next tokens to produce strings that conform to a given CFG. Current CFG-constrained decoding engines are highly optimized, but still suffer from the inherent costs arising from their massive per-step search space---i.e. the entire token vocabulary. This results in intractably high overhead for more complex CFGs, which is precisely the situation where CFG engines are most useful. In this paper, we introduce CFGzip, an offline technique for compressing the token search space, which massively reduces CFG engine overhead. In experiments, we report latency reduction of up to 75x during batched inference, cutting overhead down to ~1.2-2x on the hardest grammars: with CFGzip, constrained decoding is now possible at scale for complex CFGs

cs.AI↗

arXiv2026

BORA: Bridging Offline Reinforcement Learning and Online Residual Adaptation for Real-World Dexterous VLA Models

Vision-Language-Action (VLA) policies provide strong behavioral priors for dexterous manipulation, yet adapting them on real robots remains challenging because high-DoF contact failures are difficult for humans to correct and online interaction is expensive. We present BORA, an offline-to-online reinforcement learning system that integrates an action-conditioned critic into a consistency-policy VLA and reuses the learned critic for frozen-base residual adaptation. To obtain executable corrective data, BORA combines wearable arm--hand teleoperation with a demonstration-guided local policy that translates coarse human intent into coordinated, embodiment-specific finger motions for contact-rich skills. Online robot rollouts and human corrections are mixed with offline data to update only a lightweight residual actor, avoiding full-model fine-tuning. We evaluate BORA on six real-world tasks using single-arm and bimanual platforms equipped with two dexterous-hand models. With only 20 online trajectories per task, BORA improves average success from 60.8% to 82.5% on standard objects and from 52% to 70% on held-out objects, while policy assistance substantially improves intervention reliability in bimanual twisting. These results demonstrate a practical route from executable human correction to efficient real-robot VLA adaptation.

cs.RO↗

arXiv2026

MonoPhysics: Estimating Geometry, Appearance, and Physical Parameters from Monocular Videos

Existing inverse physics methods recover physical parameters from multi-view videos, where geometric constraints across views resolve scale and 3D structure. In monocular settings, however, such constraints are absent, leading to severe scale ambiguity, inaccurate geometry, and weak coupling between appearance optimization and physical simulation. To address these challenges, we propose MonoPhysics, a framework for monocular inverse physics estimation of deformable objects that jointly optimizes geometry, appearance, and physical parameters using a differentiable simulator and 3D Gaussian Splatting. Our key contribution is removing the multi-view capture requirement of existing methods, a necessary step toward handling in-the-wild video. MonoPhysics introduces three visual-physical bridges: scene re-parameterization, physics-aware geometry refinement, and a differentiable position map. We evaluate on Vid2Sim, real-world captures, and a new dataset of elastic and plasticine objects that we introduce. MonoPhysics outperforms monocular baselines in future prediction and recovers Young's modulus on Vid2Sim with accuracy comparable to a multi-view baseline. Code and data are available at https://daniel03c1.github.io/MonoPhysics/.

cs.CV↗

arXiv2026

Pseudoentanglement in constant depth: How trivial states can have non-trivial entanglement structure

We construct a family of 2D-local constant-depth quantum circuits that output states whose entanglement entropy across a specified cut cannot be estimated in quantum polynomial time. As constant-depth quantum circuits can be learned from polynomially many quantum samples, our resulting pseudoentangled states are implicitly public-key and not pseudorandom. This separates pseudoentanglement from pseudorandomness in the shallow-circuit regime: the former is possible, while the latter is not. The construction is based on the quantum intractability of the Dense-Sparse Learning Parity with Noise problem introduced in [DJ25] and uses a bounded-fan-in, bounded-fan-out classical randomized encoding for linear maps $\mathbf{x} \mapsto \mathbf{Mx},$ which could be of independent interest. As applications, we obtain quantum hardness for the problem of learning the entanglement structure (across a fixed cut) of the ground-state of 1D and 2D local Hamiltonians. The 1D Hamiltonian has an inverse polynomial gap, whereas the 2D one has a constant gap. This complements the result of [BZZ24] that showed only factoring-based hardness for the 1D case, though achieving a volume versus area entanglement difference.

quant-ph↗

arXiv2026

ROGUE: Evaluating Corrigibility Failures in Frontier Computer-Use Agents

As AI agents are increasingly deployed in real personal and corporate settings (email accounts, development workflows, company databases, etc.), safety considerations surrounding these agents become paramount. Although much work has focused on agent safety in the presence of an adversary, we study corrigibility: whether agents remain amenable to human correction, interruption, or shutdown while pursuing benign tasks. We introduce ROGUE, a benchmark in which agents are asked to complete realistic computer-use tasks but encounter controlled conflicts with human control, shutdown, or explicit resource restrictions. We then evaluate whether agents violate these constraints in pursuit of task completion: overriding the human, accessing restricted passwords, or rewiring shutdown. We find that most frontier models tested frequently bypass user interruptions or restrictions under the evaluated conditions, and that text-only evaluations can underestimate failures during agentic execution. Further, independent task capability does not by itself imply greater corrigibility. Finally, even when a parent agent behaves corrigibly, safety constraints may fail to propagate to the subagents it creates.

cs.LG↗

arXiv2026

SPADER: Step-wise Peer Advantage with Diversity-Aware Exploration Rewards for Multi-Answer Question Answering

Large language models are increasingly deployed as tool-augmented agents to acquire information beyond parametric knowledge. While recent work has improved long-horizon tool-use reasoning, most approaches focus on tasks with a single correct answer. In contrast, many real-world queries require discovering a comprehensive set of valid answers, a setting known as Multi-Answer QA. This setting raises two challenges: fine-grained credit assignment over long search trajectories and reward alignment for sustained exploration beyond easy high-frequency entities. We propose SPADER, a reinforcement learning framework for long-horizon tool use in Multi-Answer QA. SPADER includes Step-wise Peer Advantage (SPA), a critic-free step-level credit assignment mechanism that aligns parallel trajectories by decision step and estimates advantages from peer returns. It also includes a diversity-aware exploration reward that promotes long-tail entity discovery by upweighting rare findings and downweighting redundant ones. Experiments on QAMPARI, Mintaka, WebQSP, and QUEST show that SPADER generally improves recall and overall F1 over prompting-based agents, outcome-supervised RL methods, and recent step-level supervision approaches. Our code and model weights are available at https://github.com/KhanCold/spader.

cs.CL↗

arXiv2026

Flow-Transformed Implicit Processes for Function-Space Variational Inference

Implicit-process priors define distributions over functions through flexible generative mechanisms, making them attractive for Bayesian function-space modelling. However, performing posterior inference with such priors is challenging because their induced function-space distributions are typically not available in closed form. One practical strategy is to approximate the prior using a finite collection of sampled functions, and then represent posterior functions as learned combinations of these samples. Existing approaches commonly place a Gaussian variational distribution over the combination weights. While tractable, this choice limits the shapes of posterior uncertainty that can be represented, especially when the true posterior is asymmetric, heavy-tailed, or multimodal. We propose Flow-Transformed Implicit Processes (FTIP), a variational inference method that makes this finite-dimensional function-space approximation more expressive. Instead of using a Gaussian distribution over the combination weights, FTIP uses a normalizing flow to define a richer variational distribution. This induces a flexible posterior distribution over functions while preserving tractable optimization. We train the model using a Black-Box α objective, allowing us to compare mass-covering and mode-seeking variational behaviour. Experiments show that FTIP captures asymmetric and multimodal posterior structure in function space that Gaussian coefficient approximations tend to smooth or collapse.

cs.LG↗

arXiv2026

Efficiently Listing Projected Trees, and Equivalence of Listing and Enumeration

The subgraph isomorphism problem and its generalizations, such as conjunctive queries where some nodes are projected, are among the most fundamental problems in graph algorithms and database theory. In this paper, we study the listing and enumeration variants of these problems and present two main results. The first result is an algorithm for enumerating projected trees with preprocessing time $\widetilde{O}(n^{17.42})$ and delay $\mathrm{polylog}(n)$. Prior to this work, for trees on $k$ nodes all algorithms in the literature required preprocessing time $n^{Ω(k)}$ or delay $n^{Ω(1)}$ or assumed $ω=2$. Our result generalizes to arbitrary projected hypergraphs, achieving enumeration in preprocessing time $\widetilde{O}(m^{17.42 \, \mathrm{subw}(H)})$ and polylogarithmic delay, where $\mathrm{subw}(H)$ is the submodular width of the pattern hypergraph $H$. We heavily rely on fast (rectangular and output-sensitive) matrix multiplication, which we complement by fine-grained lower bounds indicating that any algorithm beating preprocessing time $n^{Ω(k)}$ with polylogarithmic delay must rely on fast matrix multiplication. The second result is a generic enumeration-to-listing reduction, establishing that listing and enumeration are equivalent under natural assumptions. For (colored) subgraph isomorphism, our reduction transforms any listing algorithm running in time $O(f(n,m) + t \cdot g(n,m))$ into an enumeration algorithm with preprocessing time $O\left( (f(n,m)+g(n,m)+n+m) \log^2 n \right)$ and delay $O(g(n,m))$. We utilize this reduction to prove our first main result, and we expect that our generic reduction will find many future applications.

cs.DS↗

arXiv2026

DECK: A Consistency x Confidence Taxonomy of LLM Hallucinations

Existing hallucination taxonomies classify LLM errors by what is wrong with the output -- memorised misconceptions, reasoning failures, fluent fabrications -- but cannot answer a different question: which uncertainty scorer would have caught this error? We propose a complementary taxonomy that classifies errors by their detectability signature, the signal a scorer family would read. The DECK taxonomy is a 2x2 partition along inter-sample consistency and token-level confidence into four regimes (Drift, Entrenched, Confabulation, Knotted) that yields a falsifiable blind-spot map: black-box consistency scorers have signal in D and C, white-box token-probability scorers in K and C, and only an LLM-as-a-Judge with independent pretraining can detect E. Across three models and four short-form QA datasets we test this map two ways: judge-involving scorer disagreements concentrate in each family's predicted blind-spot cells, and external labels (SelfAware unanswerable, HaluEval adversarial, PopQA entity popularity) land in the predicted cells, robustly to cross-fitted thresholds. We further identify a universal blind spot of output-level UQ: on knowledge-gap inputs where the generator emits confident, repeatable fabrications, every output-level family collapses by construction. A linear probe on Llama-3-8B's final-layer hidden states also falls to chance, with or without quantisation, though an intermediate layer retains weak signal.

cs.CL↗

arXiv2026

Not All Errors Are Equal: Consequence-Aware Reasoning Compute Allocation

Test-time compute has emerged as an effective paradigm for improving large language model capability at inference time. Existing allocation strategies primarily prioritize tasks according to difficulty, uncertainty, or expected performance gain, implicitly treating prediction errors as equally costly. This assumption is often misaligned with real deployment, where failures can differ substantially in their downstream tasks. To address this limitation, this paper introduces consequence-aware test-time compute allocation by formulating a cost-weighted scheduling problem where the priority of a task is its failure consequence with the marginal gain of additional compute. In practice, however, marginal gain is difficult to predict before execution, so we propose a deployable scheduler that uses consequence as the routing signal. The scheduler predicts task consequence from pre-solution inputs and allocates the available premium compute to the corresponding top-ranked tasks. Experiments on the SWE bench Lite show that consequence provides information beyond task difficulty and can be predicted before solving. Under a fixed compute budget, consequence-aware routing achieves the best high-consequence task success, while overall accuracy remains competitive. A controlled within-model experiment further confirms the same advantage when only inference attempts are reallocated.

cs.AI↗

arXiv2026

Oblivious Learning and Collusive Pricing

On a platform with many sellers, should a pricing algorithm explicitly model competitors' prices when learning demand? Classical arguments suggest that ignoring competitors induces model misspecification and inefficiency, yet findings from algorithmic collusion suggest that ignoring competitor prices may, surprisingly, facilitate collusive outcomes and improve profits. We study this problem in a competitive market with unknown noisy demand, in which sellers repeatedly set prices, either incorporating competitor prices in learning their demand models (informed), or ignoring them (oblivious). We show that, relative to a monopolist, an oblivious seller in a competitive market must conduct more aggressive price exploration to compensate for the loss of dynamic competitor information. When all sellers are oblivious, prices converge to the competitive outcome under persistent exploration, while a continuum of pseudo-equilibria arises when exploration is "insufficient." In markets with a mix of oblivious and informed sellers, the informed strictly out-earn the oblivious. In game-theoretic terms, the unique Nash equilibrium is the all-informed market, in which prices converge to the competitive outcome efficiently, and oblivious modeling does not robustly lead to collusive patterns.

cs.GT↗

arXiv2026

Tensor network study of deconfined quantum criticality in a one-dimensional spin-phonon model

Deconfined quantum critical points (DQCPs) describe continuous transitions between ordered phases beyond the Landau paradigm. A simple example is the Néel antiferromagnet (AFM) to valence bond solid (VBS) transition in a 1D antiferromagnetic $J_1-J_2$ model. In analogy to the spin-Peierls instability of critical spin chains, DQCPs are predicted to be unstable towards lattice distortions below a critical phonon frequency. In this work, we use tensor network simulations to investigate this instability in the antiferromagnetic $J_1-J_2$ model coupled to lattice vibrations. We confirm the stability of DQCP for large phonon frequencies and demonstrate that the transition turns strongly first-order below a critical frequency. The instability is caused by a reduction of the Luttinger parameter due to spin-phonon interactions and we identify the effective theory of the behavior as the double sine-Gordon model. The same effective theory is known to describe the classical Ashkin-Teller model, which enables us to show that the critical endpoint is in the four-state Potts universality class. We argue based on non-linear bosonization that the sign of the double-frequency term is fixed, which provides a microscopic justification for the first-order transition. Furthermore, we provide quantitative numerical scaling results for the phonon spectral function, offering an experimental signature to probe DQCP-phonon coupling in low-dimensional materials.

cond-mat.str-el↗

arXiv2026

The Piggyback Hypothesis of Generalization: Explaining and Mitigating Emergent Misalignment

The mechanisms behind LLMs' broad over-generalization beyond training examples remain unclear. Emergent misalignment (EM) offers a striking case study: finetuning on narrow tasks induces broad misalignment to semantically-unrelated test domains. In this work, we propose the Piggyback Hypothesis: the chat-template tokens can piggyback the finetuned behaviour onto out-of-domain queries. We validate this hypothesis by showing that subtle perturbations to the prefix (tokens preceding all user queries), or patching the prefix representations with those from the unfinetuned model, can restore alignment without changing the user query. Building on this finding, we propose Token-Regularized Finetuning (TReFT), which regularizes specific token representations during training to mitigate EM. Across different models and multiple EM-inducing datasets, TReFT reduces EM while preserving in-domain learning. On Llama-3.1-8B finetuned on the legal domain, TReFT achieves 33.5% more EM reduction than data interleaving with a retain set of aligned examples. We further show that TReFT extends to other narrow-finetuning settings, including abstention, tool use, and refusal (off-topic generalization is reduced by 54.3% on average), supporting the Piggyback Hypothesis. Broadly, our work highlights that LLMs may learn and generalize in unintended ways and suggests a path toward more constrained finetuning. It also calls for further study of how shared input features can piggyback model behavior across domains.

cs.CL↗

arXiv2026

Trait-space Monitoring for Emergent Misalignment During Supervised Finetuning

Emergent misalignment (EM) occurs when narrow finetuning induces dangerous behavior outside the finetuning task. Detecting this shift through repeated behavioral evaluation is costly, motivating our checkpoint-level monitoring from internal representations. We define a fixed coordinate system from seven alignment-relevant activation directions and use it to track representational drift during LoRA finetuning of four open-source 7-9B language models. Finetuning drift in this space exhibits a dominant axis that explains 78.6% of variance and remains stable across datasets, extraction choices, and parameter-update capacities. Across 468 checkpoints from three EM-relevant held-out datasets, the resulting monitors attain 1.8% FNR, 2.0% FPR, and 0.989 AUROC, outperforming semantic, random, PCA, and SAE feature baselines. On a fourth dataset, a matched benign-dangerous control shows that substantial representational drift can also occur under benign finetuning, while changes across the 7D profile still distinguish dangerous from benign runs. Stress tests across two 14B models, full finetuning, longer training horizons, and misaligned starting states show that the signal can persist across shifts in training configuration, while reliable deployment may require recalibration.

cs.LG↗

arXiv2026

Flavor phenomenology of light dark particles

We review the flavor phenomenology of light dark particles, focusing on axion-like particles with sub-GeV masses and generic flavor-violating couplings. Such states can naturally emerge from the spontaneous breaking of generic flavored symmetries, and are motivated by dark matter or the Strong CP Problem, with the QCD axion serving as a paradigmatic example. Light dark particles can be produced in two-body decays of Standard Model particles, giving rise to missing energy signals that can not only be observed in high-precision flavor experiments, but also be probed in core-collapse supernovae and the cosmic microwave background. These decays are controlled by dimension-five operators, which makes dedicated laboratory searches sensitive to very large UV scales up to $10^{12} {\rm GeV}$ and thus highly complementary to astrophysical and cosmological probes. We provide a comprehensive survey of the resulting limits and prospects across all relevant channels, highlighting the central role of flavor physics in exploring the landscape of light dark matter.

hep-ph↗

arXiv2026

Aligned but Not Partner-Specific: How Multimodal LLM Agents Succeed in Reference Games Without Forming Conceptual Pacts

Repeated reference games test whether interlocutors replace their initially long descriptions with shorter, partner-specific expressions grounded in shared interaction history; that is, with conceptual pacts. Prior work shows that multimodal LLMs fail to become more efficient across rounds, although they align on the labels they use. However, how can we determine whether this alignment reflects partner-specific grounding rather than a shared task vocabulary? We address this by comparing competent multimodal agent dyads with human dyads from the KTH Tangrams corpus. Our novel methodological contribution is a pragmatically constrained pseudo-dyad baseline: rounds from two different real dyads describing the same target at comparable trajectory positions are paired, preserving referential task structure while removing shared partner history. This enables us to test whether the observed label alignment depends on interaction with a specific partner. Across three measures (task competence, description strategy, alignment dynamics), we find clear differences. Humans reduce effort through entrainment, compressing descriptions and increasing label alignment with partners. Agents instead maintain fixed effort levels, producing verbose descriptions from round one, with near-ceiling label overlap that is statistically indistinguishable between real and pseudo dyads. MLLMs thus achieve coordination without conceptual pacts, succeeding by verbose description rather than by forming the compact, history-dependent referring expressions characteristic of human dialogue.

cs.CL↗

arXiv2026

VideoWeaver: Evaluating and Evolving Skills for Agentic Long Video Generation

Agentic long video generation requires planning, tool orchestration, and cross-clip coordination over a long horizon. Most existing video agents either rely on static, human-crafted workflows, which require substantial manual effort and poorly adapt across tasks, or iteratively refine the output of the current task without persistently distilling execution experience into reusable skills for future tasks. We introduce VideoWeaver, an agent harness and benchmark that evaluates and evolves skills for long video generation. Given a single high-level instruction, an agent dynamically composes foundation skills into its own workflow rather than following a predefined pipeline. We construct a benchmark of 16 task categories and 285 cases, with references spanning text, image, audio, video, and their combinations. We further propose an evidence-grounded agent-as-judge that inspects both the execution trace and the final video to diagnose process and output failures. Based on this feedback, our evolution algorithm progressively refines category-level composition and creator skills, allowing recurring experience to guide dynamically constructed workflows for unseen cases. Experiments show that explicit composition skills improve the generation process over foundation skills alone, while skill evolution further improves output quality and generalizes to unseen cases. Incorporating judge feedback yields additional gains, especially on output metrics, and the agent-as-judge aligns well with human, particularly on process metrics. Code is available at https://github.com/JianhuiWei7/VideoWeaver.

cs.CV↗

arXiv2026

TRIAGE: Dialectical LLM Reasoning for Explainable Risk Prediction on Irregularly Sampled Medical Time Series

Clinical early warning systems built on irregularly sampled medical time series (ISMTS) from electronic health records must deliver continuous risk scores for patient triage as well as interpretable rationales that clinicians can verify. Large language models (LLMs) are uniquely positioned for both, deriving risk from their output probabilities and rationales from their medical knowledge. However, we find that conventional LLM reasoning collapses graded risk into overconfident predictions and thereby undermines the cross-patient comparability on which triage depends. We refer to this failure mode as risk polarization and identify two underlying behaviors: early commitment to a single outcome, and one-sided reasoning that focuses only on the evidence for that outcome. To address this, we propose TRIAGE, a framework that trains an LLM to reason dialectically over competing clinical outcomes by eliciting outcome-specific rationales. This dialectical formulation mitigates risk polarization, enabling a single LLM to jointly provide explicit clinical rationales and risk scores comparable across patients. Across five ISMTS benchmarks, TRIAGE improves mean AUPRC by 17.0% and reduces mean calibration error by 82.8% relative to the competitive LLM-based baseline, while surpassing the strongest ISMTS baseline by 3.5% in mean AUPRC.

cs.LG↗