arXiv:2606.26722v1 Announce Type: new Abstract: The automation of scientific discovery has reached an inflection point. While AI systems now operate instruments, optimize parameters and generate hypotheses, most remain procedural: they execute workflows fixed by human designers. True autonomous science demands epistemic autonomy--the capacity to construct, challenge and revise physical explanations in response to evidence. Here we introduce AHOIS, a multi-agent AI scientist that embeds Socratic midwifery into closed-loop experimentation. A physics-critic agent interrogates hypotheses through causal questioning, constraint checking, counterexample generation and falsification-criteria formulation. We evaluate AHOIS on a real multimode-fibre optical platform, a high-dimensional system with complex wave transformations, indirect detection, environmental drift and multi-modal acquisition. Without prior encoding schemes, classifiers or speckle models, the system autonomously proposed and validated a random-interference encoding hypothesis, discovered task-adaptive sparse-measurement strategies, diagnosed distinct failure modes (encoding instability, fluorescence contamination and detector noise) and translated a published imaging protocol into an executable workflow on a non-original configuration. The discovered encoding yielded 16x16 measurements with effective rank 56.9 and classification accuracies of 76.97% on MNIST and 83.17% on Fashion-MNIST. Ablations show that Socratic interrogation improves physical consistency, hypothesis completeness, uncertainty calibration and experimental-plan validity. These results establish a route from workflow automation towards evidence-grounded, self-correcting autonomous discovery in complex physical environments.
Science Journals
arXiv:2606.26663v1 Announce Type: new Abstract: World Action Models (WAMs) generate actions together with predicted futures, offering a powerful interface for robot decision making. In contact-rich manipulation, however, visually plausible futures can be physically incomplete: insertion, assembly, search, and reorientation often depend on slip, jamming, contact normals, or small alignment errors that are weakly visible or hidden in RGB. A natural solution is to predict future tactile states, however, we identify tactile pollution, a failure mode where unconstrained tactile-token injection degrades video and action prediction by forcing a visual dynamics model to absorb sparse, local, event-driven contact signals. To address this, we propose Tactile-WAM, a touch-aware WAM with a Tactile Asymmetric Attention Mechanism (TAAM). TAAM combines a VideoClean mask, which blocks video-query access to tactile key/value tokens while preserving action-query access, with a touch-aware bias for action attention. The VideoClean mask protects visual prediction while keeping contact information available for action generation; the touch-aware bias is derived from predicted touch changes and modulates action attention to tactile tokens during denoising. On ManiFeel, Tactile-WAM improves the mean success rate by 38.9% overall and by 86% on contact-rich tasks.
arXiv:2606.26669v1 Announce Type: new Abstract: Agents often repeatedly solve similar task instances from scratch, leading to unnecessary reasoning cost and long execution traces. Prior work has explored workflow reuse and executable skill induction, but it remains unclear which task scenarios admit procedural skills and how the shared procedural structure should be represented across successful traces. We study this problem in FSM-defined scenarios, where successful traces can be viewed as paths in an unknown transition graph, and formulate procedural skills as reusable parameterized control-flow subgraphs. Based on this view, we introduce SkillDisCo, a distillation-and-compilation framework that distills reusable PFSM subgraphs from successful traces and compiles them into callable, executable, and verifiable procedural skills. Experiments on ALFWorld and WebArena show that SkillDisCo improves success rates and reduces agent turns across benchmarks and model scales, demonstrating the benefits of representing shared experience as reusable execution structures.
arXiv:2606.26608v1 Announce Type: new Abstract: The forest matrix of a signed graph plays an important role in network science and social opinion dynamics, yet existing algorithms are mainly designed for unsigned graphs and are difficult to extend to signed graphs. In this paper, we study the problem of efficiently estimating the forest matrix of signed graphs with n nodes and introduce the signed forest matrix theorem, which establishes the relationship between generalized spanning converging forests and the forest matrix. Based on this result, we propose a novel algorithm GSCF, built on a variant of loop-erased random walks, to generate generalized spanning converging forests in expected O(n) time. We further develop two sampling algorithms, FMDE and FMDE+, for estimating the diagonal of the forest matrix, both with time complexity O(ln), where l is the number of samples. Extensive experiments on various signed graphs show that our methods achieve high estimation accuracy, significantly improve computational efficiency, and scale to graphs with over twenty million nodes. Our source code is publicly available on https://github.com/HaoxinSun98/SignedForestDiagonal.
arXiv:2509.17899v2 Announce Type: replace Abstract: We present quasi-optimal a priori error estimates for general mixed finite element methods to approximate solutions of the Stokes problem subject to inhomogeneous Dirichlet boundary conditions. For the Scott--Vogelius element this yields pressure-robust a priori error stimates. Due to the exact divergence constraint, this requires a compatibility condition for the boundary data to hold. A key tool is a modified Fortin operator, capable of preserving this compatibility condition. Furthermore, we analyse the iterated penalty method, an Uzawa-type algorithm and we show its convergence and asymptotic pressure robustness. Numerical experiments support the theory and highlight the importance of the compatibility condition and the appropriate treatment of nearly singular vertices.
arXiv:2606.26680v1 Announce Type: new Abstract: We introduce and study a family of two-head finite automata called two head returning finite automata (2-HRFA) operating on rectangular arrays of picture languages, in which both heads move in opposite directions. We show that the class of picture languages accepted by 2-HRFA is incomparable with the class of languages generated by context-free matrix grammars (CFMG), while it forms a proper subset of the class of languages accepted by returning pushdown automata (RPDA). In addition, we define a constrained variant, both head stepping two head returning finite automata (B2-HRFA), in which both heads are required to move in a synchronized, stepwise fashion. We prove that the class of languages accepted by returning finite automata (RFA) is a proper subset of the class of languages accepted by B2-HRFA, which in turn is a proper subset of the class of languages accepted by 2-HRFA. Closure properties for both the families of languages accepted by 2-HRFA and B2-HRFA are also investigated.
arXiv:2509.22259v4 Announce Type: replace Abstract: We study the extent to which rotary position encodings (RoPE), a recent transformer position encoding algorithm broadly adopted in large language models (LLMs) and vision transformers (ViTs), can be applied to graph-structured data. We find that rotating tokens depending on the spectrum of the graph Laplacian efficiently injects structural information into the attention mechanism, boosting performance in synthetic and real-world graph learning tasks. This approach, coined _Wave-Induced Rotary Encodings_ (WIRE), enjoys intriguing theoretical properties: it recovers regular RoPE on grids, and depends asymptotically on the graph effective resistance. Unlike bias-based relative position encodings, WIRE is compatible with linear attention.
arXiv:2606.26690v1 Announce Type: new Abstract: In large-scale paid acquisition and growth advertising systems, production attribution outputs are widely used for daily budget allocation and channel diagnosis. However, paid-attributed conversions such as daily new users (DNU) may systematically overstate true incremental growth when paid channels overlap with organic demand, brand-driven traffic, or other acquisition channels. This attribution-cannibalization mismatch can distort incremental ROI measurement and budget decisions at scale. We propose an experiment-calibrated attribution correction framework that uses incrementality experiments as causal anchors to convert sparse lift measurements into daily correction estimates. To make the corrected signal actionable at production granularity, we further allocate calibrated cannibalization volume across business hierarchies under structural consistency constraints. Offline forward-in-time validation against channel-level incrementality experiment readouts shows that the proposed framework substantially reduces calibration error relative to raw attribution and fine-grained ML baselines. Deployed across multiple global TikTok markets, the system supported budget and traffic strategy adjustments that were followed by an approximately 15-percentage-point reduction in the measured cannibalization rate.
arXiv:2606.26313v1 Announce Type: new Abstract: This paper presents a hierarchical control framework using model predictive control (MPC) and reinforcement learning (RL) for active roll control to manage lateral load transfer during autonomous racing of a wheeled quadruped. The framework integrates offline time-optimal raceline generation, an online MPC planner that actively minimizes the lateral Load Transfer Ratio (LTR), and a low-level, whole-body RL policy deployed directly onto the robot's 16 actuators. The MPC is based on a vehicle dynamics bicycle model of the Unitree Go2-W platform. The robot's leg actuators act as active suspension where knee joints generate anti-roll torque to bank into turns. Physical track experiments demonstrate that active roll control reduces mean LTR by up to 44%, improves the fastest lap time by 8.7%, and boosts peak lateral acceleration capability by 21.3% to 1.98 $m/s^2$, maintaining robust high-speed stability beyond the range of a non-tilting baseline controller. Supplementary code and video can be found at https://github.com/meisman-ucb/go2w-roll-control-mpc
arXiv:2606.26505v1 Announce Type: new Abstract: Modern software development increasingly involves the use of large language models (LLMs) to generate code. Despite their rapid advancement, LLMs remain prone to errors and hallucinations, emphasizing the importance of careful code inspection. However, in practice, developers' trust in LLM-generated code and their willingness to review it thoroughly may differ from these recommendations. How developers actually behave when reviewing LLM-generated code remains largely unexplored. In this study, we conduct a Wizard-of-Oz experiment to examine how software engineers behave when code is explicitly labeled as LLM-generated during a code review task. We collect both behavioral data and participant feedback through eye-tracking and exit interviews. Combining Bayesian data analysis with qualitative analysis, we found that while the thoroughness of code review did not change for participants, they spent more time fixating on LLM-labelled code, indicating that the label itself influences attention. Practitioners also adapted their review strategy for LLM-labelled code by assessing the code based on specific criteria (e.g., logical correctness), or using the prompt to guide their review. These findings inform LLM-based tool design on labelling while incorporating the prompt as a software artifact. Our study reveals a gap between reviewers' intentions and actual reviewing behaviour, highlighting the need for software companies to revisit their AI policies (particularly regarding LLM-assisted development) to better support developers in reviewing LLM-generated code.
arXiv:2606.26348v1 Announce Type: new Abstract: Multimodal large language models (MLLMs) can process diverse inputs, e.g., text, images, audio, and video, and generate textual responses. While their capabilities have advanced rapidly, evaluation of such models has not kept pace. Most existing evaluation benchmarks are limited to isolated tasks and reveal little about whether a model integrates information across modalities. We examine current means for evaluating MLLMs and review the existing benchmark taxonomy to identify gaps, including temporal-spatial coherence, physical world understanding, multimodal consistency, and selective attention. Addressing these gaps is essential for measuring real progress in multimodal intelligence and exposing capability boundaries.
arXiv:2606.26382v1 Announce Type: new Abstract: Social-physical human-robot interaction (spHRI) has grown rapidly across robotics, human-computer interaction, human-robot interaction, and haptics. Yet, fragmented terminology and inconsistent methodologies make systematic synthesis difficult. To support scalable review practices, we evaluated the extent to which small language models (SLMs; < 1.5B parameters) can assist with title and abstract screening for a large spHRI systematic review. While no SLMs matched human reviewers' performance, the models operated locally and screened papers orders of magnitude faster. The combined SLM ensemble identified 39 papers reviewers missed, representing 10.29% of the final relevant dataset. These results demonstrate that SLMs can augment, rather than replace, expert reviewers and make large-scale literature reviews accessible and sustainable.
arXiv:2606.26356v1 Announce Type: new Abstract: Practitioners of prompt-composed agentic systems report a recurring failure mode: editing one prompt module silently shifts the behavior of others despite no shared variable or executable dependency. We formalize this as compositional behavioral leakage (CBL): interference between modules sharing a context window. CBL is enabled by architectural non-isolation: transformer self-attention provides no formal boundary between concatenated modules. We probe CBL on a deployed job-evaluation agent (Claude Sonnet 4.6, 144 trials) through a reusable three-channel protocol that perturbs non-focal modules along volume, content, and form. Only the content channel produces a detectable paired effect (Cohen's d = 0.63, bootstrap 95% CI excluding zero); no recommendation flipped -- a sub-threshold regime invisible to standard QA but compounding across the thousands of decisions a deployed agent makes. CBL is orthogonal to known agent-failure axes (adversarial injection, cognitive degradation, multi-agent fault propagation, privacy leakage). We contribute an operational definition, a reusable protocol, a falsifiable prediction set, and a system-class characterization, establishing cross-module interference measurement as a requirement for prompt-composed agent evaluation.
arXiv:2510.07716v2 Announce Type: replace Abstract: We propose refined GRFs (GRFs++), a new class of Graph Random Features (GRFs) for efficient and accurate computations involving kernels defined on the nodes of a graph. GRFs++ resolve some of the long-standing limitations of regular GRFs, including difficulty modeling relationships between more distant nodes. They reduce dependence on sampling long graph random walks via a novel walk-stitching technique, concatenating several shorter walks without breaking unbiasedness. By applying these techniques, GRFs++ inherit the approximation quality provided by longer walks but with greater efficiency, trading sequential, inefficient sampling of a long walk for parallel computation of short walks and matrix-matrix multiplication. Furthermore, GRFs++ extend the simplistic GRFs walk termination mechanism (Bernoulli schemes with fixed halting probabilities) to a broader class of strategies, applying general distributions on the walks' lengths. This improves the approximation accuracy of graph kernels, without incurring extra computational cost. We provide empirical evaluations to showcase all our claims and complement our results with theoretical analysis.
arXiv:2606.26692v1 Announce Type: new Abstract: Total variation diminishing (TVD) and total variation bounded (TVB) properties are crucial for controlling spurious oscillations in numerical solutions of conservation laws. In the classical Runge--Kutta (RK) discontinuous Galerkin (DG) framework, enforcing these properties is intrinsically tied to TVD time integrators, more commonly known today as strong-stability-preserving (SSP) methods. This reliance imposes severe structural restrictions, including order barriers and incompatibility with various fully discrete DG formulations, ranging from the recent RKDG method with compact stencils (cRKDG) to the widely established Arbitrary DERivative (ADER) DG method. To bypass these constraints, we propose a novel trace-limited corrector framework that preserves the TVD/TVB-in-the-means properties using generic, non-SSP time stepping. Based on Harten's lemma, our key insight is that total variation stability is dictated solely by the cell-average update in the final corrector stage. Consequently, we modify the traces in the numerical fluxes exclusively in the final stage, leaving the intermediate predictor stages unconstrained. This strategy decouples oscillation control from the SSP restriction, accommodates standard RKDG, cRKDG, and ADER-DG predictors, and retains the compactness of the cRKDG framework. We also prove that the limiter does not activate in smooth regions, thereby preserving the underlying accuracy. Finally, numerical experiments are presented to demonstrate the capabilities and robustness of the method.
arXiv:2606.26713v1 Announce Type: new Abstract: As semiconductor technology nodes scale, computational lithography is essential for ensuring yield and performance. However, lithography is a continuous physical process involving mask optimization, optical imaging, resist exposure, and development, which existing models fail to capture. To overcome this limitation, we present LithoDreamer, the first physics-informed World Model (WM) framework for computational lithography, which formulates the ``Layout-Mask-Resist Image-After Development Image (ADI)'' pipeline as a decision-driven multi-step evolution system. LithoDreamer captures feature changes between adjacent states to model stage-specific physics-informed latent spaces, in which it controls process intervention exploration and drives subsequent state transitions. To achieve interpretable intervention optimization without continuous supervision, we propose a contrastive variational optimization paradigm that contrasts the latent differences between intervention paths with variational evolution constraints, guiding the model to generate evolutions consistent with real lithography physics. Experiments show LithoDreamer achieves state-of-the-art performance in forward evolution and inverse planning. Our lithography dataset is publicly available at GitHub (https://github.com/7jiangyq/lithodreamer.git).
arXiv:2606.26714v1 Announce Type: new Abstract: Balancing exploration and exploitation remains a central challenge in metaheuristic optimization. To address this issue, this paper proposes B\'ezier Walk Evolution (BWE), a geometry-driven optimization framework that reformulates evolutionary search as adaptive trajectory construction in the decision space. BWE integrates B\'ezier curve modeling with a distance-aware random walk mechanism to generate topology-guided search trajectories. By adaptively varying the curve order during evolution, the proposed method enables a smooth transition from diversified global exploration to refined local exploitation. Higher-order B\'ezier curves leverage multiple population-derived control points to enhance search diversity, while lower-order curves generate near-linear trajectories to improve convergence efficiency. This adaptive geometric search mechanism provides an interpretable alternative to conventional nature-inspired designs. Extensive experiments on 41 benchmark functions from the CEC2017 and CEC2022 suites, spanning dimensions from 10 to 100, show that BWE achieves strong overall performance and favorable scalability compared with 7 classical and 6 state-of-the-art optimizers, including L-SHADE and CMA-ES. Additional evaluations on five constrained engineering design problems further demonstrate the practical applicability and robustness of BWE.
arXiv:2606.26829v1 Announce Type: new Abstract: Robots operating in real-world environments must in general be able to recognize previously unseen objects. As robotic systems move toward open-world autonomy, there is a growing, yet largely unmet, need for open vocabulary object detectors that are prompt-free and efficient enough for continuous deployment. We present AnomNOVIC, a two-stage known-workspace framework that combines a masked autoencoder (MAE) trained for anomaly detection, with NOVIC, a powerful real-time prompt-free open vocabulary image classifier. The MAE produces generic object-agnostic bounding boxes, allowing NOVIC to classify salient image regions without requiring a predefined candidate class list. We evaluate AnomNOVIC against strong open vocabulary baselines in a tabletop robot-object environment featuring the NICOL humanoid robot, reaching 47.1% AP / 57.5% AP50 for prompt-free recognition, and 59.0% AP / 72.5% AP50 if class candidates are provided. Across additional datasets, including an in-the-wild test set with 48 unique objects, AnomNOVIC reaches up to 82.6% prompt-free detection and classification accuracy. These results significantly surpass all tested open vocabulary baselines, including YOLO-World-v2, OWLv2, and YOLOE.
arXiv:2606.26719v1 Announce Type: new Abstract: Unlearnable examples (UEs) protect training data by injecting imperceptible perturbations so that models fail to extract exploitable representations. In this paper, we reveal that existing UEs exhibit a critical failure once low-pass filtering is applied, indicating that the effective perturbation signals for unlearnability concentrate predominantly in high frequencies. Hence, we argue that reliable UEs should remain effective across the full spectrum. To this end, we propose Full-spectrum Unlearnable examples via Spectral Equalization (FUSE), which aims to generate spectrum-agnostic perturbations by equalizing the contributions from different bands and enforcing cross-band consistency. Specifically, FUSE adopts a Random Spectral Masking (RSM) strategy during generator training, which randomly removes a contiguous frequency band, forcing the remaining bands to maintain unlearnability. In addition, FUSE further integrates Cross-Band Guidance (CBG), which enforces mutual consistency between high- and low-frequency components, thereby further enhancing low-frequency unlearnability and regulating high-frequency perturbations to preserve the semantic fidelity of images. Extensive experiments across multiple datasets, architectures, and spectral filtering demonstrate the strong protection achieved by FUSE.
arXiv:2606.26724v1 Announce Type: new Abstract: High penetration of distributed energy resources increasingly creates congestion in low-voltage distribution networks, while local energy markets (LEMs) optimise community welfare without explicitly internalising network constraints. This paper investigates whether a profit-seeking aggregator embedded within a welfare-oriented LEM can partially internalise distribution-level congestion through market participation. We develop a post-clearing, price-protected intervention in which the aggregator injects additional supply and triggers re-clearing, with network feasibility validated using nonlinear AC power flow subject to a non-deterioration constraint on maximum line loading. The mechanism is benchmarked against Distribution System Operator (DSO)-only corrective control and a hybrid regime with residual DSO action following aggregator intervention. Results on a UK LV feeder show that aggregator participation reduces thermal loading and preserves community welfare relative to DSO-only control, though it does not fully restore compliance under severe stress. The hybrid regime achieves the strongest technical performance while maintaining lower welfare loss. Overall, aggregator intervention remains privately profitable, indicating partial incentive alignment.
arXiv:2606.26725v1 Announce Type: new Abstract: This paper introduces a computational cognitive model to investigate how information grouping impacts visual search, a key consideration in user interface design. The model uses computational rationality to view user behavior as an adaptation to cognitive and task constraints. Our work highlights that humans use hierarchical task representations, exploiting semantic and visual structures to improve search efficiency within the constraints of the visual system. We validate this model with data from two human studies focused on visual search and semantic categorization, demonstrating that semantic grouping improves search performance when it aligns with spatial grouping. Our model replicates task durations and eye movement patterns. By improving understanding of how hierarchical memory structures are utilized in human cognition, the model extends previous visual search models. We showcase our model in the rapid prototyping and evaluation of semantic visual groupings within user interface wireframes, suggesting a pathway toward applications in more complex, real-world interface design.
arXiv:2510.17538v3 Announce Type: replace Abstract: Randomized Krylov subspace methods that employ the sketch-and-solve paradigm to substantially reduce orthogonalization cost have recently shown great promise in speeding up computations for many core linear algebra tasks (e.g., solving linear systems, eigenvalue problems and matrix equations, as well as approximating the action of matrix functions on vectors) whenever a nonsymmetric matrix is involved. An important application that requires approximating the action of matrix functions on vectors is the implementation of exponential integration schemes for ordinary differential equations. In this paper, we specifically analyze randomized Krylov methods from this point of view. In particular, we use the residual of the underlying differential equation to derive a new, reliable a posteriori error estimate that can be used to monitor convergence and decide when to stop the iteration. To do so, we first develop a very general framework for Krylov ODE residuals that unifies existing results, simplifies their derivation and allows extending the concept to a wide variety of methods beyond randomized Arnoldi (e.g., rational Krylov methods, Krylov methods using a non-standard inner product, ...). In addition, we discuss certain aspects regarding the efficient implementation of sketched Krylov methods. Numerical experiments on large-scale ODE models from real-world applications illustrate the use of the sketched residual norm as stopping criterion as well as the general competitiveness of sketched Krylov methods for ODEs in comparison to other Krylov-based methods.
arXiv:2606.26748v1 Announce Type: new Abstract: We introduce and analyze several grouped variants of the method of reverse importance sampling (RIS) for estimating a partition function from samples of the Boltzmann distribution $p(x)=e^{ \betaU(x)}/Z(\beta)$. Ordinary RIS weighs each sample separately. By contrast, our proposed grouped RIS (GRIS) methods are based on assigning the samples into groups (or batches) of size $k\ge 2$ and applying a joint weight function to each group. The focal point of the research is the quest for a tractable weight function that would yield the smallest possible mean squared error (MSE). A simple identity relates the normalized MSE to the chi-squared divergence between the joint-weight distribution and the distribution of the $k$-fold sum of independent energies. Our first theoretical finding is that any weight that improves on ordinary RIS ($k=1$) must couple the group components. In other words, it must not be a product-form function across those components, as product-form weight functions always worsen the MSE. Our second, and more important, finding is that, without loss of optimality, it is sufficient to seek weight functions that depend only on the total energy, $\sum_iU(x_i)$, of the group (group-energy weight functions); for the sliding-window variants, the analogous result is open. This finding simplifies both the theoretical analysis and the application of the method substantially. For $k=2$ and $k=3$, the MSE associated with non-overlapping (NOL) groups is reduced by $20$--$65\%$ across three examples. We then propose two additional variants of GRIS, both based on sliding-window grouping (as opposed to NOL grouping). The first applies a fixed weight sliding window (FSW) across all (cyclic) shifts of the sliding window, and the second allows a variable-weight sliding window (VSW). The FSW scheme improves on the NOL one, and the VSW improves even further, as will be demonstrated numerically.
arXiv:2606.26753v1 Announce Type: new Abstract: Conversational memory retrieval optimizes relevance, yet a retrieved memory can be relevant and simultaneously outdated: a later turn updates, corrects, or supersedes it. ConvMemory v3 adds a validity context layer that detects and surfaces this update evidence through target-conditioned relation verification, sitting after the v1/v2 retrieval path. The core mechanism is a dual-evidence gate that conditions a relation judgment on the specific target proposition, scoring a (target, source) pair through the product of a MiniLM slot head and a DeBERTa-v3 slot head and gating it by conservative event/operation evidence. On a synthetic multi-hop validity benchmark the gate reaches 90.12% +/- 1.73 accuracy; through a real-data feedback loop that mines failure patterns but trains on synthetic pairs only, the verifier transfers to Memora role binding with zero target-side labels, reaching 98.8% +/- 0.9 group-all-correct. The deployed layer preserves retrieval by default: a context mode attaches structured validity metadata while keeping the candidate set and rank order fixed, and a query-conditioned demote mode is an explicit opt-in for dense current-state workloads, where it raises current-active H@1 from a never-demote baseline of 45.1% to 95.7% +/- 1.2 while protecting non-superseded memories at 99.4% recall. Six machine-verifiable safety contracts pin the layer's behavior. Multi-hop graph propagation is validated as a mechanism; fully automatic construction of strict prerequisite edges is characterized as a boundary, since strict necessity requires counterfactual world knowledge. This report extends ConvMemory v1 (arXiv:2605.28062) and v2 (arXiv:2606.10842).
arXiv:2606.26757v1 Announce Type: new Abstract: Edible insects offer an efficient source of alternative protein, requiring less land, water and emitting less greenhouse gas than conventional livestock. However, their successful integration into the food supply chain demands reliable species authentication to control allergen exposure, prevent adulteration, and meet regulatory standards. Near-infrared spectroscopy provides a rapid analytical tool, but its performance drops when applied to production batches unseen during training due to batch-to-batch variation in spectral measurements. We introduce the Batch-Invariant Spectral Network (BISN), an end-to-end framework that combines a learnable preprocessing module, initialised with Savitzky-Golay filtering, with an entropy-regularised adversarial objective to suppress batch-specific spectral variation. In contrast to Domain-Adversarial Neural Networks, which enforce domain adaptation only after feature extraction, BISN suppress batch-effects before species-specific features are learned. Using 2,700 spectra from three species (Acheta domesticus, Hermetia illucens, and Tenebrio molitor) collected across three independent production batches, BISN achieves a mean leave-one-batch-out accuracy of 0.93 (standard deviation 0.04), outperforming the strongest baseline by four percent. Further insights gained by using explainable AI confirm that model decisions consistently rely on the lipid and protein absorption regions across all folds, connecting predictive performance to known insect biochemistry. BISN addresses both cross-batch robustness and biochemical interpretability for automated insect species authentication under realistic industrial conditions. The source code and dataset are publicly available at https://github.com/majharB/bisn.