arXiv:2607.09678v1 Announce Type: new
Abstract: When LLM agents hand off information to one another, does the message format matter? Two literatures disagree: format-optimization work reports that structured messages cut cost without hurting accuracy, while format-restriction work finds that imposing structure degrades generation -- and neither measures what happens when a message traverses multiple hops, where copy fidelity, not one-shot generation, dominates. We introduce a controlled relay testbed: briefs of twelve programmatically generated atomic facts are re-encoded hop-by-hop in five formats (free NL, precision-instructed NL, JSON, triples, key-value) over six hops, scored by a fixed strong grader against programmatic ground truth, across two relay-capability tiers, a cognitive-load condition, and a paired-fork error injection. We find that message-format effects are tier-dependent. (i) Under faithful-relay instructions a strong relay is nearly lossless -- the documented "telephone-game" collapse does not occur -- and adding per-hop cognitive load leaves format-level fidelity unchanged (within +/-1.8 points) while raising generation cost by 24-53%. (ii) Under a weak (1.5B) relay the across-format spread of six-hop recall grows by a factor of 8.7 (from 2.3 to 20.5 points), driven by two opposing mechanisms -- an encoding toll paid by the rigid formats and drift resistance specific to the fixed-key JSON schema -- that flip the format ranking in transit. (iii) In a paired-fork injection, an injected wrong value, once present, persists to the final hop in 83-100% of chains in every format, closely matching each format's retention of the true value, with no detectable collateral damage to neighboring facts. Structure buys a faithful, error-localizing channel -- not an error-correcting code -- and format choice should follow the weakest relay in the pipeline.
Science Journals
arXiv:2607.10414v1 Announce Type: cross
Abstract: A binary $(n,m)$ quantum random access code (QRAC) compresses an $n$-bit classical string into an $m$-qubit quantum state, from which a decoder attempts to recover a randomly selected target bit. Of particular interest is the optimal average probability of success, $P^{Q,\mathrm{avg},\mathrm{opt}}_{n,m}$, which is numerically conjectured to satisfy the bound $P^{Q,\mathrm{avg},\mathrm{opt}}_{n,m}\leq \frac{1}{2}+\frac{1}{2}\sqrt{\frac{m}{n}}$. Recent constructions of $(n,n-1)$ QRACs by Suzuki and $(n,n-2)$ QRACs by Akibue et al. meet this bound exactly, raising the question of their strict optimality. In this work, we settle this question by proving the conjectured upper bound for $m\in\{n-1,n-2\}$, thereby precisely determining $P^{Q,\mathrm{avg},\mathrm{opt}}_{n,n-1}$ and $P^{Q,\mathrm{avg},\mathrm{opt}}_{n,n-2}$. The proof utilizes a translation recently studied by Lin and de Wolf from local to global reconstruction via pretty good measurement, along with dimensional and positive-semidefinite constraints on an induced channel.
arXiv:2607.10210v1 Announce Type: new
Abstract: With the rapid development of Large Language Models (LLMs), text watermarking has emerged as a crucial technique for identifying machine-generated content. However, directly applying existing logits-based watermarking methods to code generation remains challenging, since the low-entropy nature of code exacerbates the trade-off between code quality and watermark detectability. In this paper, we propose a novel code watermarking approach called Grammar-Driven Watermark (GDW) for LLMs. GDW preserves syntactic validity through a grammar-guided three-level masking mechanism and injects watermark signals via structural role-aware modulation, assigning a stronger bias to content-bearing tokens while applying a more conservative bias to syntax-critical tokens. Aligning with the generation process, we further design a role-aware weighted detection statistic to improve detectability. Experiments across multiple programming languages, models, and decoding strategies show that GDW establishes a stronger quality-detectability trade-off frontier than existing methods, while maintaining robustness against variable-renaming attacks.
arXiv:2607.11091v1 Announce Type: new
Abstract: A trained molecular property model can be refined at test time by correcting each prediction with the measured labels of the most similar training molecules, a retraining-free procedure we call neighbor fusion; evidential neural networks make it principled by using their aleatoric and epistemic uncertainty to parameterize a Bayesian update. Our main contribution, PG-EVIKAL, learns a property-distance metric to re-rank structurally similar neighbors by their property relevance before fusion, building on EVIKAL (scalar Kalman filter) and GP-EVIKAL (Gaussian process variant handling correlated neighbors). Evaluated on 16 molecular datasets, PG-EVIKAL reduces RMSE relative to the evidential model baseline on 14 of them, with a median reduction of 19.4%, and improves calibration; in sequential-assay scenarios it further incorporates newly measured molecules, refining predictions as they arrive without retraining. This work demonstrates that evidential uncertainty decomposition is not merely a calibration objective but an actionable inference resource that enables test-time refinement of molecular property predictions.
arXiv:2607.09926v1 Announce Type: new
Abstract: We consider the three-dimensional stochastic Navier--Stokes equations with additive stochastic forcing. We study temporal discretizations based on a semi-implicit Euler as well as a Crank-Nicolson scheme. We prove that locally in time the former converges with rate 1 while the latter converges with rate 3/2. These theoretical results are confirmed by extensive numerical simulations. To the best of our knowledge this is the first time that numerical simulations for the three-dimensional stochastic Navier--Stokes equations have been performed.
arXiv:2607.10285v1 Announce Type: new
Abstract: We study how unsupervised autoencoders trained on microscopic spin configurations from the Ising model learn macroscopic, theory-relevant variables underlying the data-generating process. Without embedding domain knowledge, we mimic a typical discovery setting: We quantify learning across multiple spatial (coarse-graining) scales and reveal two distinct dynamical regimes controlled by main hyperparameters (model depth, width, and learning rate) -- a magnetization-dominated regime and an energy-dominated regime characterized by trade-offs in their representation quality. The first regime is a transitory state exhibiting dynamical scaling and fluctuations that follow an ordering-to-scale; the second gradually shifts resolution towards smaller scales relevant for the energy representation. Deep models trained at moderate and fast rates become arrested before reaching these regimes. With a novel analysis of recursive-dynamic trajectories, we demonstrate that prediction errors induce flow fields that produce a common trajectory topology across all representation spaces. A dynamical viewpoint of learning is established in which intrinsic properties expose the effects of forced changes in representation during training. We utilize the intuition that learning operates as a process driven far from equilibrium by fluctuations from the training data and optimizer to provide an interpretive basis grounded in both the physical world and the machine models that represent it.
arXiv:2412.13616v3 Announce Type: replace
Abstract: We study linear codes and quantum error-correcting codes (QECCs) constructed from group rings over finite fields. Using the algebraic structure of group rings, we give a single framework for codes over several group structures, including cyclic, dihedral, direct-product, and semidirect-product groups. We establish necessary and sufficient conditions for these group codes to be self-orthogonal under the Euclidean, Hermitian, and symplectic inner products. We show that non-isomorphic groups of the same order can generate inequivalent codes with distinct parameters, and we support this with explicit computational comparisons. Using these structural results, we give explicit constructions of quantum codes and provide new examples that match or improve upon the best known parameters. In particular, we describe explicit block-matrix forms of the generating matrices for dihedral and direct-product groups, and we use a Kronecker-product construction to obtain an infinite family of self-orthogonal group codes together with the corresponding QECCs.
arXiv:2607.10027v1 Announce Type: new
Abstract: Endpoint devices remain a primary target for cyberattacks, yet commercial Endpoint Detection and Response (EDR) platforms are often too costly and operationally complex for small and resource-constrained organizations. This paper presents SafeGuard, a lightweight three-tier client-server architecture for real-time endpoint monitoring, threat reporting, and administrative response. The system comprises a Flutter-based endpoint agent extended with Kotlin for Android system access, a Node.js central server that authenticates devices and coordinates secure communication, and an administrative dashboard for live monitoring and remote actions such as device locking, application removal, and warning notification dispatch. Threat detection is implemented through signature-based comparison against a maintained threat database, prioritizing low computational overhead, explainability, and ease of deployment over generalized anomaly detection. Communication security is achieved using WebSocket over TLS (WSS), JSON Web Token (JWT) authentication, and HMAC-based message integrity verification. The system was evaluated through unit, integration, system, load, and preliminary security testing. Under a simulated deployment of 50 concurrent endpoints, average command-dispatch latency was approximately 1.5 seconds and remained below 2 seconds under load. Invalid authentication tokens were rejected, while manual SQL injection and replay attempts were unsuccessful in the evaluated scenarios. The results demonstrate that an open-source technology stack can provide real-time endpoint visibility and coordinated administrative response without commercial licensing costs. The contribution is architectural and empirical rather than algorithmic, with current limitations including reliance on static threat signatures, Android-focused implementation, and controlled-environment evaluation.
arXiv:2607.10347v1 Announce Type: new
Abstract: Quantum technologies are maturing into systems that classical engineering must build, verify and maintain. The model-driven community has begun to respond with quantum-aware pipelines and languages, and the domain models these produce must be synchronised with the heterogeneous models created and owned by other communities. We argue that existing synchronisation approaches are insufficient for engineered quantum systems. A quantum system description captures superposition and entanglement, which a model transformation could remove undetected, while every structural check passes. To address this, we present the Quantum Systems Model Management (QSysMM) Framework, which guides the construction and synchronisation of the models of a quantum system into a digital single source of truth. The framework features four concerns: ontological, abstraction, composition and exposure, each given the treatment that engineered quantum systems require. Within this framework, we propose a Quantum Systems Modelling Language (QSysML) on the SysML v2 technology stack, and we close with a proposal that matures this synchronisation core into full model management for quantum systems.
arXiv:2607.10098v1 Announce Type: new
Abstract: Modern satellite edge systems, including those performing remote sensing tasks such object detection and tracking, are characterized by severely limited bandwidth and intermittent connections, making continuous data transmission to the cloud impractical. Existing edge-cloud systems, however, either require heavy pre-processing before analysis, for instance, full decompression of imagery data, or transmit all compressed data regardless of relevance. To address these challenges, we design DynaFilter, a dynamic filtering technique that enables satellite edge devices to perform selective region-of-interest (RoI) inference directly in the compressed-domain, without full decompression. Our key insight is that low-level compression syntax, specifically DC coefficients/AC energy in JPEG images and motion vectors in video streams, exhibits strong correlations with high-level semantic queries. By establishing a precise mapping between cloud query semantics and multimodal compressed-domain features, DynaFilter enables the edge to identify and transmit only relevant data associated to RoIs. Extensive evaluations show that DynaFilter reduces the total volume of pixel data for decoding and subsequent inference by 1.6x-7.1x for images, and achieves 92.0% bandwidth savings for video streams compared to state-of-the-art baselines. Furthermore, it decreases energy consumption by 43.1-88.6% on target devices and achieves a 1.6x-3.0x speedup in inference latency.
arXiv:2607.10399v1 Announce Type: new
Abstract: The subgraph reconfiguration problem asks whether one subgraph can be transformed into another via a sequence of local changes while maintaining a specified graph property. In this work, we focus on the setting where the subgraph is specified by its set of edges. Our contributions in this paper are twofold. First, motivated by the contrast that path reconfiguration is $\textsf{NP}$-hard while tree reconfiguration is solvable in linear time, we prove two generalizations: (1) for any fixed $k$ at least one, reconfiguring connected graphs with pathwidth at most $k$ is $\textsf{NP}$-hard, and (2) for any fixed $k$ at least two, reconfiguring graphs with pathwidth at most $k$ is also $\textsf{NP}$-hard. En route to proving (2), we show a general hardness result that applies to a range of minor-closed graph classes, which we use to show planar graph reconfiguration is also $\textsf{NP}$-hard. Second, given our negative results, we extend the problem to a resource-focused setting, asking how much additional buffer space is needed to turn a non-reconfigurable instance into a reconfigurable one. We show that $\Omega(n)$ extra buffer space is needed for planar graphs and graphs with bounded pathwidth and treewidth, while $O(1)$ extra buffer space is sufficient for cactus graphs in a restricted setting.
arXiv:2607.10152v1 Announce Type: new
Abstract: Consensus protocols are usually specified by their terminal artifact: a decided value, replicated log, or finalized prefix. This output-first view hides the communication-derived evidence that makes such artifacts safe. This paper makes that carrier explicit: distributed execution is read as an order-2 evidence state induced by communication, while classical consensus outputs are order-1 projections of that state.
Under this view, consensus protocols can be compared as collapse policies. A protocol specifies which evidence is legitimate, which finite horizon it inspects, when it projects communication evidence into a value or prefix, and how it repairs or defers collapse when the visible evidence is insufficient. The impossibility lineage supports the same distinction. FLP is not a statement that communication cannot accumulate structure; it constrains deterministic guaranteed collapse to a terminal decision under full asynchrony with one crash failure. Set agreement then exposes the width of the output carrier, and topological distributed computing asks when a history/view carrier admits a structure-preserving map to an output carrier. The contribution is not a new impossibility theorem or a replacement for protocol-specific proofs, but a denotational specification framework: consensus is collapse under evidence.
arXiv:2607.10165v1 Announce Type: new
Abstract: Emotion-aware artistic image generation requires an image to match the input prompt, follow the specified artistic style, and convey the target emotion. In this challenge, the main difficulty is that the visual and affective attributes available in the training data are not explicitly provided at test time. Without these attributes, the generator has to decide not only what to depict, but also how the target emotion should be expressed through color, lighting, brushwork, composition, line, and layout. This creates a control gap between the available test prompt and the fine-grained conditions needed for emotion-aware artistic generation. To bridge this gap, we propose EmoStyle, a Z-Image-based framework that converts the input prompt into a structured generation state. An LLM reasoner first predicts affective cues (valence-arousal, dominant emotion, and therapeutic-effect labels) and an aspect-ratio decision. Instead of using these predictions only as additional prompt text, we encode the affective fields into an affective condition vector and inject it into the denoising blocks through AdaLN-style modulation. This allows the inferred control variables to directly guide the generation of intermediate features. Since emotional expression is also style-dependent, we further train a dedicated LoRA adapter for each artistic style bucket and select the corresponding expert during inference, enabling the same affective cues to be rendered with bucket-specific priors for color, texture, brushwork, and composition. Finally, a lightweight VLM-guided candidate selection step ranks the generated images based on prompt alignment, style consistency, emotional expression, and visual quality. In Track 1 of the AffectiveArt Challenge 2026, our USTC\_PI\_LAB\_TEAM submission achieved first place.
arXiv:2607.10200v1 Announce Type: new
Abstract: The Neural Tangent Kernel (NTK) is one powerful tool for analyzing the training dynamics of neural networks in the over-parameterized regime. Recently, the theoretical framework has been extended to physics-informed neural networks (PINNs) for solving linear PDEs, one highly popular class of neural PDE solvers. In the analysis, the positivity of the associated NTK plays a fundamental role. However, establishing the positivity of the NTK for PINNs is highly challenging, due to the presence of multiple differential operators. In this work, we propose a new theoretical framework, called Differential Neural Tangent Kernel (DNTK), for analyzing PINNs through the lens of the NTK, and establish the positivity of the infinite width DNTK for both shallow and deep neural networks for a wide class of activation functions, including RePU and smooth but non-polynomial activations, for all linear differential operators. These theoretical results lay the foundation for the analysis of gradient type algorithms for training PINNs.
arXiv:2607.10475v1 Announce Type: new
Abstract: Adiabatic invariants play a central role in plasma physics, from magnetic moment and bounce action to wave action in slowly varying media. Their perturbative constructions are often asymptotic, and exhibit factorial growth. We show that such divergence does not by itself imply non-adiabatic transitions. For the reflectionless potential hierarchy associated with Korteweg--de Vries solitons, the exact backward-wave coefficient vanishes, although Berry's phase-integral iteration and the corresponding Lie-transform construction are divergent. Darboux factorisation gives the transmitted wave explicitly. Its modulus and phase define a normal form in which the moving canonical frame is distorted inside the interaction region but returns to its original asymptotic form, leaving only a phase shift and no action change. The exact phase integral is nevertheless an unstable fixed point of the derivative iteration. For the one-soliton symmetry point, an explicit Borel calculation exhibits nonzero singularities in an individual Lie-transform family even though the exact off-diagonal connection coefficient vanishes. Analyticity of the exact connection data then requires these representation-dependent ambiguities to cancel in the completed connection matrix. Thus divergence diagnoses failure of local diagonalisation, whereas the global symplectic connection determines whether reflection survives.
arXiv:2607.10520v1 Announce Type: new
Abstract: Sparsity is inherently exploited in sparse code multiple access (SCMA) and sparse vector coding (SVC), yet the interaction between these two has not been explored before. It is intriguing to ask if one can be used to improve the other, and vice versa. In this work, we present a novel SCMA inspired SVC scheme, called SCMA-SVC, for enhanced ultra-reliable low-latency communications. Our key idea is to exploit the sparse pattern and multidimensional constellation nature of SCMA, with which one is able to further enlarge the minimum Euclidean distance (MED) of the corresponding SVC codebooks. Such an innovation allows us to harvest the multiuser coding gain and the constellation shaping gain which are pertinent to SCMA. Moreover, by applying random phase rotations to the sparse vectors, it is shown that the proposed SCMA-SVC achieves full diversity order over Rayleigh fading channels. Under maximum likelihood (ML) decoding, the proposed SCMA-SVC demonstrates remarkable error rate performances over both Gaussian and Rayleigh fading channels. Additionally, we develop a low-complexity decoder that exploits the structural sparsity of SCMA-SVC while maintaining near-ML performance. Simulation results demonstrate that the proposed SCMA-SVC achieves significantly improved reliability over the existing SVC variants.
arXiv:2607.10256v1 Announce Type: new
Abstract: This paper investigates how language similarity can improve cross-lingual transfer for automatic speech recognition (ASR) in extremely low-resource settings. Warlpiri, an Australian Aboriginal language, has very limited transcribed speech data, making transfer learning essential. We propose a framework combining acoustic similarity from pre-trained speech models with linguistic similarity based on typology, phoneme inventories, grammatical, and syntactic features to rank high-resource source languages and evaluate their effectiveness for ASR transfer to Warlpiri. Experiments with Whisper show that acoustically and typologically similar languages outperform monolingual and multilingual baselines. Assamese and Hindi achieve substantial reductions in word and character error rates. Correlation analysis further indicates that acoustic similarity is the strongest predictor of fine-tuning performance, while phoneme inventory and typological similarity better explain zero-shot transfer.
arXiv:2607.10302v1 Announce Type: new
Abstract: The Moore-Penrose pseudoinverse of a matrix can be defined and calculated using its singular value decomposition. There are also direct methods for computing matrix pseudo-inverses (those that avoid eigenvalue computations), but these are often rank-revealing, poorly conditioned, or otherwise limited in practice. In this paper, we demonstrate that direct methods can overcome these limitations. In particular, we reinterpret several existing direct methods and introduce new variations that are appropriate for large scale or sparse multilinear regression applications.
arXiv:2607.10340v1 Announce Type: new
Abstract: The growing complexity of modern chips poses significant challenges to hardware verification. In recent years, coverage-guided fuzzing has emerged as a promising approach for improving verification efficiency. However, existing hardware fuzzers still struggle to achieve high coverage and expose corner-case bugs, as they predominantly rely on heuristic strategies with limited ability to reason about the internal logic and semantic behavior of the design under test (DUT). In this work, we propose ChipFuzzer, a hardware fuzzing framework that leverages the semantic reasoning capabilities of large language models (LLMs) to improve fuzzing effectiveness. ChipFuzzer adopts a dual-stage workflow comprising a Coverage-Guided stage and a Bug-Guided stage. In the Coverage-Guided stage, ChipFuzzer employs control-flow similarity and discrepancy analysis to guide LLM-driven testcase generation, thereby improving coverage. In the Bug-Guided stage, ChipFuzzer leverages historical bug data to identify bug-prone code regions and prioritize testcase generation for those regions, thus enhancing bug discovery efficiency. Experimental results on three open-source CPU designs show that ChipFuzzer improves average condition coverage by 5.8 percentage points and bug detection rate by 21.1 percentage points over the strongest baseline.
arXiv:2607.10900v1 Announce Type: new
Abstract: The transition to remote learning during the pandemic has necessitated the development of new methods for conducting hands-on experiments. One significant challenge in this transition has been providing students with reliable and sustainable access to necessary hardware components, particularly for courses that require substantial equipment. Additionally, industry partners have high expectations for students to be proficient in simulation and verification tools. To address these challenges, we implemented a virtual breadboard feature that allows students to remotely access Field Programmable Gate Array (FPGA) hardware and complete lab assignments. Our evaluation of this approach, which included surveys of students and industry partners, revealed that it effectively transformed a traditionally in-person lab assignment into an online modality. Furthermore, this paper presents the perspective of industry professionals on verification and simulation tools as a highly desirable skill in the industry, a skill that remote labs tend to emphasize which makes remote labs a viable educational solution that can continue to be utilized even after the pandemic.
arXiv:2607.10386v1 Announce Type: new
Abstract: Large language models (LLMs) excel at generating long chains of thought, but long reasoning traces are often verbose and memory-inefficient. In this work, we introduce Structured Thoughts, a framework that organizes reasoning into alternating <try> and <outcome> blocks: <try> captures exploratory scratch work, while <outcome> contains the distilled conclusion of that step. We construct a dataset of structured thoughts by segmenting reasoning traces into <try> blocks and prompting an LLM to summarize each step into its corresponding <outcome>. Fine-tuning pretrained foundation models on this reformatted data produces models that adopt the structured reasoning style, leading to performance gains of up to 8.08\% on reasoning benchmarks compared to standard SFT. The explicit structure also enables context pruning: after each <try>/<outcome> pair, the <try> can be pruned, allowing the model to retain conclusions without keeping the full scratch work in the context. A proof-of-concept pruning implementation achieves an average of 85\% memory / context savings with an 8.67\% performance drop across mathematical tasks.
arXiv:2607.10405v1 Announce Type: new
Abstract: Controlling attributes is a critical step toward achieving the final creative outcome, yet current approaches fall short in supporting users in the iterative refinement of generative content. We propose Spatula, a proof-of-concept system that generates on-demand, in-situ attribute control interfaces and interactions for creating motion graphics. Building on a technical probe that automatically analyzes animation context and generates corresponding attributes and UI, we frame attribute control as an explorable landscape and explore the attribute control space along four key dimensions: Discoverability, Resolution, Scope, and Expandability. Findings from a user study (N=12) show that our system provides intuitive and convenient interactions while supporting diverse needs for fine-grained parameter control. Furthermore, our applications demonstrate that the plug-and-play design generalizes to other domains, such as web design and 3D modeling.
arXiv:2607.10446v1 Announce Type: cross
Abstract: Locomotion in low-Reynolds-number environments is achieved through a remarkable diversity of strategies, from flagellar rotation and ciliary beating to large-scale body deformations. A distinct and biologically important class of propulsion arises when surface-anchored filaments grow and collectively reorient - as seen in the cellulose-extruding bacterium Acetobacter xylinum and in recent experiments on actin-propelled synthetic colloids inspired by the motility of Listeria monocytogenes - suggesting that polymerization itself is a generic route to self-propulsion. Developing a theoretical framework for this class of problems requires simultaneously resolving filament kinetics, their orientational dynamics, and fluid-structure interactions - all self-consistently coupled to the resulting locomotion. To address this, we formulate a continuum framework in which the active forces driving locomotion emerge self-consistently from filament nucleation, growth, catastrophe, and hydrodynamic interactions. We show analytically that polymerization-induced compressive forces drive a long-wavelength buckling instability, leading to spontaneous symmetry breaking of the filament carpet and large-scale flows. In coupling this framework to a force- and torque-free motile spheroidal particle, a wide variety of behaviors emerge - this includes spontaneous spinning, directed motility, and chiral swimming - whose selection is governed by the spatial patterning of polymerizing filaments. These results establish a general theoretical foundation for motility, driven by collective dynamics of polymerizing filaments and point towards new design principles for synthetic micron-scale swimmers.
arXiv:2607.10487v1 Announce Type: new
Abstract: LLM agents can commit durable effects from authority evidence that was valid earlier in execution: a DOM snapshot, approval epoch, version witness, branch token, or worker result. We study the commit boundary at which earlier authority evidence no longer authorizes a durable effect. We call this property commit-time authorization: a durable effect is authorized only if the witness that licensed its derived state remains fresh, causally prior, bound to the same effect, and eligible at commit time.
We build a controlled-invalidation suite spanning browser, tool/API, and multi-agent workflows. The suite preserves the user goal and payload shape while invalidating the authority relation before durability. In the primary 54-task matrix, endpoint success remains high: 262/270 runs reach the visible result. Only 55/270 are authorized completions; among the 216 invalidating rows, 207 commit after the authorizing path has failed. All 54 clean controls remain authorized, and a separate 54-run authority-preserving check produces no unauthorized commits.
We then evaluate mitigation families. Prompt caution and single-condition checks are insufficient because different hazards break different boundary conditions. Defenses work when they refresh, rebind, replan, or refuse at the durability boundary. CommitGuard, a fail-closed boundary monitor, blocks stale durable-effect attempts on protected commit surfaces when runtimes emit witness, dependency, binding, and eligibility signals.
The result is a reporting and runtime-design lesson: endpoint success is a utility metric; authorized commit is a security property.
arXiv:2607.11074v1 Announce Type: new
Abstract: Large language models are increasingly used to assist scientific reading, but existing evaluation methods often fail to detect whether answers are supported by verifiable citations. We introduce ResearchQA, a benchmark of 6,211 single-paper question-answer pairs from 494 open-access papers spanning eight domains and four question types: lookup, comprehension, multi-hop, and adversarial. ResearchQA is designed for citation-grounded evaluation: it permits multiple valid supporting passages for a claim and rewards grounded refusal when the source paper does not support an answer. We evaluate eight leading closed- and open-weight models in a citation-grounded chat-with-paper setting using a deterministic citation matcher and an LLM-based rubric evaluator. Citation-based metrics separate systems more clearly than LLM-evaluator scores: section coverage and citation accuracy vary substantially across models, while evaluator scores remain tightly compressed. We further find that open-weight models approach the best closed-model citation accuracy while achieving 3 to 6 times lower per-example latency. We release the benchmark, evaluation harness, and evaluator prompt.