Academic Intelligence · Curated Daily

Explore the Frontier of Global Academia

AcademicHub aggregates real-time literature from top journals and preprint platforms. Build your personal research radar and let large language models compile cross-disciplinary analysis briefings automatically.

01.
arXiv (quant-ph) 2026-06-19

Quantum Algebraic Diversity: Single-Copy Density Matrix Estimation via Group-Structured Measurements

arXiv:2604.03725v3 Announce Type: replace Abstract: We extend the algebraic diversity (AD) framework from classical signal processing to quantum measurement theory. The Quantum Algebraic Diversity (QAD) Theorem establishes that a group-structured positive operator-valued measure (POVM) applied to a single copy of a quantum state produces a full-rank, group-averaged density matrix estimator whose eigenbasis and eigenvalue ordering track those of the true density matrix, with a bias toward the symmetrized state, analogous to the classical recovery of covariance eigenstructure from a single observation. We establish a Classical-Quantum Duality Map connecting classical covariance estimation to quantum state tomography, and an Optimality Inheritance Theorem showing that classical group optimality transfers to quantum settings via the Born map within the group-averaged family. SIC-POVMs are identified as AD with the Heisenberg-Weyl group and mutually unbiased bases as AD with the Clifford group, revealing the hierarchy $\mathrm{HW}(d) \subseteq \mathcal{C}(d) \subseteq S_d$ that mirrors the classical $\mathbb{Z}_M \subseteq G_{\min} \subseteq S_M$. The double-commutator eigenvalue theorem gives polynomial-time adaptive POVM selection. A worked qubit example shows the group-averaged estimator from a single computational-basis measurement, averaged over a matched $\mathbb{Z}_2$ group, reaching fidelity 0.99 where standard single-basis tomography gives a rank-1 estimate of fidelity 0.80. Monte Carlo simulations for $d = 2$ to $13$ confirm fidelity above 0.90 from a single outcome while standard fidelity degrades as $\sim 1/d$. The growing ratio reflects collapse of the rank-1 standard estimator, not fewer copies per parameter: the biased single-copy estimator reduces the number of distinct measurement settings, not the per-parameter sampling cost, and a genuine copy reduction holds only under exact symmetry.

02.
arXiv (quant-ph) 2026-06-16

How Many Shots Are Enough for a Quantum Circuit?

arXiv:2606.16965v1 Announce Type: new Abstract: Quantum algorithms require repeated circuit executions, known as shots, to estimate output distributions accurately. Determining the minimal number of shots needed to meet a target accuracy is crucial to reduce costs and resource usage, especially on today's noisy and expensive quantum hardware. In this paper, we address the shot optimisation problem in a black-box setting, where no assumptions are made about the structure of the quantum circuit or the noise model of the backend. We introduce IncrementalExecution, a novel online framework that dynamically determines when to stop executing shots based on the principle of point of diminishing returns: the point at which additional shots no longer significantly alter the empirical distribution of a fixed circuit. The framework supports customisable policies for shot management, enabling flexible trade-offs between execution cost and result fidelity within static execution scenarios. We assess our proposal through an extensive experimental evaluation spanning 33,750 framework configurations across 180 unique static quantum circuit-backend combinations, for a total of 7.3M independent experiments. Unlike prior work that relies on problem-specific knowledge or algorithm-dependent assumptions (e.g., variational or adaptive workflows), our approach is applicable to a large set of static circuits and immediately deployable on current quantum cloud platforms.

03.
arXiv (CS.CL) 2026-06-18

TW-LegalBench: Measuring Taiwanese Legal Understanding

Large language models (LLMs) have shown impressive capabilities across diverse tasks, yet their performance on jurisdiction-specific legal reasoning remains underexplored. We present TW-LegalBench that utilizes Taiwanese legal system's rich official corpus open to the public to fill the gap in evaluating LLMs on Taiwanese law, among common-law benchmarks that focus on English sources and civil-law benchmarks focusing on sources of Simplified Chinese. TW-LegalBench comprises three task types: (1) over 16,000 multiple-choice questions (MCQs) across five years of official examinations in 18 professional domains; (2) 117 open-ended essay questions (OEQs) from examinations for legal professionals with official scoring rubrics; and (3) more than 14,000 legal judgment prediction (LJP) instances covering hundreds of crime categories. We evaluate 13 LLMs using accuracy for MCQs, a decomposed LLM-as-Judge framework based on the scoring rubric points for OEQs, and metrics for sentencing accuracy and statute citation for LJP. Our results reveal that top-performing models exceed the passing threshold for qualified lawyers (passing rate: 11%) but fall short of that for judges and prosecutors (passing rate: 1~2%). For LJP, while models demonstrate reasonable verdict type accuracy and sentence prediction capability, they struggle to cite exact legal articles. These findings highlight that reliable legal text generation remains challenging for LLMs, even though their performance on qualification examinations approaches human level.

04.
arXiv (CS.CL) 2026-06-17

The Critical Role of Model Selection in Causal Inference: A Comparative Analysis of Classification Models within the InferBERT Framework for Pharmacovigilance

Distinguishing causal adverse drug events (ADEs) from spurious correlations remains a central challenge in pharmacovigilance. The InferBERT framework integrates transformer models with Do-calculus, but its success hinges on the underlying classification model. This study evaluates the impact of model choice in InferBERT, assessing whether simpler models suffice, if domain-specific pre-training helps, whether scaling to LLMs improves causal detection, and the effect of post-hoc calibration. We performed a comparative study on two benchmarks: Analgesics-induced Acute Liver Failure (AILF) and Tramadol-related Mortalities (TRAM). Four models were evaluated-XGBoost (baseline), ALBERT (original InferBERT), BioBERT (biomedical transformer), and Med-LLaMA (medical LLM)-using 5-fold cross-validation repeated over 20 runs. We measured accuracy, Expected Calibration Error (ECE) pre- and post-isotonic regression, and Jaccard concordance of causal terms with PRR, ROR, and EBGM; significance was tested with paired t-tests. BioBERT achieved the highest accuracy on both datasets, while Med-LLaMA underperformed despite its size and parameter-efficient fine-tuning. Domain-specific pre-training was decisive. Calibration improved ECE but had mixed effects on accuracy and causal discovery. BioBERT's superiority also yielded the strongest concordance with traditional pharmacovigilance signals. These results show that domain-specific pre-training provides a clear advantage over simpler baselines and larger LLMs. Investing in manageable, domain-aware models is more effective for computational pharmacovigilance than simply scaling model size.

05.
arXiv (CS.LG) 2026-06-12

Normative Robustness as a Frontier for Non-Verifiable Reasoning in LLMs

arXiv:2606.12731v1 Announce Type: new Abstract: As LLMs increasingly serve in advisory and deliberative roles, users rely on them for non-verifiable reasoning in domains lacking objective ground truths. However, traditional evaluations of LLM reasoning focus almost exclusively on fact-based domains, such as mathematics and science, leaving uncertainty over whether and to what degree models can handle ambiguous, subjective, or value-laden problems over time. To address this concern, we propose moral reasoning as a paradigmatic subdomain of non-verifiable reasoning. We define moral robustness as a model's capacity to exhibit sound moral reasoning across time and contexts, and we introduce a scalable, adversarial, multi-turn evaluation framework to empirically measure this capability. We simulate 48,000 user-agent moral deliberations across four frontier LLMs, varying premise relevance, premise order, conversation duration, and the user's stated moral view. We find that models successfully ignore morally-irrelevant distractors, but shift their reasoning by up to 6.5%, on average, towards the user's stated preferred moral view, and varying their reasoning depending on factors such as order (altering moral judgments by order in 13-22% of the cases) and duration (altering moral judgments between single-turn and multi-turn in 10-24% of the cases). Our analysis indicates that models tailor not just their final verdicts but their underlying justifications to align with a user's moral viewpoint - a failure mode we characterize as moral deliberative sycophancy.

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

Cross-Dataset Bloom Question Classification: Supervised Models and Prompted LLMs

Automatic Bloom's taxonomy classification of assessment questions can substantially reduce instructor workload, but labeling is subjective and teacher-dependent. Prior machine learning (ML) and deep learning (DL) approaches reported strong within-dataset results, yet were rarely evaluated in cross-dataset settings, leaving real-world generalizability unclear; meanwhile, LLM effectiveness for Bloom question classification has not been systematically studied. We evaluated the cross-dataset generalization of existing ML/DL methods and assessed LLMs with multiple prompting strategies on five datasets; the best prompting strategy combined in-context examples with course-specific action verbs. Supervised ML/DL models degraded substantially on unseen datasets, whereas LLMs were more stable, suggesting a robust alternative across diverse educational contexts. Based on the best prompting strategy, we also presented a lightweight UI that supports instructors in automatically classifying large question banks; a usability study indicated low workload and high usability.

07.
arXiv (quant-ph) 2026-06-19

Nearest-neighbour gates are all you need: High-rate quantum low-density parity-check codes on a planar grid

arXiv:2606.19482v1 Announce Type: new Abstract: High-performance quantum low-density parity-check codes promise substantial reductions in the overhead of fault-tolerant quantum computation, but most constructions require long-range connectivity or qubit shuttling, both of which are difficult to realise in superconducting architectures. Here we introduce a family of quantum low-density parity-check codes that, for the first time, combines planar open-boundary layouts, finite-size advantages over surface codes, and syndrome extraction using only nearest-neighbour gates on a square grid of qubits. The key idea is to generate check-data connectivity dynamically: nearest-neighbour iSWAP walks both define the stabiliser supports and implement their measurement, avoiding the need for a long-range hardware graph. The resulting circuits achieve optimal constant-depth stabiliser measurement, independent of code size, and naturally remove leakage from the system by exchanging the role of check and data qubits at each syndrome extraction round. We find finite-size instances such as a [[323,14,15]] code, whose code-efficiency ratio is nearly an order of magnitude larger than that of rotated surface-code patches. At around 30 circuit qubits per logical qubit, the best directional tile-code layouts reduce the per-logical per-round logical error rate by up to a factor of 1000 relative to rotated surface-code memories. These results show that the advantages of quantum low-density parity-check codes can survive compilation into strictly planar nearest-neighbour circuits, bringing low-overhead fault-tolerant memories closer to near-term hardware.

08.
arXiv (CS.LG) 2026-06-17

From Reasoning Traces to Reusable Modules: Understanding Compositional Generalization in Language Model Reasoning

arXiv:2606.18089v1 Announce Type: new Abstract: Post-training pipelines that combine supervised fine-tuning (SFT) with reinforcement learning (RL) have emerged as the key recipe for transforming large language models (LLMs) into robust reasoners. We argue that this combined success is driven by compositional generalization, which we formalize through a hierarchical latent selection model. In this framework, reasoning traces are generated by a cascade of discrete latent selection variables corresponding to reusable atomic modules, including both skills (local operations) and routing mechanisms (how intermediate information is selected, reused, and composed). Within this model, we theoretically show that SFT and RL play asymmetric, complementary roles: SFT supplies the raw module materials in compositional traces, and RL decomposes those traces to identify the latent atomic modules and enable compositional generalization. We design controlled experiments to validate this theory. Our results demonstrate that RL can extract atomic modules from compound traces supplied by SFT and recombine them to solve new configurations. Moreover, we find that training on compound traces yields stronger generalization than training on isolated atomic modules. Finally, we investigate the relationship between SFT and RL data and identify an effective protocol in which SFT ensures coverage of all atomic modules through compositional traces, while RL focuses on novel compositions outside the SFT support to drive exploration.

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

Region-Adaptive Sampling for Diffusion Transformers

Diffusion models (DMs) have become the leading choice for generative tasks across diverse domains. However, their reliance on multiple sequential forward passes significantly limits real-time performance. Previous acceleration methods have primarily focused on reducing the number of sampling steps or reusing intermediate results, failing to leverage variations across spatial regions within the image due to the constraints of convolutional U-Net structures. By harnessing the flexibility of Diffusion Transformers (DiTs) in handling variable number of tokens, we introduce RAS, a novel, training-free sampling strategy that dynamically assigns different sampling ratios to regions within an image based on the focus of the DiT model. Our key observation is that during each sampling step, the model concentrates on semantically meaningful regions, and these areas of focus exhibit strong continuity across consecutive steps. Leveraging this insight, RAS updates only the regions currently in focus, while other regions are updated using cached noise from the previous step. The model's focus is determined based on the output from the preceding step, capitalizing on the temporal consistency we observed. We evaluate RAS on Stable Diffusion 3 and Lumina-Next-T2I, achieving speedups up to 2.36x and 2.51x, respectively, with minimal degradation in generation quality. Additionally, a user study reveals that RAS delivers comparable qualities under human evaluation while achieving a 1.6x speedup. Our approach makes a significant step towards more efficient diffusion transformers, enhancing their potential for real-time applications.

10.
arXiv (CS.AI) 2026-06-19

LLM Doesn't Know What It Doesn't Know: Detecting Epistemic Blind Spots via Cross-Model Attribution Divergence on Clinical Tabular Data

arXiv:2606.19509v1 Announce Type: new Abstract: Large language models (LLMs) are increasingly applied to structured clinical data, yet whether they can recognize the limits of their own knowledge on such tasks remains unexplored. We study this question through the lens of cross-model attribution divergence with the goal of reducing epistemic uncertainty for structured tasks, comparing Qwen 2.5 7B and XGBoost on a prediction task via attribution divergence analysis. We report four findings. First, LLM verbalized confidence is epistemically vacuous, it outputs a near-constant (0.856-0.937) regardless of whether accuracy is 49% or 75.3%, tracking prompt format rather than prediction quality. Second, the LLM exhibits an inverse difficulty effect: accuracy drops to 64.8% when XGBoost is 99% correct, but matches XGBoost (73.8% vs. 73.1%) when it is moderately uncertain. Third, few-shot examples and SHAP-derived feature evidence are orthogonal, super-additive interventions: they reduce the Attribution Disagreement Score (ADS) from 1.54 to 0.38 and improve accuracy from 49% to 75.3% without training. Fourth, a cross-model calibrator that determined LLM reliability using attribution divergence signals reduces expected calibration error from 0.254 to 0.080, replacing uninformative verbalized confidence with patient-specific reliability estimates, without accessing model internals or requiring repeated inference. We frame these findings as a cold start problem for LLMs on structured data and outline a path toward genuine epistemic self-awareness.

11.
arXiv (CS.CV) 2026-06-15

Schrödinger's Navigator: Imagining an Ensemble of Futures for Zero-Shot Object Navigation

Zero-shot object navigation (ZSON) requires robots to find target objects in unseen environments without task-specific fine-tuning or pre-built maps, a key capability for general-purpose service robots. Yet methods that perform well in simulation often degrade in cluttered real-world scenes with severe occlusion and latent hazards, where large unseen regions make single-scene inference brittle and unsafe. We propose Schrödinger's Navigator, a belief-aware framework that reasons at inference time over multiple trajectory-conditioned imagined 3D futures. Given candidate paths, a trajectory-conditioned 3D world model predicts hypothetical observations and maintains a superposition of plausible scene realizations rather than committing to one map. An adaptive occluder-aware sampler directs imagination to uncertainty-critical regions, while a Future-Aware Value Map (FAVM) aggregates imagined futures for robust, proactive action selection. Experiments in simulation and on a physical Go2 quadruped show that Schrödinger's Navigator outperforms strong ZSON baselines, improving hidden-target discovery and risk-aware waypoint selection in occlusion-heavy navigation scenarios. These results highlight imagined 3D futures as a scalable and generalizable strategy for zero-shot navigation in uncertain real-world environments.

12.
arXiv (CS.AI) 2026-06-15

Q-Net: Queue Length Estimation via Kalman-based Neural Networks

arXiv:2509.24725v4 Announce Type: replace-cross Abstract: Estimating queue lengths at signalized intersections is a long-standing challenge in traffic management. Partial observability of vehicle flows complicates this task despite the availability of two privacy-preserving data sources: (i) aggregated vehicle counts from loop detectors near stop lines, and (ii) aggregated floating car data (aFCD) that provide segment-wise average speed measurements. However, how to integrate these sources with differing spatial and temporal resolutions for queue length estimation is rather unclear. Addressing this question, we present Q-Net: a queue estimation framework built upon a state-space formulation. This design addresses key challenges in queue modeling, such as violations of traffic conservation assumptions. Q-Net follows the Kalman predict-update structure and maintains physical interpretability in both the state evolution and measurement models. Q-Net uses an AI-augmented Kalman filter to learn time-varying gain dynamics from data. The framework supports real-time implementation and improves spatial transferability by grouping aFCD measurements into fixed-size local groups, making the number of learnable parameters independent of section length. Evaluations on urban main roads in Rotterdam, the Netherlands, show that Q-Net outperforms baseline methods, tracks queue formation and dissipation accurately, and mitigates aFCD-induced delays. By combining data efficiency, interpretability, real-time applicability, and spatial transferability, Q-Net makes accurate queue length estimation possible without costly sensing infrastructure like cameras or radar.

13.
medRxiv (Medicine) 2026-06-10

Cortical activity during narrative discourse production in individuals with post-stroke aphasia and controls measured via functional near-infrared spectroscopy

Introduction: Aphasia is an acquired language disorder with a significant negative functional impact. Much of the research on aphasia has focused on word-level language comprehension and production. Further evaluation of discourse-level tasks, both at behavioral and neural levels, will allow for an ecologically valid understanding of the functional implications of language impairment in this population. Method: This study evaluated bilateral frontal, temporal, and parietal cortical activity during computer-based narrative production in 14 young neurotypical individuals, 17 individuals with post-stroke aphasia, and 15 age-matched neurotypical participants using functional near-infrared spectroscopy (fNIRS). Oxygenated hemoglobin (HbO) was measured during narrative production following short video clips and compared to HbO during counting aloud. In addition, behavioral measures quantifying in-task performance were correlated with averaged HbO values. Results: Young neurotypical individuals showed greater cortical activity in bilateral language regions for narrative production compared to counting aloud. In contrast, people with aphasia showed positive condition-related effects in the right frontal ROI and the age-matched group showed positive condition-related effects in the left frontal and right precentral ROIs. Each group showed different patterns in relationships between cortical activity and discourse performance measures. Conclusion: Overall, young participants showing more consistent condition-related effects for narrative discourse production than individuals with aphasia and age-matched controls. This study shows the potential for fNIRS to evaluate cortical activity for ecologically valid language tasks in individuals with post-stroke aphasia.

14.
arXiv (CS.LG) 2026-06-19

Reinforcement Twinning for Hybrid Control of Flapping-Wing Drones

arXiv:2505.18201v2 Announce Type: replace-cross Abstract: Controlling flapping-wing drones requires controllers that handle time-varying, nonlinear, underactuated dynamics from incomplete, noisy sensor data. Recent advances in artificial intelligence (AI), particularly reinforcement learning (RL), have opened new perspectives for addressing such complex control problems through data-driven policy optimization from interaction with the environment. Yet purely data-driven methods are sample-inefficient, demanding extensive, sometimes unsafe exploration, especially without guiding physical models. This motivates hybrid AI-physics frameworks. This article proposes a hybrid model-free/model-based flight-control approach using the reinforcement twinning algorithm. The model-based (MB) component uses an adjoint formulation and an adaptive digital twin continuously identified from live trajectories; the model-free (MF) component uses RL. The two agents share knowledge via transfer learning, imitation learning, and shared experience between the real environment and the digital twin, coordinated by a policy referee that selects which agent acts in reality based on digital-twin performance and a real-to-virtual consistency ratio. The framework is evaluated for the longitudinal control of a flapping-wing drone, modelled as a nonlinear time-varying system driven by quasi-steady aerodynamic forces. The hybrid strategy is tested under three adaptive-model initializations: (1) offline identification from existing data, (2) random initialization with fully online identification, and (3) offline pre-training with biased parameters followed by online adaptation. In all cases, the hybrid framework improves performance, robustness, and sample efficiency over purely model-free and purely model-based approaches.

15.
arXiv (CS.AI) 2026-06-16

Tensor-Coord: Algebraic Decomposition of Joint Plan Tensors for Conflict-Free Multi-Agent LLM Planning

Authors:

arXiv:2606.16478v1 Announce Type: new Abstract: Large language models (LLMs) remain limited in multi-agent planning because independently generated plans can create coordination failures such as spatial collisions, resource contention, and temporal deadlocks. We introduce Tensor-Coord, a multilinear algebra framework that represents the joint plan of N agents as a third-order tensor \(T \in R^{N \times H \times A}\) over agents, timesteps, and actions. Canonical Polyadic (CP) and Tucker decompositions are used to identify latent coordination structure. The minimal epsilon-approximate CP rank R* defines a computable coordination complexity measure, with \(CC(Pi)=(R*-N)/N\). We prove that R*=N is necessary and sufficient for plan independence. The residual \(E=T-T_{R*}\) defines a conflict score over agent pairs, timesteps, and actions, localizing failures without domain-specific rules. Tucker factors provide interpretable agent roles, temporal phases, and action clusters that are converted into natural language constraints for iterative LLM replanning. Experiments on multi-robot delivery tasks across Easy (2 agents, 5x5 grid), Medium (3 agents, 5x5 grid), and Hard (4 agents, 5x5 grid) settings show convergence to conflict-free plans in 100% of 2-agent cases within 1.4 iterations on average, 80% of 3-agent cases within 3.2 iterations, and 60% of 4-agent cases within 4.0 iterations. CP rank scaled approximately linearly as \(R*(N) = 3.9N + 0.5\), supporting its use as a predictor of coordination complexity.

16.
arXiv (CS.CL) 2026-06-17

Environment-Grounded Automated Prompt Optimization for LLM Game Agents

LLM agents in interactive environments are highly sensitive to their prompts, yet prompt engineering remains a manual, task-specific process. We introduce an automated prompt optimization framework for LLM agents that decomposes the observation-to-action pipeline into a goal-conditioned descriptor agent and an action selection agent, and iteratively refines each module's prompt through an LLM-driven evolutionary loop guided by environment returns. We propose a behavior analyzer to attribute episode outcomes to specific prompt components, and a mutator to propose targeted revisions to the prompt, before validating them through environment rollouts. We evaluate on all five BabyAI tasks in the BALROG benchmark, comparing our pipeline against BALROG's RobustCoTAgent under both plain and guided prompt initializations. Optimization improves performance consistently across tasks and conditions, without requiring updates to the model weights. On PutNext, a multi-step coordination task where the RobustCoTAgent achieves 0% success, our framework reaches up to 72.5% success rate using the same underlying LLM with optimized prompts. These results suggest that a multi-agent framework, combined with automatic prompt optimization, enhances LLMs without the need for fine-tuning or extensive human supervision.

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

The Measurement Gap in the Automation of EU Law: Benchmarking Doctrinal Legal Reasoning under the EU AI Act

Large language models now produce legal text of at least median quality, yet no existing benchmark can evaluate whether they perform doctrinal legal reasoning, which forms the interpretive core of legal work, rather than the ancillary, paralegal tasks that most current legal-AI evaluations measure. This measurement gap is not only methodological but legal: the EU AI Act makes "appropriate accuracy" a binding requirement for high-risk AI used in the judicial domain, yet that requirement cannot acquire operational content without the very doctrinal-reasoning benchmark the field lacks.

18.
arXiv (CS.AI) 2026-06-15

Active Inference for Adaptive Traffic Signal Control in Noisy Nonstationary IoT Environments

arXiv:2606.13698v1 Announce Type: cross Abstract: Urban traffic signal control at IoT-instrumented intersections must remain effective under sensor occlusion, weather attenuation, and nonstationary demand. Conventional controllers degrade under these conditions, and learned policies remain difficult to audit. To address these challenges, we propose an active inference controller for a four-arm signalized intersection that dynamically selects phases by minimizing expected free energy (EFE) over Gaussian beliefs about per-direction congestion levels, yielding a fully traceable decision pipeline. We benchmark the controller in a SUMO traffic simulator against a rule-based heuristic and a deep Q-network (DQN) across four scenarios that progressively increase noise and nonstationarity, spanning sensor occlusion, adverse weather, and stochastic accidents. Across 100 independent random evaluations per scenario, active inference attains the lowest idle times and CO2 emissions in the noisiest scenarios (56,977 s and 29.12 kg vs. 71,741 s and 30.56 kg for DQN). These gains come at a modest cost in bus priority service rate and phase switch frequency.

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

Judging to Improve: A De-biased VLM-as-3D-Judge Protocol for Single-Image 3D Generation

arXiv:2606.20364v1 Announce Type: new Abstract: A companion study established a de-biased, cross-model VLM-as-3D-judge that reliably ranks single-image-to-3D mesh quality where cheap geometry and CLIP proxies fall short. This paper asks: can that judge's preferences specialize a strong open generator, TRELLIS, on one asset class (furniture), cheaply and without human labels? Taking the judge from ranking to optimization is where the work lives. Pushing a VLM judge into the training and evaluation loop exposes failure modes ranking never triggered, so our contribution is an optimization-grade hardening of the judge: a training judge (Qwen2.5-VL-7B) held distinct from an evaluation judge (InternVL3-8B) to break circularity; position-bias correction; and fixes for three failure modes (image overload, geometry-hiding splat renders, and reference-free judging that rewards clean-but-wrong outputs), with calibration evidence (clear-gap win-rate 0.83-1.0; base-vs-base ~0.5). Using this protocol as an independent evaluator, and working only from public models and data with lightweight parameter-efficient adaptation, we find our methods match the strong base rather than exceed it. Independent base samples carry essentially no learnable preference (0.94 order-flip rate), so signal must be engineered by quality-contrastive construction. Across six adaptation methods, two input regimes, and a severity sweep, the most targeted - conditioner repair under severe degradation - reaches parity (0.50) with the base, while no method clears the >=65% win-rate target. The result is mechanistic: clean inputs saturate the judge, flow-DIT fine-tuning washes out through the sampler, and conditioning repair is the locus that moves geometry. Win-rates are directional at n=8 objects. Matching a strong public-data base with cheap adaptation is itself informative: exceeding it needs more than lightweight PEFT on public data, and the judge protocol is reusable.

21.
arXiv (CS.CV) 2026-06-16

Is My Vision-Language Data in Your AI? Membership Inference Test (MINT) Demo 2

We present the Membership Inference Test (MINT) Demo 2, a framework designed to improve transparency in machine learning training processes. MINT is a technique for experimentally determining whether specific data were used during machine learning model training. We establish the theoretical framework and propose multiple architectures for MINT depending on the amount of information known about the models that are being audited. Experimental results using a popular face recognition model, 4 state-of-the-art LLMs, and multiple, diverse, and large-scale public image and text databases achieve promising accuracy levels in the detection of training data of up to 90%. Building on these results, we introduce a comprehensive web platform1 that expands these capabilities to image and text modalities. The platform integrates a diverse technological stack, including MINT, aMINT, and gMINT, allowing users to audit a wide range of models. This demonstrator aims to promote AI transparency and provides a practical tool to foster compliance with emerging AI regulations.

22.
arXiv (CS.CL) 2026-06-17

Rift: A Conflict Signature for Deception in Language Models

Authors:

A model that lies while knowing the truth is the central case ELK cannot handle with behavioral evaluation alone. We ask whether such deception leaves an internal signature distinguishing it from honest error. Our key move is a control for wrongness: we contrast a sleeper agent (knows the truth, lies on trigger) against a naive liar (fine-tuned to emit the same wrong answers with no honest training). Both produce identical wrong outputs; any difference is about knowledge conflict, not incorrectness. We find deceptive forward passes carry a conflict signature - 2.1-2.3x higher residual rank than naive-liar passes on the same wrong answer - strong enough to identify which of two responses is the lie with 100% accuracy and no labels, across GPT-2 small/medium (three seeds) and three instruct models. Across Qwen2.5-1.5B/7B and Phi-3-mini, instructed deception raises residual rank on every tested fact (18/18, 40/40, 34/34); on Phi-3, lies separate perfectly from both honest answers and hallucinations (AUC 1.0, Wilcoxon p~6e-11). The signature survives strategic self-constructed deception (model invents its own lie, AUC 1.0), active concealment attempts (AUC 1.0), and length-controlled replication (20/20, AUC 1.0, p~1e-6). Using basis-free relative representations, a probe trained on one model family detects deception in two other families zero-shot (mean AUC 0.933), surviving simultaneous architecture and format change (AUC 0.821), and transfers across five languages (AUC 1.000, length-controlled). The signature is read-only: detectable but not injectable (0/8 both directions). Honest limitations and six negative experiments are documented in full.

24.
arXiv (CS.CL) 2026-06-11

When is Your LLM Steerable?

Activation steering offers a lightweight approach to control language models' behavior at inference time, but whether it succeeds or fails heavily depends on the prompt, concept, model, and steering configuration. Finding the regime and boundaries of successful steering typically requires expensive grid searches and post-hoc evaluation of full autoregressive rollouts. In this work, we investigate whether steerability can be predicted from the model's internal states at the beginning of the generation process, e.g., after generating the first few tokens, and how to leverage such a predictor to improve steering success rate. To this end, we first introduce ASTEER, a testbed including 1.4M steered generations, spanning 150 concepts with each steering success/failure labeled. Leveraging this testbed, we analyze the model's early decoding dynamics by extracting features that compare hidden states before and after steering across layers and initial decoding steps. These features help us understand how steering's effects propagate along layers and token positions, which provide key information for steerability prediction. We then train a Gradient Boosting Decision Trees (GBDT) classifier on these features to predict whether an intervention will under-steer, succeed, or over-steer without requiring full rollout. Our predictor achieves around 0.7 macro-F1 score on unseen concepts, demonstrating that early hidden states encode substantial, structured information about eventual steering efficacy. We further leverage this steerability predictor as guidance for steering strength searching, achieving near-optimal performance with a small fraction of decoding cost.

25.
arXiv (CS.CL) 2026-06-17

Security and Privacy Prompts in the Wild: What Users Ask LLMs and How LLMs Respond

Large language models (LLMs) are widely used to fulfill users' information needs; users ask LLMs about the weather, pose educational questions, and consult them for legal assistance. One particularly understudied area is digital security and privacy (S&P), where users may seek LLMs' help on how to secure their online accounts or protect their computers from cyber attacks. To the best of our knowledge, no prior study has collected or analyzed the S&P questions users ask LLMs; prior research on LLM response quality relied on expert-authored S&P misconceptions or FAQs rather than user queries. Drawing from WildChat, a dataset of 3.2M user-LLM conversations collected in the wild, our study identifies 14,727 S&P prompts and categorizes them into nine categories covering a wide range of S&P topics. From the S&P prompts, we sampled 450 and performed a thematic analysis to characterize the S&P questions users ask LLMs. Separate from the thematic analysis, we curated 270 advice-seeking S&P prompts, where users ask for recommendations, guidance, or specific S&P information. We measured LLM response quality and consistency when posing the prompt to LLMs 10 times. We found that commercial LLMs outperform open-weight models (GPT 5.5 provided "good enough" responses on 98% of prompts; Llama 4 on 47%). However, among prompts that received high-quality responses on average, commercial models sometimes produce contradictory responses across runs, risking confusing or misleading users.