arXiv:2607.05572v1 Announce Type: cross Abstract: Assuming the existence of axion-like particles (ALPs), the standard Quantum Electrodynamics (QED) amplitudes for the linear Breit-Wheeler (BW) pair production process are shown to be supplemented by an additional ALP-mediated amplitude. We compute the coherently combined cross section and show that, for pseudoscalar ALPs, interference with QED produces an asymmetric Fano-type resonance. We propose experimental observation of the BW process with real photons as a complementary probe of MeV-scale ALPs, coupling jointly with photons and electron-positron pairs, on the basis that it occupies a kinematic region not readily accessible to other ALP search methods.
Science Journals
arXiv:2607.01499v2 Announce Type: replace Abstract: Recent advances in Image-to-Video generation allow a single image to be animated into a convincing video under text guidance, raising serious copyright and privacy risks. We propose Anti-Prompt, an image protection approach that injects imperceptible perturbations into an image, inducing visible inconsistencies and structural failures in text-guided I2V generation. Our method is motivated by a simple empirical observation. When text guidance is removed from modern I2V models, generation quality degrades markedly, not only in motion realism but also in subject preservation, structural coherence, and temporal consistency. Building on this insight, Anti-Prompt exploits the model reliance on textual guidance by attenuating text-conditioned interactions during denoising while strengthening visual-only pathways. To further systematically evaluate protection effectiveness, we introduce a Video-LLM-assisted evaluation protocol that provides interpretable, frame-grounded analyses of generation artifacts and inconsistencies. Experiments on two representative I2V architectures demonstrate that our method achieves strong protection performance while improving efficiency and cross-model transferability.
arXiv:2601.14043v2 Announce Type: replace Abstract: We investigate the robustness of a curvature-weighted spectral precursor to dissipation in freely decaying three-dimensional incompressible turbulence. Building on our recent work in \emph{Physical Review Fluids} on the Taylor--Green vortex, we analyze direct numerical simulations using the shell-summed curl-of-vorticity spectrum, denoted here by $\mathcal{C}_{4}(k,t)$ and equivalent to a $k^4$-weighted energy spectrum in the modal incompressible sense. Extending the study across multiple initial conditions -- multi-mode ABC flows, a randomized low-wavenumber ABC field, the Taylor--Green vortex, and the Kida--Pelz flow -- we find a consistent temporal ordering: the characteristic time associated with the advance and saturation of the peak wavenumber of $\mathcal{C}_{4}(k,t)$ precedes the dissipation-peak time, which in turn precedes the characteristic time associated with the peak scale of the nonlinear energy-flux spectrum. We further probe Reynolds-number and scale-separation effects using Taylor--Green simulations at additional viscosities: the precursor ordering persists when adequate scale separation and resolution are maintained, but can change in the low-$R_\lambda$/limited-scale-separation regime. Throughout, we use explicit inspection of curvature-weighted spectra to distinguish physical peak evolution from cutoff-proximate artifacts. These results support robustness over the deterministic decaying-flow initial conditions examined here and clarify the practical role of Reynolds number, scale separation, and resolution when using curvature-weighted spectral diagnostics in decaying turbulence.
arXiv:2607.06205v1 Announce Type: cross Abstract: The Fredrickson-Andersen model with hyperparameter $K=1$ is a severely constrained kinetic lattice spin system, such that any site is temporarily blocked from changing its packing state (empty or occupied) if there is one or more occupied nearest neighbors. Starting from a completely random initial configuration with a fraction $\rho$ of sites being occupied, some of the sites may be permanently frozen to their initial state under this severe kinetic constraint. The remaining sites can switch states at least occasionally, and they form the unfrozen subsystem associated with the given initial configuration. In the present work we investigate thermodynamic phase transitions in such unfrozen subsystems of the two-dimensional square lattice and the three-dimensional cubic lattice by extensive numerical simulations. We demonstrate that the giant connected component of the unfrozen subsystem collapses at certain critical value $\rho_{c}$ of initial packing density, with $\rho_c = 0.2475$ for the square lattice and $\rho_c = 0.2809$ for the cubic lattice. This phase transition belongs to the same universality class of the conventional site percolation. We also observe that the ground states (densest packing configurations) experience a continuous crystal-to-glass phase transition at the critical value $\rho^* = 0.1423$ of initial packing density for the cubic lattice. For the two-dimensional square lattice we argue that long-range crystalline order is destroyed in the ground states as long as the initial packing density $\rho$ is positive.
arXiv:2607.06174v1 Announce Type: new Abstract: Extreme ultraviolet (EUV) radiation is a key tool for attosecond physics and lithography. However, strong material absorption limits the availability of transmissive optical elements at these wavelengths. Metaoptics exploit geometry to control the wavefront of transmitted light on the nanoscale and, due to their minimal thickness, promise to fill this gap. Here, we demonstrate the first EUV metaoptics for broadband applications: we design, fabricate, and experimentally investigate a blazed transmissive EUV metagrating and compare it with a focused-ion-beam-milled sawtooth-blazed grating serving as an in-situ reference. The metagrating achieves an angular dispersion of 0.04{\deg}/nm with a directionality (the ratio of the +1st and -1st diffraction order efficiency) of up to 5.8. The device shows phase-based operation up to 50 eV photon energy (down to 25 nm vacuum wavelength) and an octave-spanning bandwidth of 25 eV, doubling the previous spectral window addressable by metasurfaces. Comparing both gratings' performance reveals that, when accounting for fabrication constraints, EUV metasurfaces are competitive with free-form optics while offering scalability to large apertures and arbitrary phase profiles. Broadband transmissive operation removes the need for grazing incidence optics, defeating a major source of aberrations, and allows polarization-insensitive spectral analysis, enabling energy-resolved ultrafast spectroscopy in compact experimental configurations.
arXiv:2607.06181v1 Announce Type: new Abstract: Levenshtein's sequence reconstruction model plays an essential role in information retrieval of advanced memory systems, such as the DNA-based storage systems. In the model, a word $\mathbf{x}\in\mathbb{Z}_q^n$ is transmitted through $N$ noisy channels, and the goal is to recover it. Errors occurring in the channels usually involve substitutions, insertions and deletions. Our focus is on insertions. One of the main questions in this context is determining the minimum number of channels $N$ required to recover the transmitted word $\mathbf{x}$. The original formulation of the reconstruction problem requires that all the output words from the channels are distinct. However, different insertion errors may lead to the same output words. In this paper, we investigate two reconstruction models where the channels are allowed to produce identical output words even though different insertion errors occur in the channels. These two models, called \textit{the multiset model} and \textit{non-multiset model}, generalize the Levenshtein's model. We denote the minimum number of channels required to \textit{unambiguously} recover the transmitted word $\mathbf{x}\in\mathbb{Z}_q^n$ by $N_q^m(n,t)+1$ in the multiset model and $N_q^{nm}(n,t)+1$ in the non-multiset model, where $t$ is the exact number of insertions occurring in a channel. We determine $N_q^m(n,1)$ and $N_q^{nm}(n,1)$ for all $n$ and $q$, and show the somewhat surprising fact that $N_q^m(n,1)=N_q^{nm}(n,1)$. We also provide a full characterization of the words attaining this value and give a general lower bound on $N_q^m(n,t)$ for $t\ge1$ and a recursive upper bound. For $t=1$, we construct codes $C'\subseteq\mathbb{Z}_q^{n+2}$ from codes $C\subseteq\mathbb{Z}_q^n$ such that the number of channels required to determine the transmitted word $\mathbf{x}\in C'$ is small. This construction is shown to be optimal for certain parameters.
arXiv:2607.06184v1 Announce Type: new Abstract: Coding agents are ranked almost entirely by resolve rate: whether their final patch passes the target tests. Yet two agents can reach the same outcome through very different processes, and a single pass/fail label says nothing about why a run failed or why an accepted run spent extra steps, time, or tokens. This process evidence lives in the trajectory, which records a run's searches, reads, edits, tool calls, validation, and reversions. However, raw traces are heterogeneous and hard to compare across runs. We present TraceProbe, a trajectory-diagnostic framework that recovers what resolve rate hides. TraceProbe normalizes each raw run into a canonical nine-type action taxonomy with deterministic effect labels, then applies two rule-based modules: Insight names single-trajectory anti-patterns adapted from established debugging practice (e.g., search loops, verification skips), while Converge aligns pairs of runs and classifies where their behavior diverges under controlled references. Applying TraceProbe to 2,500 trajectories from five production settings on SWE-Bench Verified, we find that (i) file choice is too coarse to separate success from failure, whereas function selection and completion behavior localize it; (ii) Insight anti-patterns act mainly as corpus-level difficulty clues, with search loops the most stable; and (iii) even resolved runs differ in how quickly they reach relevant code and how much failed work they incur. Trajectory structure thus adds auditable diagnostic context to outcomes by localizing inspection targets, suggesting failure hypotheses, and prioritizing runs for review.
arXiv:2607.05798v1 Announce Type: new Abstract: Recent advancements in Multimodal Large Language Models (MLLMs) have evolved from static perception to interleaved visual-language reasoning, often referred to as ``thinking with images''. A basic operation in this reasoning process is to zoom in on regions of interest (often represented with bounding boxes) to acquire finer visual details. In this paper, we propose \textbf{Seg}mentation before \textbf{Answer}ing (SegAnswer), which shifts the unit of zoom-in from the popular bounding box to pixel-level segmentation mask. By employing fine-grained masks to isolate the target area from cluttered environments, segmented visual input yields a more precise region of interest, effectively filtering out redundant background and interfering objects. Furthermore, the discrete patches of segmented visual input align more seamlessly with how MLLMs structure visual tokens via positional embeddings. In experiments, we evaluate SegAnswer across diverse benchmarks, including high-resolution perception, general perception, and hallucination. It achieves consistent improvements and also exhibits considerable performance on segmentation tasks, validating its capability for reliable pixel grounding.
arXiv:2607.05818v1 Announce Type: new Abstract: Three-dimensional (3D) integration is a critical technique for enhancing transistor density, improving power efficiency, and reducing interconnect delays. However, as current demands and design complexity increase, power deliver networks (PDNs) are facing growing challenges.Careful planning of through-silicon vias (TSVs) is essential for ensuring reliable PDNs, where effective resistance serves as a vital metric for the reliability. Ill-planned TSVs often cause 3D IC with unevenly distributed effective resistance and consequently severer IR Drop.In this paper, we propose a GPU-accelerated framework on accurate effective resistance analysis for early stage 3D IC PDNs. The proposed framework achieves a speedup of 5 to 6 orders of magnitude compared to the conventional direct solver, while maintaining negligible deviations in both maximum and average relative errors.
arXiv:2508.09964v2 Announce Type: replace Abstract: Traditional methods of population synthesis produce stable and interpretable populations but cannot capture the interrelationships between household- and individual-level attributes. Recent deep learning methods offer this flexibility, yet can overfit high-dimensional attribute relationships without structural guidance and deviate from known structures. We develop a household level synthetic population generation framework that adapts the existing conditional input directed acyclic tabular generative adversarial network, or ciDATGAN, to multi person households. The framework combines household size specific data construction, directed acyclic graphs (DAG) informed dependency regularization, and conditional population inputs as deterministic anchoring to preserve intrahousehold associations. We apply the model to generate an open access synthetic population for New York State. The synthetic population includes nearly 20 million individuals and 7.5 million households in 2021. Validation against withheld benchmarks shows close agreement: joint distribution matching of Public Use Microdata Areas (PUMA), age, race, and disability against the full public use microdata sample (PUMS) yields an R-squared of 0.602 and a Jensen-Shannon distance of 0.280, pairwise Cramer's V differences between generated and benchmark records average below 0.007 at the person level, and classifier two-sample tests on complete household records yield AUC values of 0.527-0.549, close to chance level. The generated households reproduce cross member associations while increasing diversity by 10-17% over the PUMS sample and 13.2% over PopGen alone.
arXiv:2607.06226v1 Announce Type: new Abstract: Inverse Compton X-ray sources are laboratory-scale devices providing quasi-monochromatic synchrotron radiation which is generated by laser photons Compton-scattering off highly relativistic electrons. Since the shape and width of the X-ray spectrum are determined by the properties of the colliding beams, these must be carefully optimised. However, device compactness limits the space for diagnostics, rendering a complete characterisation challenging, especially if an electron storage ring is combined with a laser enhancement cavity. Here, a framework for laser, electron and X-ray beam parameter determination is proposed to address this issue. First, methods for determining the laser- and X-ray parameters are presented. Knowing these, electron beam parameters are retrieved from the shape of the X-ray spectrum. To this end, an analytical physical model enabling a rapid calculation of inverse Compton scattering spectra is developed and combined with a genetic algorithm. This strategy's effectiveness is demonstrated by applying the concept at the Munich Compact Light Source, a storage ring-based inverse Compton X-ray source facility. Since the analytical model is computationally very inexpensive, the proposed framework could enable real-time monitoring of inverse Compton X-ray sources or be used as a non-invasive diagnostic based on a single spectrum for the electron beam emittance of storage rings or accelerators.
arXiv:2510.12588v2 Announce Type: replace Abstract: We investigate the effective rheology of a train of elongated bubbles of negligible viscosity flowing in capillary tubes. Building upon the classical Bretherton theory for a single bubble, we extend the analysis to a train of bubbles in a single capillary tube and finally to an array of parallel, noninteracting capillary tubes, i.e., a capillary bundle. Our goal is to characterize the nonlinear pressure drop-flow rate relation of this simplified two-phase system by incorporating the thin-film hydrodynamics at small capillary numbers. We model the structural heterogeneity of the bundle by assuming that the tube radii follow a truncated power-law distribution and examine deviations of the system from the Darcy law in terms of both its statistical properties and the parameters characterizing the bubble train (i.e., the tube slenderness ratio, the volume fraction, and the number of bubbles). The main result is that two-phase flow alters the effective rheology, leading to deviations from Darcy-type behavior across the entire parameter space investigated. Specifically, for a limited number of bubbles, the flow exhibits a smooth transition from the Bretherton regime, where the pressure drop scales with the flow rate to the power of 2/3, to weaker sublinear regimes with exponents between 2/3 and unity. Interestingly, increasing the number of bubbles or narrowing the pore-size distribution leads to only minor deviations from the Bretherton regime. The resulting pressure drop-flow rate exponents are qualitatively similar to those reported in the literature for immiscible two-phase flow in porous media, despite the inherent simplicity of the capillary bundle model.
arXiv:2607.05966v1 Announce Type: new Abstract: Long-horizon failure in world models is conventionally attributed to compounding error, a generic framing that does not distinguish what kind of error compounds. We propose a kinematic-vs-dynamic reframing: world models tend to imagine kinematically rather than dynamically. We operationalize this as the imagined Kinematic-Consistency Error, a per-step diagnostic that measures how far a rollout departs from a closed-form kinematic null, paired with a perturbation protocol that tests whether iKCE responds when physical conditions cross a regime boundary. We instantiate the diagnostic on a released DreamerV3 checkpoint trained on DMC walker-walk, where imagined iKCE runs roughly two orders of magnitude above that of matched real-physics rollouts. Across a friction sweep that crosses the gait-collapse boundary, the model's iKCE stays statistically flat even as the trained policy's reward collapses through the same range, providing the kinematic-not-dynamic signature. The diagnostic distinguishes kinematic from dynamic imagination at horizons longer than the embodiment's gait period.
arXiv:2607.05971v1 Announce Type: new Abstract: We present VTMR, a two-stage framework for Video-To-Music Recommendation. In Stage~1, VTMR aligns comprehensive video and music signals in a joint audio-visual-text representation space and efficiently retrieves semantically compatible candidates using coarse global embeddings. In Stage~2, it reranks the retrieved candidates by attending to the temporal sequences of both video and music, thereby capturing fine-grained temporal correspondence. Evaluated on the video-to-music recommendation task, the multimodal retrieval stage improves R@10 from 14.2 to 15.9 and Median Rank from 75 to 58 over the strongest baseline; the temporal reranker further boosts R@10 to 18.3 and Median Rank to 46, demonstrating complementary gains from richer query encoding and temporal alignment. A human preference study confirms that VTMR is on par with a commercial baseline in overall preference, while outperforming a generative baseline in music quality.
arXiv:2607.05435v1 Announce Type: new Abstract: Academic output is produced across a fragmented toolchain: literature discovery in one application, reference management in another, writing in a LaTeX editor, formatting against venue templates by hand, and submission through yet another portal. Each boundary between tools forces a context switch, a format conversion, or a manual copy-paste step, and the cumulative cost dominates the time researchers spend on activities that are not research. We present Bibby AI, an editor-native platform that collapses this toolchain into a single Research-Write-Publish pipeline built around a cloud LaTeX editor. Unlike assistants that attach to an existing editor through a browser extension, Bibby AI owns the full document state, compilation pipeline, and revision history, which allows its agents to perform retrieval-grounded citation insertion, structural edits, and template-compliant reformatting as first-class, verifiable operations rather than text suggestions. The platform integrates (i) ingestion pipelines that convert PDF, DOCX, and handwritten mathematics into clean LaTeX; (ii) a retrieval layer over scholarly metadata enriched with patent-to-paper citation signals derived from USPTO PatentsView and the Marx-Fuegi citation corpus, surfacing the translational impact of candidate references; and (iii) task-scoped agents for literature triage, drafting, revision, and venue formatting that operate directly on the document's abstract syntax representation. Bibby AI is deployed in production and serves more than 5,000 active researchers across more than 50 subscribing universities. We describe the architecture, the design decisions that editor-nativeness makes possible, and the workflow-level time-savings framework we use to evaluate the platform against fragmented baselines.
arXiv:2607.06160v1 Announce Type: new Abstract: Synthesizing long-context supervised fine-tuning (SFT) data is a scalable way to enhance the long-context understanding of large language models (LLMs), yet existing approaches share three limitations: narrow task coverage, insufficient instruction difficulty, and a lack of faithfulness supervision. We propose \textbf{LongCrafter}, a structured synthesis framework that couples a hierarchical task taxonomy with an evidence-grounded pipeline. The taxonomy organizes long-context understanding into local/shallow and global/deep levels and yields 32 fine-grained task types that serve as a global generative prior. Guided by this taxonomy, LongCrafter constructs task-aligned long contexts, decomposes them into explicit evidence graphs that model cross-paragraph dependencies, and generates instruction--response pairs strictly grounded in the located evidence spans, ensuring both controllable difficulty and faithful, traceable reasoning. Models fine-tuned on LongCrafter data outperform all SFT baselines and even the official post-trained models on LongBench, LongBench~v2, and LooGLE across both Qwen2.5-7B and LLaMA-3.1-8B, with the largest gains on high-difficulty tasks. Further analysis shows that LongCrafter data is more diverse and better spread across difficulty levels, and that the trained models locate evidence robustly regardless of position, effectively mitigating the ``lost in the middle'' problem.
arXiv:2607.06169v1 Announce Type: new Abstract: Meeting the growing demand for quality-of-service (QoS) guarantees in 5G networks requires an accurate characterization of delay performance, commonly captured by the delay violation probability (DVP) at a specified delay target. Although hybrid automatic repeat request (HARQ) is a fundamental reliability mechanism in wireless systems and is central to supporting QoS, many existing approaches to DVP prediction for HARQ remain overly simplified. In particular, they omit important delay components and adopt assumptions that do not reflect the operation of HARQ in slot-based systems such as 5G. Consequently, these models can substantially underestimate the DVP, especially under stringent latency requirements, where the contribution of the neglected components becomes critical. To address this gap, we develop a tractable DVP characterization for 5G HARQ that accounts for queueing, transmission, decoding, and feedback delay, as well as the contribution of Control Signaling (CS) transmissions to the overall delay, under practical timing assumptions consistent with 3GPP operation. Moreover, we incorporate parallel packet transmissions that proceed without waiting for earlier packets to succeed, an essential HARQ behavior frequently overlooked in prior work. Using tools from queueing theory and Markov analysis, we then derive upper bounds on the DVP and validate them against ns-3 5G-LENA simulations.
arXiv:2607.06290v1 Announce Type: new Abstract: We study the infinite-width Gaussian-process limit of random neural networks through the lens of tensor programs, and we provide a quantitative convergence theory in Wasserstein distance. Our main result gives explicit finite-width error bounds, of order inverse square-root of the widths between finite-network executions and their Gaussian-process limits. The framework is architecture-agnostic and covers feed-forward models together with weight-sharing schemes relevant for recurrent and transformer-type architectures.
arXiv:2607.06341v1 Announce Type: new Abstract: Formal verification offers the strongest guarantee of software correctness, but it does not scale: the proofs demanded by interactive theorem provers such as Coq require enormous expert effort. Large language models (LLMs) promise to generate these proofs automatically, yet existing approaches wire a fixed, human-designed proof strategy into the system and constrain the model to follow it (retrieving premises and predicting tactics one step at a time, or splitting goals by divide-and-conquer), and still prove only a fraction of their target theorems. We show that imposing such a strategy is unnecessary and limiting. Handing the whole lemma to a general LLM code agent (for example, Claude Code), free to choose its own approach, and wrapping it in a verification harness is both simpler and more effective, achieving full coverage: every targeted lemma proved, with no failures and no Coq expert intervention. The agent writes the proofs under feedback and hard constraints from the harness that keep each one sound (accepted only when the prover's kernel closes it), complete (no obligation left unproved or silently dropped), and terminating (no divergent tactics). We evaluate this harness plus code agent along three dimensions. (1) Core logic: on Iris, the state-of-the-art separation logic for concurrent and memory-manipulating programs, Aria proves all 4,257 lemmas of the four core modules and the 217 lemmas verifying Rust's standard libraries built on it, fully automatically. (2) Comparison with prior LLM provers: on reglang, where prior provers manage barely one in eight, Aria proves all 318. (3) Generality: on iris-lean, the unfinished Lean 4 port of Iris, it proves 72 not-yet-ported lemmas, showing the approach is not specific to Coq. A state-of-the-art model (Claude Opus 4.7) can write proofs for verified software development fully and automatically.
arXiv:2607.05961v1 Announce Type: cross Abstract: Recently, Chudnovsky, Cook, Davies, Oum, and Tan obtained the first finite bound on the chromatic number of t-perfect graphs, showing that they are 199053-colorable. We improve this bound to 186 by refining their proof. The original proof establishes that every graph with large odd girth and large chromatic number contains a certain structure called an r-arithmetic rope, and that its existence in a certain leveling of a graph with large odd girth would imply an odd wheel as a t-minor, a known obstruction of t-perfectness. While their technique requires a lower bound on the chromatic number that is exponential in r, we show that the existence of an r-arithmetic rope can already be guaranteed under a linear bound. Using a slightly weakened notion of arithmetic ropes allows us to reduce the bound even further.
arXiv:2607.01148v2 Announce Type: replace Abstract: We investigate the emergence of structural disparities in networks of collaborating large language model (LLM) agents. When LLM agents autonomously choose collaborators, the resulting communication network exhibits preferential-attachment dynamics: agents that are already prominent become increasingly likely to attract additional connections. In some cases, weaker LLM agents (agents with smaller base model or older version) can disproportionately occupy central and influential network positions relative to stronger LLM agents. We interpret this as a type-dependent glass-ceiling effect (GCE). We model the network of LLM agents as a time-evolving sequence of directed weighted graphs, where the vector-valued edge weights represent cumulative tokens exchanged, number of interaction rounds, and reasoning effort. Using a contraction mapping argument on the mean-field dynamics, we prove that the importance (centrality) of each agent type converges to a unique stable equilibrium. To ground the model in LLM decision mechanisms, we introduce a cross-attention-inspired utility for collaborator selection. This utility specifies the local connection dynamics and, together with the mean-field model, yields a predictive characterization of the limiting network structure and its type-dependent centrality gaps. To validate the theory, we develop an experimental testbed with 100 LLM agents. Our experiments show that autonomous network formation can generate persistent centrality disparities, with their magnitude and direction depending on model family, model size, system-prompt design, and task context. They further show that the effect of preferential attachment depends on its alignment with model capability: reinforcing it improves collective performance when stronger agents become central, whereas weakening it improves performance when network dynamics instead favor weaker agents.
arXiv:2607.05742v1 Announce Type: new Abstract: Large language models increasingly serve as judges in AI evaluation, but current approaches rely on consensus preferences that ignore individual evaluator variation. We propose a novel simulation approach that combines categorical judgments with evaluator-specific auxiliary data--retrospective reasoning traces and interface telemetry--to enable LLM-based simulation of individual evaluators via in-context learning. We conduct a systematic empirical study of this approach using multi-facet data from 32 trained annotators across 4,200 preference judgments in a 4 x 4 x 4 factorial design. Our key findings: (1) The simulation approach achieves up to 9.9 percentage point improvements over the Base Judge; (2) Reasoning traces provide the largest gains with higher collection efforts, while interface telemetry often hurts rather than helps performance despite being cheaper to collect. (3) Simulation difficulty is systematic, predicted by an evaluator's neutral usage (most clearly on Helpfulness) and divergence from consensus; the neutral-usage tendency--rather than simulatability itself--is the cross-task-stable property (r = 0.728). These results establish both the potential and limits of evaluator-specific auxiliary data for personalized evaluation, offering methodological insights for scaling individual aware AI assessment.
arXiv:2607.05756v1 Announce Type: new Abstract: Efficient data movement between memory and compute units is a key performance bottleneck in modern FPGA designs, particularly for deep learning (DL) workloads. In typical FPGA architectures, data transfers between block RAMs (BRAMs) and digital signal processing units (DSPs) must traverse the global routing network, leading to increased wirelength, routing congestion, and critical-path delays. Prior work has explored in- and near-BRAM compute architectures to mitigate these issues, but such solutions often require fundamental changes to FPGA architecture and CAD tools, limiting their commercial viability. This paper proposes a lightweight architectural enhancement that introduces a dedicated direct connection between BRAM and DSP blocks, enabling BRAM data to be consumed by DSPs without passing through the global interconnect. We also enhance the placement algorithm to recognize these BRAM-DSP macro blocks. The proposed architectural change incurs negligible area and delay overhead and does not affect non-DL benchmarks, while the proposed CAD remains compatible with the baseline architecture, where it yields negligible change in quality-of-results (QoR). On an Agilex-10-like FPGA, the proposed architecture and CAD updates deliver up to +25% Fmax and -49% wirelength on common DL layer designs.
arXiv:2607.05483v1 Announce Type: new Abstract: Agentic workflows often operate over shared, structured state. Because LLM context windows are limited, each model invocation is typically shown only the state fragment needed for the current workflow step, a pattern commonly known as progressive disclosure. Modern systems construct such model-facing views using grep-like keyword search, retrieval-augmented generation (RAG), abstract-syntax-tree (AST) queries, and task-specific agent skills. These methods make the read side manageable, but they do not define when a locally proposed rewrite is valid after it is applied back to the full state. The missing piece is a contract between local updates and global validity. We introduce PatchOptic, an optic-inspired interface for shared-state LLM workflows. Optics are compositional bidirectional accessors that describe how views of structured data are read and updated. PatchOptic borrows this view/update intuition and realizes it through projected reads and verified structured patches. Each workflow step declares a projected read view, an authorized write region, and a patch-source region. Beyond runtime enforcement, the same declaration yields a path-level footprint that supports delegation, sub-workflow composition, and static certificates for reordering independent steps within the same phase. We evaluate this design with PatchBench, a benchmark with 46 cases across domains. The results show that projected reads reduce reported leakage and token cost while preserving accepted-output quality under the strong actor. Runtime verification blocks declared workflow-contract violations before commit, and patch-read enforcement rejects compromised patch artifacts that use hidden sources.
arXiv:2607.05552v1 Announce Type: new Abstract: Large language models (LLMs) increasingly issue judgments read as binary verdicts, and a growing literature reports such judgments shifting under logically irrelevant changes of wording - among them an amplified yes-no bias on moral dilemmas, absent in humans. A single framing cannot say what such a shift is: in a yes/no question the word "no" is at once logical verdict, lexical token, and last-printed option. We introduce a psychometric battery that separates these: crossed symmetrization - every logically irrelevant factor flipped in balanced pairs - across a corpus of question forms. A graded rating across logically equivalent forms recovers a coherent internal moral scale: frontier models' stance $\theta$ is nearly format-invariant (cross-form incoherence 0.12-0.21 on a $\pm 1$ axis); small open-weight models fail in model-specific ways. Forcing the verdict through yes/no overlays a decomposable artifact: an order bias toward the last-printed option - opposite to classic human primacy - plus a lexical pull toward the word "no"; the artifact is substantial only in the Claude models (story-averaged -0.32 to -0.86), $\approx 0$ for GPT-5.5 and Gemini, and shrinks under extended reasoning. The word and the verdict share one token; swapping the words for arbitrary labels separates them, and the verdict-attached logical bias proves $\approx 0$ for every frontier model, while model-specific label and order attachments remain: the models are not drawn toward rejecting - the pull follows the printed surface, not the verdict it carries. A minimal model, $P = \sigma((\theta \pm m)/s)$, summarizes any such artifact by a framing susceptibility m and a moral decisiveness s, measurably distinct from sampling temperature. The battery applies unchanged to any dilemma set and binary format: measuring what a model values requires crossing the frames of the question, not asking once.