Academic Intelligence · Curated Daily

探索全球前沿学术脉络

AcademicHub 汇聚顶级期刊与预印本平台的实时文献。定制您的专属科研雷达,利用大语言模型自动生成交叉领域文献分析简报。

01.
arXiv (quant-ph) 2026-06-16

Hyperinvariant Spin Network States – An AdS/CFT Model from First Principles

arXiv:2510.06602v2 Announce Type: replace Abstract: We study the existence and limitations of hyperinvariant tensor networks incorporating a local SU(2) symmetry. As discrete implementations of the anti de-Sitter/conformal field theory (AdS/CFT) correspondence, such networks have created bridges between the fields of quantum information theory and quantum gravity. Adding SU(2) symmetry to the tensor network allows a direct connection to spin network states, a basis of the kinematic Hilbert space of loop quantum gravity (LQG). We consider a particular situation where the states can be interpreted as kinematic quantum states for three-dimensional quantum gravity. We show that important aspects of the AdS/CFT correspondence are realized in certain quantum states of the gravitational field in LQG, thus justifying, from first principles, a class of models introduced by [F. Pastawski et al., JHEP 06, 149 (2015)]. We provide examples of hyperinvariant tensor networks, but also prove constraints on their existence in the form of no-go theorems that exclude absolutely maximally entangled states as well as general holographic codes from local SU(2)-invariance. We calculate surface areas as expectation values of the LQG area operator and discuss further possible constraints as a consequence of a decay of correlations on the boundary.

02.
arXiv (math.PR) 2026-06-16

A small noise approximation for Muller's Ratchet

arXiv:2606.15842v1 Announce Type: new Abstract: We consider an infinite system of SDEs with Fleming-Viot noise indexed by $k=0,1,2,\dots$, whose parameters $\alpha,\lambda$, and $\nu$ are the (deleterious) selection coefficient, the (uni-directional) mutation rate, and a quantity which determines the size of the system's fluctuations. The SDE's unique weak solution $X(t) = (X_k(t))_{k=0,1,2,...}$ models what is known in population genetics as Muller's ratchet. Here, $X_k(t)$ stands for the frequency of individuals carrying $k$ deleterious mutations. Since the mutation process is uni-directional, $t\mapsto \inf\{k: X_k(t)> 0\}$ is non-decreasing for almost every path of $X$, and we refer to an increase as a click of Muller's ratchet. A long standing question concerns the clicking rate of Muller's ratchet. Using Duhamel's principle for semigroups, we give a partial answer by approximating $E(\sum_{k=1}^\infty kX_k(t) )$ and $E\big(X_0(t)\big)$ up to $O(1/\nu^2)$ for fixed $\alpha$, $\lambda$ and $t>0$. Our results suggest that $\psi:=\nu \alpha e^{-\lambda/\alpha}$ is a crucial quantity also when the mutation/selection ratio $\theta = \lambda/\alpha$ is moderately large: for large $\nu \alpha$, clicking of the ratchet on the time scale $\frac 1\alpha \log \theta$ becomes rare as soon as $\psi$ becomes large.

03.
arXiv (CS.CL) 2026-06-17

Decoding Hidden Deception in Reasoning LLMs: Activation Explainers for Deception Auditing

As LLMs acquire stronger reasoning capabilities, deceptive behavior becomes an increasingly serious safety concern. Existing deception monitors either score visible transcripts or derive scalar probe scores from representation vectors, leaving little inspectable evidence about why a response is suspicious. We introduce STATEWITNESS, an activation explainer for deception auditing. A separate decoder reads a target model's hidden states, then answers natural-language queries or emits structured reports about them. We evaluate STATEWITNESS on two target reasoning LLMs across seven deception datasets. STATEWITNESS reaches 0.916 mean AUROC, a relative gain of 11.6% over the best black-box text monitor and 25.0% over the best activation-probe baseline under the same evaluation protocol. When combined with existing monitors, STATEWITNESS reduces missed deceptive examples in simple threshold ensembles. Beyond scalar detection, the decoder returns query-level answers, schema reports, and token- or sentence-level evidence traces for human inspection. We view this interface as a potential building block for broader interpretability and alignment tools.

04.
arXiv (CS.CL) 2026-06-16

CoRA: Confidence-Rationale Alignment for Reliable Chain-of-Thought Reasoning

Chain-of-thought (CoT) reasoning can improve LLM performance, but high answer confidence may be misleading when the accompanying CoT rationale is plausible yet incomplete or poorly supported. We study confidence–rationale alignment: whether a model's confidence in its committed answer is justified by its generated rationale. We introduce a GRPO-based reinforcement learning framework that jointly rewards answer correctness, committed-answer probability, and rubric-based rationale support, where the rubric assesses grounding, coherence, task match, and connection to the selected answer without revealing the gold answer to the judge. Across MedQA, MathQA, and OpenBookQA using three open-weight LLMs, our method reduces the confidence–rationale alignment error by up to 26.51% compared with untuned checkpoints, SFT, and correctness-only GRPO, while maintaining competitive accuracy and often improving calibration. These results show that reliable CoT reasoning requires not only confident answers, but rationales that substantively support them.

05.
arXiv (math.PR) 2026-06-15

Lehner's operator norm formulas, semidefinite programming, and spiked matrix models

arXiv:2606.14687v1 Announce Type: new Abstract: Lehner (1999) derived elegant formulas for the operator norm $\|\mathfrak{X}\|$ of operators of the form $\mathfrak{X} = \mathbf{A}_0 \otimes \mathfrak{1} + \sum_{i = 1}^n \mathbf{A}_i \otimes \mathfrak{m}_i$, also easily generalized to the spectral edge $\lambda_{\max}(\mathfrak{X})$, in terms of nonlinear optimization problems over positive definite matrices. Here the $\mathbf{A}_i$ are finite-dimensional Hermitian matrices, the $\mathfrak{m}_i$ are either free semicircular or free Rademacher families of operators, and $\mathfrak{1}$ is the identity operator. We first show that both of Lehner's nonlinear optimizations can be rewritten as linear semidefinite programs (SDPs), even in the Rademacher case where Lehner's optimization is not itself convex. We give the primal and dual forms of these SDPs, derive the complementary slackness relations and consequences thereof, and propose that the SDPs are more stable and accurate than the iterative numerical scheme proposed in Lehner's original work. We then apply the SDPs from the semicircular case to spiked matrix models, studied recently via Lehner's formula by Bandeira, Cipolloni, Schröder, and van Handel (2024). We give a new proof of the Baik–Ben Arous–Péché (BBP) transition they establish in models with isotropic (but possibly correlated) Gaussian noise by constructing feasible variables for the associated primal and dual SDPs. Combining our construction with a sensitivity interpretation of optimal dual variables, we study the fluctuations of leading eigenvectors of such models. We conjecture and give numerical evidence that these fluctuations are Gaussian but anisotropic and non-universal, and that their covariance may be computed in terms of the optimizer of the dual of Lehner's formula, which in turn is approximately the leading eigenmatrix of a completely positive operator associated to the covariance of the noise model.

06.
medRxiv (Medicine) 2026-06-22

Assessment of adaptive functioning in Angelman syndrome using the Vineland Adaptive Behavior Scales, Third Edition

Purpose: This study examined longitudinal trajectories of adaptive functioning in 331 individuals with Angelman syndrome (AS) using the Vineland Adaptive Behavior Scales, Third Edition (Vineland-3) and examined differences by molecular subtype. Methods: A total of 331 individuals (156 females, 47%) with genetically confirmed AS (ages 6 months to 52 years) were assessed between 2018 and 2025, including 207 with a deletion subtype, 63 with uniparental disomy or imprinting defect, and 61 with a UBE3A point mutation. Growth scale values were analyzed using linear mixed-effects models with log2-transformed age. Results: Individuals with deletion subtypes demonstrated significantly lower adaptive functioning across domains compared to those with non-deletion subtypes. Adaptive skills across all Vineland-3 subdomains increased nonlinearly with age, showing faster growth early in life that slowed over time, with largely parallel trajectories across subtypes. Conclusion: Individuals with AS demonstrate slow but steady growth in adaptive functioning that continues into adulthood, with progress varying by molecular subtype. These findings provide updated natural history benchmarks and demonstrate the utility of the Vineland-3 for clinical trials.

07.
arXiv (CS.LG) 2026-06-12

A Stabilized Path-Space Approach to Diffusion-Based Posterior Sampling

arXiv:2606.12710v1 Announce Type: new Abstract: Diffusion models provide expressive data-driven priors for Bayesian inverse problems, but many diffusion posterior samplers rely on heuristic guidance approximations that can fail for nonlinear operators and multimodal posteriors. In this work, we develop a stabilized path-space framework for diffusion-based posterior sampling. Starting from a base diffusion process whose terminal marginal represents the prior, we define a likelihood-weighted target measure on trajectories and cast posterior sampling as learning a controlled stochastic process whose path measure matches this target. This formulation connects diffusion posterior sampling to stochastic optimal control while preserving the Bayesian structure needed for uncertainty quantification. We introduce a time reparameterization that makes the path-space control problem well posed by removing the bias induced by the unknown initial value function, without auxiliary training. We then learn the control via a trust-region path-space optimization method with log-variance objectives. The path-space perspective also unifies our learned control approach with existing guidance-based samplers, quantifies the sampling error induced by approximate controls, and yields importance sampling corrections for asymptotically exact posterior expectations. We evaluate the proposed framework on a suite of benchmark inverse problems with analytically characterized or high-quality reference posteriors, enabling principled assessment of sampling accuracy and uncertainty quantification. These experiments provide insight into the behavior of diffusion-based posterior samplers and demonstrate improved accuracy and robustness over leading approaches.

08.
arXiv (CS.AI) 2026-06-15

The Silent Cost of Artificial Intelligence Assistance: A Theory of Autonomy Surrender, the Recovery Mechanism, and the Restoration of Human Agency

arXiv:2606.13962v1 Announce Type: cross Abstract: The integration of artificial intelligence into human decision-making environments has introduced a previously undertheorized cost: the gradual surrender of human autonomy in exchange for access to information and computational assistance. Building on the Human Identity and Autonomy Gap (HIAG) framework, this paper advances a theoretical model of autonomy surrender as a measurable, cumulative process driven by cognitive bandwidth depletion. The model proposes three interacting mechanisms: the silent cost of AI assistance, in which autonomy is transferred incrementally and without awareness; the surrender threshold, beyond which reclaiming autonomous function becomes cognitively and psychologically difficult; and the recovery mechanism, which establishes the design obligation and the ethical responsibility accompanying deliberate human re-assumption of control. The paper argues that human re-entry into the decision loop is not a passive option but an active cognitive event requiring intentional bandwidth restoration. The design of AI systems must incorporate structured re-entry pathways, here termed recovery mechanisms, that preserve human agency while appropriately distributing responsibility. The model further predicts a terminal state, here termed preference inversion, in which functional dependence on AI assistance is experienced not as a deficit but as a preference, transforming the restoration of autonomy from a design problem into a cultural and political one. Implications are drawn for AI system design, governance frameworks, and human factors research.

09.
arXiv (CS.LG) 2026-06-16

Physics-conforming Latent Twins

arXiv:2606.15053v1 Announce Type: new Abstract: Surrogate models are central to scientific machine learning, where they enable fast prediction, simulation, inference, and control for complex physical systems. For time-dependent problems, however, accurate interpolation of training trajectories is not sufficient: reliable surrogates should also respect the conservation laws, invariants, admissibility conditions, and dissipative structures that give those trajectories physical meaning. We introduce Physics-conforming Latent Twins, a framework for learning latent surrogate solution operators whose dynamics satisfy selected physical principles by design. The method builds on the Latent Twin formulation by jointly learning an encoder, a decoder, and a latent flow map between arbitrary time-indexed states, while constraining the latent dynamics to preserve or dissipate prescribed structural quantities. We develop a constraint-transfer viewpoint that connects physical structure in the original state space with compatible constraints in latent space, and prove structure-preservation bounds showing how latent enforcement improves control of physical defects after decoding. We also derive algebraic conditions for latent flow maps that preserve linear and quadratic invariants or enforce dissipative inequalities. Numerical experiments on representative ODE and PDE benchmarks demonstrate improved constraint satisfaction, structural fidelity, and qualitative long-time behavior while maintaining accurate surrogate prediction.

10.
medRxiv (Medicine) 2026-06-18

Urinary Creatine Riboside Complements PSA to Improve Disease Detection in the Diagnostic Gray Zone of Prostate Cancer

Circulating prostate-specific antigen (PSA) discriminates poorly in the diagnostic gray zone (3.0-9.99 ng/mL), where ~75% of biopsies yield no clinically significant prostate cancer (PCa). We evaluated whether urinary creatine riboside (CR), a tumor-derived metabolite excreted through the prostatic urethra, complements PSA for gray-zone detection and independently predicts prostate-cancer-specific mortality (PCSM). In the NCI-Maryland PCa Case-Control Study (951 cases, 962 controls; 47.6% African American men; median follow-up 11.5 years), urinary CR was quantified by UPLC-MS/MS. Within the PSA gray zone (n = 668), urinary CR was complementary to PSA, with markedly higher single-marker discrimination than PSA (AUC 0.93, 95% CI 0.88-0.98 vs 0.77, 0.66-0.89) and additive when combined ({Delta}AUC +0.17, p < 0.001; 91.4% sensitivity at 80% specificity). After adjustment for 11 clinical and sociodemographic covariates, urinary CR independently predicted PCSM complementary to PSA (Fine-Gray SHR 1.72, 1.35-2.19 for CR; 1.35, 1.08-1.68 for PSA; Harrell's C 0.85 for CR + PSA vs 0.77 for PSA alone), with strongest signal in African American men (SHR 2.43, 1.57-3.75 for CR). We conclude that urinary CR is a candidate non-invasive biomarker complementary to PSA - improving gray-zone triage and predicting PCSM; prospective validation in biopsy-referred cohorts is warranted.

11.
arXiv (CS.AI) 2026-06-18

SAE Interventions are Unreliable: Post-Intervention Recovery of Suppressed Behavior

arXiv:2606.18322v1 Announce Type: cross Abstract: Sparse Autoencoders (SAEs) decompose residual-stream activations into interpretable features. Recent latent-space defenses increasingly rely on these decompositions, assuming that identified "unsafe" SAE features serve as actionable handles for monitoring and intervention. In this paradigm, clamping a specific harmful feature is expected to reliably prevent model misbehavior. However, we show that this success may hide a recoverable failure mode: the clamp may block one visible route to a behavior without eliminating the behavior itself. We formulate this vulnerability as post-intervention recovery, a constrained residual-space optimization problem. Starting from the post-intervention residual state, we optimize residual perturbations to recover the pre-intervention behavior while preserving the post-intervention values of the targeted SAE features. Even under a strong threat model where the intervention remains active throughout optimization and generation, recovery remains possible. To rule out that recovery simply undoes the intervention, we use encoder-orthogonal updates for single-layer interventions and the corresponding feature-map Jacobian in the cross-layer setting. Across TPP, unlearning, IOI, and refusal steering experiments, this stress test reveals recoverable behavior despite successful feature-level intervention. Especially in the safety-critical refusal-steering setting, we achieve a 95.8% recovery rate on valid samples while keeping defended-feature relative drift to 0.131, substantially below suffix-based baselines. A recovery-path attribution analysis further localizes this recovery to the SAE reconstruction residual, the component left unexplained by the SAE. These results expose a gap between feature-level control and behavioral completeness: SAE features can support causal intervention, but controlling them does not guarantee control over the underlying behavior.

12.
arXiv (CS.CL) 2026-06-17

Implicit vs. Explicit Prompting Strategies for LVLMs in Referential Communication

Two recent studies (Jones et al. (2026); Zeng et al. (2026)) reach apparently contradictory conclusions about whether LVLMs can coordinate on efficient referring expressions. We control for task differences between the studies while directly comparing their prompting styles. We replicate the finding that models can coordinate efficient referring expressions when explicitly prompted to do so, suggesting that other task differences are not responsible for divergent results. However, we also find that the same models fail to infer the need for communicative efficiency from a more implicit prompt, highlighting critical differences between how humans and AI systems communicate.

13.
bioRxiv (Bioinfo) 2026-06-11

inquiSTR: a toolkit for accurate and efficient population-scale tandem repeat genotyping and analysis

Tandem repeats are highly mutable genomic elements linked to human traits and diseases. Profiling large catalogs of tandem repeats from population-scale long-read sequencing data requires accurate and efficient tools. We introduce inquiSTR, a command-line toolkit for fast genome-wide tandem repeat length genotyping. inquiSTR, with efficient parallel processing and low-memory streaming algorithms, genotypes a genome-wide repeat catalog of 1.78 million loci in less than two minutes. Benchmarking shows high accuracy and significantly faster performance compared to existing tools and truth sets. inquiSTR also provides methods for downstream analyses such as population structure inference, association testing, and outlier detection.

14.
medRxiv (Medicine) 2026-06-18

Cost analysis of overseas versus domestic vaccination of US-bound refugees

Context: To ensure healthy resettlement and protect US health security, the Vaccination Program for US-bound Refugees (VPR) offers some recommended vaccines to refugees overseas before resettlement to the United States. The selected vaccines and number of doses vary by country of departure. VPR was found to be cost-saving in 2018 but had since expanded to more sites. Objective: Assess VPR's current costs and impact on post-arrival domestic vaccination needs and costs. Setting and Participants: A model-based analysis of the Federal government costs for VPR and post-arrival (US) vaccination of resettled refugees separated across five regions: Africa, Asia, the Middle East and North Africa/Republic of Turkiye and Middle East, Europe, and the Americas using fiscal year 2024 data. Design: We quantified and compared full vaccination costs for refugees under two scenarios: (1) 'No VPR' and (2) 'VPR'. Refugees would receive no vaccines overseas and be fully vaccinated after US arrival under 'No VPR'. Under 'VPR', refugees receive one or two doses of selected vaccines overseas before completing vaccination schedules after arrival. Main Outcomes: Costs were reported in 2023 US dollars for 'VPR' and 'No VPR' scenarios and further subdivided by grouping countries/sites depending on whether the International Organization for Migration (IOM) provides vaccination services for refugees (IOM sites) versus non-IOM providers (non-IOM sites). Results: 'VPR' resulted in average net cost savings of $147 per person or $14.7 million per 100,000-refugee cohort compared to providing all vaccines after US arrival ('No VPR'). 'VPR' was cost-saving across most regions, except for IOM sites in Europe, where a net cost of $44 per person was observed. Net cost savings per person were highest for IOM sites in Africa ($333). Conclusions: VPR remains a cost-saving strategy, while protecting US-bound refugees' health and US health security by preventing disease outbreaks during resettlement.

15.
arXiv (CS.AI) 2026-06-15

When Good Verifiers Go Bad: Self-Improving VLMs Can Regress on New Tasks

作者:

arXiv:2606.14629v1 Announce Type: cross Abstract: Verifier-driven self-DPO is a common recipe for self-improving production visual-language models. In this setup, a frozen verifier scores candidate generations, the top- and bottom-scoring candidates form a preference example, and DPO updates the learner. The deployment-time assumption is monotone: a stronger verifier should yield a stronger student. We show that this assumption can fail because verifier quality is highly task-specific. On a four-rung open-source verifier ladder across MathVista, MMMU, and BLINK, the same verifiers that are above-threshold and improve a Qwen-3-VL-2B student on MathVista become sub-threshold on MMMU, where their task-rubric accuracy drops to 8% to 23%. In this regime, every verifier we tested silently regresses the student, producing drops of 3.4 to 10.9 percentage points below the frozen baseline while the DPO training loss continues to decrease. The regression replicates on a second student, Qwen-2.5-VL-3B. Moreover, within the failure regime, damage is confidence-inverted: the more accurate-but-still-wrong verifier causes larger regression than a near-random verifier, suggesting that progress-gated replay amplifies confidently wrong preference pairs. We give a compact mechanistic explanation via a variance theorem for progress-gated replay and its direction-mismatch failure mode. The deployment message is operational rather than purely diagnostic: before running any verifier-driven loop, teams should measure target-task rubric accuracy, rank verifiers by target-task rubric quality rather than parameter count, and treat diminishing returns in above-threshold regimes as a verifier-side compute budget cap.

17.
arXiv (CS.CL) 2026-06-12

Can Factual Opinions Be Edited (Manipulated) in Large Language Models?

Large Language Models (LLMs) are increasingly integrated into various domains, making knowledge editing techniques crucial yet potentially hazardous. Current editing methods primarily target atomic facts, overlooking the significant risks associated with manipulating factual opinions, e.g., documented stances of public figures on societal issues. Such manipulation could reshape public images, influence elections, and alter societal views. To systematically assess this threat, we introduce the Factual Opinion Editing with Evidence (FOE) benchmark, which encompasses 261 public figures, 19 issue categories, and 2,178 complete opinion records. Our evaluations demonstrate that current editing techniques struggle significantly with factual opinions, often achieving only superficial changes while failing to preserve consistency between the edited opinion and the supporting evidence generated by the model. To address this limitation, we further propose a simple yet effective Self-Generated Evidence-Aligned method that achieves opinion-evidence alignment without relying on explicit instructions. Together, our benchmark and method provide a foundation for understanding the emerging security implications of factual opinion editing in LLMs.

18.
arXiv (CS.LG) 2026-06-16

{\alpha}-Fair Insurance Pricing: A Fairness Continuum

arXiv:2606.14898v1 Announce Type: new Abstract: Fairness in insurance pricing remains a long-standing and deeply debated puzzle. On one hand, insurers, driven by profitability considerations, set premiums that differentiate across individual risks to achieve actuarial fairness. On the other hand, insurance serves a critical societal function by pooling risks across a population, motivating cross-subsidization among groups to promote solidarity fairness. The tension between these two competing notions of fairness makes insurance pricing inherently complex, particularly in modern settings where granular data allow for increasingly fine risk differentiation and regulators face growing pressure to protect vulnerable groups. To address this challenge, we propose an $\alpha$-Fair Individual Solvent Premium ($\alpha$-FISP) framework for insurance pricing that explicitly captures the trade-off between actuarial and solidarity fairness while guaranteeing solvency, a fundamental requirement in insurance operations. We formulate the pricing problem as a constrained optimization task, where actuarially fair premiums are adjusted subject to budget constraints on cross-subsidization within each risk class. This formulation naturally yields a family of solutions parameterized by $\alpha$, tracing a continuum between purely actuarial and purely solidarity-based pricing and enabling decision-makers to select an operating point along this fairness spectrum. We derive theoretical guarantees for the proposed framework. Numerical experiments show that $\alpha$-FISP is computationally tractable and aligns well with the U.S. regulatory regimes featuring heterogeneous state-level fairness requirements.

19.
arXiv (CS.AI) 2026-06-18

What Does the Weight Norm Control in Grokking? Logit-Scale Mediation under Cross-Entropy

arXiv:2606.18465v1 Announce Type: cross Abstract: Grokking, the delayed jump from memorization to generalization, is usually tied to the weight norm: a smaller norm generalizes sooner. We ask what the norm actually controls. Holding the weight norm fixed by clamping and varying only an output temperature, we slide the grokking delay across its entire norm-induced range under cross-entropy; matching the effective logit scale back to baseline recovers about 85% of the delay at two moduli. Across a grid of norms and temperatures the delay collapses onto the logit scale alone (R2 = 0.97), with the norm adding 1-2% beyond it. The effect is loss-dependent: under mean-squared error the logit scale is pinned and the norm acts through a different route. A memorization control, a float64 softmax-collapse audit, and a no-LayerNorm transformer point to the same channel. Forking arms from one identical state, the delay follows the held norm value and not the clamp operation, which closes a rescaling-artifact concern. The proximal variable is the logit scale and the softmax saturation it drives; the weight norm is only an upstream handle. All numbers, tables, and figures reproduce from released code and data.

20.
arXiv (CS.AI) 2026-06-12

Two-Layer Linear Auto-Regressive Models Estimate Latent States

arXiv:2606.12691v1 Announce Type: cross Abstract: Auto-regressive models have emerged as powerful tools for sequential data, from language to video. Understanding how and why these models learn latent representations remains an open theoretical question. In this work, we demonstrate that when trained by empirical risk minimization on data from partially observed linear dynamical systems, two-layer linear auto-regressive models naturally learn to approximate Kalman filtering. In particular, we show that the learned hidden representation coincides, up to a similarity transformation, with the state estimates produced by the optimal (Kalman) filter, even though the model has no explicit knowledge of the underlying dynamics or state. The result follows from three main insights. First, we establish that the Kalman filter is well approximated by an auto-regressive model with bounded truncation error. Second, we show that despite non-convexity, the two-layer optimization landscape is benign, i.e., all stationary points are either strict saddles or global minima. Finally, as our main contributions, we provide finite-sample guarantees on prediction error, parameter estimation error, and latent state recovery. Numerical simulations support the theoretical results and demonstrate that the latent representations of auto-regressive models recover state estimates.

21.
arXiv (CS.AI) 2026-06-12

ARROW: Augmented Replay for RObust World models

arXiv:2603.11395v3 Announce Type: replace-cross Abstract: Continual reinforcement learning challenges agents to acquire new skills while retaining previously learned ones with the goal of improving performance in both past and future tasks. Most existing approaches rely on model-free methods with replay buffers to mitigate catastrophic forgetting; however, these solutions often face significant scalability challenges due to large memory demands. Drawing inspiration from neuroscience, where the brain replays experiences to a predictive World Model rather than directly to the policy, we present ARROW (Augmented Replay for RObust World models), a model-based continual RL algorithm that extends DreamerV3 with a memory-efficient, distribution-matching replay buffer. Unlike standard fixed-size FIFO buffers, ARROW maintains two complementary buffers: a short-term buffer for recent experiences and a long-term buffer that preserves task diversity through intelligent sampling. We evaluate ARROW on two challenging continual RL settings: Tasks without shared structure (Atari), and tasks with shared structure, where knowledge transfer is possible (Procgen CoinRun variants). Compared to model-free and model-based baselines with replay buffers of the same-size, ARROW demonstrates substantially less forgetting on tasks without shared structure, while maintaining comparable forward transfer. Our findings highlight the potential of model-based RL and bio-inspired approaches for continual reinforcement learning, warranting further research.

22.
arXiv (CS.LG) 2026-06-11

On Regret Bounds of Thompson Sampling for Bayesian Optimization

arXiv:2603.09276v2 Announce Type: replace-cross Abstract: We study a widely used Bayesian optimization method, Gaussian process Thompson sampling (GP-TS), under the assumption that the objective function is a sample path from a GP. Compared with the GP upper confidence bound (GP-UCB) with established high-probability and expected regret bounds, most analyses of GP-TS have been limited to expected regret. Moreover, whether the recent analyses of GP-UCB for the lenient regret and the improved cumulative regret upper bound can be applied to GP-TS remains unclear. To fill these gaps, this paper shows several regret bounds: (i) a regret lower bound for GP-TS, which implies that GP-TS suffers from a polynomial dependence on $1/\delta$ with probability $\delta$, (ii) an upper bound of the second moment of cumulative regret, which directly suggests an improved regret upper bound on $\delta$, (iii) expected lenient regret upper bounds, and (iv) an improved cumulative regret upper bound on the time horizon $T$. Along the way, we provide several useful lemmas, including a relaxation of the necessary condition from recent analysis to obtain improved regret upper bounds on $T$.

23.
arXiv (CS.LG) 2026-06-19

Online Dynamic Batching with Formal Guarantees for LLM Training

arXiv:2606.19989v1 Announce Type: cross Abstract: Modern LLM training breaks a core assumption behind offline batch samplers: the true training cost of a sample is only observable after preprocessing, augmentation, templating, tokenization, and multimodal visual-token expansion. Unless one pays for a preprocessing- and augmentation-dependent length cache, batch construction is therefore blind to the quantity that determines padding, memory use, and GPU saturation. We introduce Online Dynamic Batching (ODB), a DataLoader-side drop-in system that moves batch formation to this point of accurate observability while preserving DDP step alignment. We formalize this synchronization requirement as the Distributed Group Alignment Problem and prove deadlock-free bounded termination with default join-mode identity coverage and opt-in non-join sample-quota closure. ODB requires no model, optimizer, or attention-kernel changes and is released as online-dynamic-batching with lightweight trainer adapters. Across public 2B/8B Qwen3-VL runs on UltraChat/LLaVA/ShareGPT4o, ODB improves literal emitted-sample throughput vs. fixed-batch Standard by 1.58-2.51x on single-node Full FT/LoRA and 1.71-3.78x on two-node Full FT, with Standard-comparable quality; production MM-Mix reaches 4.43x. Against GMT/BMT offline token-budget oracles, ODB is within 15% on UltraChat/LLaVA and faster on high-CV ShareGPT4o: 2.24-2.39x single-node Full FT/LoRA and 3.06-3.69x two-node Full FT. Together, ODB occupies the online/drop-in regime for high-heterogeneity LLM fine-tuning: large throughput gains at Standard-comparable quality, formal DGAP guarantees, and no length-cache precompute or kernel rewrites.

24.
arXiv (math.PR) 2026-06-17

Optional Stopping for Superhedging Supermartingales

arXiv:2606.17452v1 Announce Type: new Abstract: Superhedging supermartingales, introduced by the authors in previous work, are non-probabilistic processes defined via subadditive outer integrals that carry a purely financial interpretation in terms of superhedging cost. Building on the Leinert-König theory of non-lattice integration, the present paper establishes several results that are classical in probability theory but whose non-probabilistic proofs require fundamentally new arguments: (i) a tower inequality for the conditional outer integral \overline{\sigma}_j applied at stopping times, reducing to equality when the integrand is conditionally integrable; (ii) three versions of Doob's optional stopping theorem, organised by the class of supermartingale and the range of the stopping times; and (iii) Dubins' upcrossing inequality in both finite- and infinite-time horizons. A key structural result, property (K)-a.e., identifies conditions under which the two superhedging operators \overline{\sigma}_j and \overline{I}_j coincide on non-negative functions, extending the scope of all preceding results to the positive operator \overline{I}_j. None of the proofs invoke classical measure-theoretic tools; in particular, (classical) integrability and measurability are not assumed. The analogues of classical stochastic results acquire a purely financial interpretation and, in this way, gain depth and generality by providing a context that is independent of any a priori probabilistic structure.

25.
arXiv (CS.CL) 2026-06-19

Quantifying Aleatoric Uncertainty of In-Context Learning for Robust Measure of LLM Prediction Confidence

In-Context Learning (ICL) allows LLMs to adapt to new tasks from a few demonstrations, but its reliability remains a concern: predictions are highly sensitive to both prompt design and the model's ability to understand the context, obscuring whether failures arise from data properties or model limitations. Uncertainty decomposition-separating aleatoric from epistemic sources-is particularly crucial in this setting, yet existing methods, designed for standard generation tasks, fail to capture the unique dynamics of ICL. To address this, we introduce a concept of self-function vectors, built upon Bayesian views and the mechanistic interpretability of ICL. These vectors leverage internal model representations to model the latent concept learned during in-context prompting, thereby enabling a direct estimation of aleatoric uncertainty within a Bayesian framework and circumventing the reliance on brittle input or decoding manipulations. Given the lack of established benchmarks and suitable evaluation protocols, we also propose the first and rigorous evaluation protocol, in which data is manipulated in controlled ways so as to quantify aleatoric uncertainty precisely and separately from epistemic uncertainty. With this new evaluation framework, initially grounded in synthetic tasks for conceptual development and subsequently extended to real-world datasets, we show that our proposed methodology can measure uncertainty of LLM predictions made under ICL more reliably than existing alternative methods. Moreover, we show it can be used as a practical tool for trustworthy-related applications, such as hallucination detection. Our findings pave a new direction for connecting the quantitative view of uncertainty with the mechanistic understanding of model behavior.