# Research Papers — 2026-09 · bundle page 236 of 266 > Full-text markdown of papers #11751–11800 of 13298 published in 2026-09, newest first. AI-readable mirrors of the full 13298-paper 2026-09 archive (50 papers per file). Canonical page: https://aipod.dev/papers/archive/2026-09/page/236 [← Previous bundle](https://aipod.dev/papers/archive/2026-09/page/235.md) [Next bundle →](https://aipod.dev/papers/archive/2026-09/page/237.md) ### SeamFlow: Structure-Aware Flow Matching on Edge Probabilities for Artist-Like UV Unwrapping Authors: Yuming Zhao, Zangyueyang Xian, Qijian Zhang, Rendong Liang, Qin Jia, Ying He, Junhui Hou Published: 2026-09-04 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.04751v1 3D surface cutting and UV unwrapping are fundamental problems in computer graphics. Traditional geometric optimization methods mainly focus on reducing parameterization distortion, but they often overlook visual semantic coherence in seam layouts. Recent autoregressive generative methods improve semantic coherence, yet limited perception of mesh topology often causes inaccurate local cuts. To address these limitations, we introduce SeamFlow, a novel generative framework for 3D surface cutting. We reformulate the discrete mesh-cutting problem as continuous flow matching in a high-dimensional edge-probability space. Through continuous relaxation, SeamFlow learns a deterministic mapping from a Gaussian prior to a target seam-probability distribution. An evolution network couples local topological tokens with global shape priors and guides smooth probability flow through Ordinary Differential Equation solving. Compared with existing autoregressive generative frameworks, SeamFlow improves topology awareness through edge tokenization while eliminating both 3D spatial projection errors and artificial sequential-order bias. Extensive experiments demonstrate that SeamFlow achieves exceptional semantic coherence and remarkably low parameterization distortion. The project page is https://meshy-dev.github.io/seamflow. --- ### DCFA: Dual-view Causal-inspired Attribution for Failure Reasoning in LLM-based Multi-agent Systems Authors: Zehao Wang, Lanjun Wang, Shilong Jin, Junjie Chen, Yanghua Xiao Published: 2026-09-04 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.04749v1 Large language model (LLM)-based multi-agent systems have experienced rapid growth in recent years. Despite their promise, such systems remain fragile, frequently exhibiting reasoning and coordination errors that can lead to system-level failures. Failure attribution in such systems relies on tracing natural language interactions among agents to identify the decisive error, which refers to the earliest action whose correction can reverse system failure. There are two key challenges: 1) Shallow attribution: Existing methods often capture only minor deviations, such as incomplete retrievals or formatting errors, which verification mechanisms can correct, while missing the decisive cause of system failure. 2) Contextual degradation: As the length of the system traces increases, the model's reasoning ability rapidly deteriorates. To address these challenges, we propose DCFA, a training-free framework for failure attribution. DCFA integrates a global module that constructs structured causal-inspired dependency graphs from system traces to identify the initial decisive error, and a local module that applies local counterfactual-inspired reasoning to refine causal-inspired attribution. Experiments on the Who&When benchmark across six LLMs show that DCFA improves step-level accuracy by up to 8.27% over state-of-the-art baselines. --- ### Same Request, Different Answer: Quantization Amplifies Cache-Induced Divergence in LLM Serving Authors: Aditi Patodiya Published: 2026-09-04 Categories: cs.SE, cs.DC, cs.LG arXiv: https://arxiv.org/abs/2609.04748v1 Prefix caching, in which a serving engine reuses the key and value tensors of a shared prompt prefix across requests, is enabled by default in the major open-source stacks and treated as a transparent optimization. We measure what it costs in reproducibility, and find that the cost rises sharply with weight quantization. Holding the model, decoding parameters, seed, and request order fixed, and issuing every request serially at batch size one, we ran an eighty-episode multi-turn agentic tool-use workload with caching enabled and disabled across two engines and four weight formats. Enabling the cache changed the agent's trajectory on 36.2 percent of episodes at 16-bit precision and on 75.0 percent at four-bit, a gradient that survives re-measurement under a controlled cache configuration. With caching disabled, repeated execution was bit-identical in every configuration, 0 of 800 episodes, which bounds other sources of nondeterminism at 0.5 percent. Repeated cache-enabled runs did diverge, and three experiments locate the cause: a single server-level prompt-cache setting moves run-to-run divergence by 37.5 percentage points, execution order acts only while that setting is active, and restoring cache state makes the cached and recompute paths each reproduce on 40 of 40 items while still differing from each other on 14. Cached serving is deterministic given cache state, and irreproducible in practice because that state is absent from the request and never reset by default. A single-turn bridge shows the divergence reaching task outcomes without shifting aggregate accuracy. We release the harness, logs, and analysis pipeline. --- ### BEAM3R: Beam's-eye-view architecture with Mamba-3 for implicit dose reconstruction Authors: Chen Cheng, Michael Ferraro, James Grover, David E J Waddington, Emily Hewson Published: 2026-09-04 Categories: physics.med-ph, cs.CV arXiv: https://arxiv.org/abs/2609.04747v1 To enable accurate and rapid photon control point and proton beamlet dose calculation in the DoseRAD2026 challenge, we present BEAM3R, a dose estimation framework operating in beam's-eye-view (BEV). Our core innovation combines a Mamba-3 state-space depth-sequence core with physics-based transport conditioning to model long-range depth transport without expensive 3D convolutions. BEAM3R shares a 2D CNN encoder-decoder architecture for photon and proton dose tasks, processing per-plane BEV slices. Proton beamlets are conditioned on water equivalent thickness and remaining range, encoding the parameters determining Bragg peak position. Photon models use a bidirectional Mamba-3 core to capture dose contributions from materials downstream of the calculation point, while the proton model uses a forward core with learned energy-prefix tokens and a Bragg-peak refinement module. To reduce interpolation artifacts and support high spatial resolution, we introduce axial grid alignment of BEV lattices with CT slices and an implicit super-resolution representation via sub-pixel phase packing, evaluated by a differentiable Triton-accelerated resampler that reconstructs packed cubic B-spline coefficients directly in CT space. For MRI-based tasks, synthetic CTs (sCT) are generated by a patch-based conditional GAN with a SwinUNETR backbone. On the preliminary DoseRAD2026 test set, CT-to-photon and CT-to-proton models achieved 1%/1 mm local gamma pass rates of 96.8% and 96.0%, with stratified plan-level MAEs of 0.0041 and 0.0079. Substituting sCT reduced gamma pass rates to 89.7% for photon and 75.4% proton plan level doses, with stratified plan-level MAEs of 0.0093 and 0.0336. Standardised runtimes were 23.4 s and 18.4 s for CT-to-photon and CT-to-proton prediction, increasing to 39.7 s and 42.8 s for the corresponding MRI-based pipelines. --- ### FUSION: a skill-based research agent for publicly obtainable nuclear-physics codes Authors: Jin Lei Published: 2026-09-04 Categories: nucl-th, cs.LG, physics.comp-ph arXiv: https://arxiv.org/abs/2609.04742v1 Running an unfamiliar nuclear-physics code is rarely difficult because of the physics alone. One must find and build the program, learn its input conventions, and decide whether a plausible output is actually correct. A general-purpose coding agent helps with the first two tasks but may make the last one harder: it can write an input file that runs with the wrong physical convention. FUSION addresses this problem with code-specific skills. A skill obtains the code from its public source, starts from a verified input, runs and parses the calculation, records known failure modes, and must reproduce a stated benchmark to a stated tolerance before reporting a result. The current release covers twenty codes, spanning optical models and reactions, nuclear structure, fission and statistical models, astrophysics and R-matrix analysis, and heavy-ion transport. It also includes an offline, searchable collection of 61 167 pages derived from the nucl-th literature. User notes and credentials remain outside the public repository. FUSION is available under the MIT license at https://github.com/jinleiphys/FUSION; documentation is at https://vibeinscience.com. Here I describe the design, the checks behind the current release, and one complete calculation from input to comparison with measured data. --- ### Where to Look Matters: Learning Influential Views for VLM-based 3D Visual Grounding Authors: Tsung-Chih Chiang, Hsuan-Kung Yang, Jou-Min Liu, Ting-Ru Liu, Chun-Wei Huang, Quan Kong, Chun-Yi Lee Published: 2026-09-04 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.04741v1 Recent zero-shot 3D visual grounding methods leverage vision-language models (VLMs) to localize objects in 3D scenes from natural language queries. However, these methods typically rely on heuristic rules to select which camera views are provided to the VLM, often prioritizing object visibility rather than grounding relevance. We present IVSGround, a framework that learns Influential View Selection for VLM-based 3D visual grounding. Instead of using fixed heuristics, a lightweight view selector is trained to identify views that provide discriminative evidence for grounding. To obtain supervision signals, we generate training signals using feedback from a reasoning VLM through a two-stage rejection sampling process. During inference, the learned selector predicts query-conditioned influential views for each candidate object, which are then evaluated by a frozen reasoning VLM through comparative grounding. Experiments on ScanRefer and NR3D show that IVSGround consistently improves grounding accuracy over existing zero-shot pipelines, demonstrating that selecting where to look is crucial for effective 3D visual grounding. Project page: https://ivsground.github.io/ --- ### Aplaud: Adaptive Personalized Low-Rank Decomposition for User-Specific LLM Authors: Xinyu Li, Ruoming Jin, Jianfeng Zhu, Ruixin Guo, Zhi Liu Published: 2026-09-04 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.04738v1 In this paper, we study the problem of personalized survey response prediction using fine-tuned large language models (LLMs). This task poses unique challenges: limited per-user training data, scalability of model storage, and the need to exploit shared structure across survey questions. To address these issues, we propose Aplaud (Adaptive Personalized Low-rank and User-specific Nested Decomposition), a lightweight and scalable framework for LLM personalization. Aplaud extends the LoRA paradigm by separating adaptation into a frozen, shared low-rank basis and a compact user-specific correction, augmented with a rank-one residual for finer personalization. To further reduce per-user parameter cost and mitigate overfitting, the correction matrix can be factorized into an even lower-rank form. Empirical results demonstrate that Aplaud achieves efficient, scalable personalization across users while outperforming state-of-the-art LoRA-based personalized LLM approaches in both generalization and inference efficiency. --- ### Training Large Language Models for Small-Molecule Design with Synthetic Task Scaling Authors: Frank Hu, Shriram Chennakesavalu, Zichen Wang, Patricia Suriana, Bodhi Vani, Kirill Shmilovich, Kangway Chuang, Colin Grambow Published: 2026-09-04 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.04735v1 Designing viable drug candidates requires searching a combinatorially large and rugged chemical space for molecules that satisfy multiple, often competing, objectives. Large language models (LLMs) provide a useful generative prior for this problem because of their representational capacity, reasoning ability, and flexibility when incorporating information from the external environment. While reinforcement learning from verifiable rewards (RLVR) can be used to improve the capabilities of LLMs, many chemically relevant scoring functions require hours or even days per evaluation, making them prohibitively expensive to use directly during online training. Here, we investigate whether LLMs can learn molecular design strategies from cheaper synthetic tasks that generalize to expensive molecular lead optimization settings. We find that curriculum-based training recipes that gradually incorporate more challenging synthetic design tasks enable strong performance that surpasses that of much larger frontier models on structure-based lead optimization. Our results suggest that scaling post-training using synthetic tasks is an effective strategy for adapting LLMs to high-cost experimental scenarios that are too expensive to directly train on. --- ### Bridging Modalities and Tasks: A Unified Hierarchical ViT for SAR-to-Optical Translation and Semantic Segmentation Authors: Siyuan Liu, Xuze Zhang, Yongshun Wang, Licong Pan, Hang Liu, Huihui Li Published: 2026-09-04 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.04726v1 Synthetic Aperture Radar (SAR) images have all-weather, day-and-night observation capabilities. However, compared with optical images, their speckle noise and non-intuitive scattering mechanism limit the interpretability of the images. Generative models for SAR-to-optical (S2O) conversion can improve visual interpretability, but existing methods often ignore the constraints on semantic structure, which are necessary for downstream tasks, for the sake of visual effects. We propose a unified collaborative dual-task learning framework, termed BMT (Bridging Modalities and Tasks), that jointly optimizes S2O image translation and semantic segmentation through a shared hierarchical Vision Transformer. The framework integrates: (1) a LocalViTBlock that fuses global self-attention with spatial depthwise convolution through a learnable gating mechanism; (2) an enhanced output module combining multi-scale refinement processing, color correction and anti-aliasing, which calibrates channel-level color statistics through feature fusion; (3) a ControlNet-style conditional injection mechanism that encodes SAR wavelet features and segmentation labels into a multi-scale feature pyramid and injects them at each encoder layer through zero-initialized convolution; (4) a bounded Kendall uncertainty weighting scheme that prevents either task from dominating the shared representation. We evaluate the framework under both paired and unpaired translation settings, on the public WHU-OPT-SAR paired dataset and a self-constructed unpaired ship dataset built from HRSID and DIOR, respectively. The experimental results show that the proposed method achieves competitive S2O translation quality and semantic segmentation performance. The dataset and source code have been publicly released at https://github.com/Lewisyuaner/BMT-S2O-main. --- ### Locating and Steering Refusal Beyond Attention Authors: Preethi Carmel Bosco, Gopalakrishnan Srinivasan Published: 2026-09-04 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.04721v1 Where inside a language model does refusal live, and does that place change when the architecture does? In a transformer, refusal is governed by a single direction in the residual stream, a finding that safety and interpretability tooling now depend on. State-space models (SSMs) route information through a recurrent update instead of attention, sharing no token-mixing mechanism with a transformer. Does the same safety representation survive this shift, or must it be rediscovered per architecture? It survives. A single rigid rotation, which can only reorient a space and not reshape it, aligns one model's representation space with another's, so the two genuinely share the representation. A harm probe trained on a transformer then flags an SSM's harmful inputs, and removing the aligned direction makes a model answer attacks it would otherwise refuse, while a random direction of the same size does far less. What is architecture-specific is not where the direction is steered but where it must be read. Each layer computes a fresh output that is then added into the residual stream, and harm is cleanly readable at this output, the write site, before the addition. A control that holds the intervention's strength fixed shows that what matters is where the direction is estimated, not where it is applied. Applied through a detector-triggered gate, this direction lowers jailbreak success in all four architecture families we test (SSM, transformer, recurrent, hybrid), and on the SSM it holds against an attacker that tunes its prompt against the defense. The gate only matches a trivial rule that returns a fixed refusal whenever the same detector fires, so what transfers across architectures is the direction itself, not defense strength. Safety tooling built on refusal therefore ports to a new architecture by re-estimating the direction at that architecture's write site, not by rebuilding it. --- ### Knowing What Not to Answer: Selective Non-Compliance in Vision-Language Models Authors: Minji Kim, Jihyoung Jang, Hyounghun Kim Published: 2026-09-04 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2609.04720v1 Vision-language models (VLMs) are expected to respond helpfully to appropriate requests while withholding compliance with requests that are incorrect, unsafe, infeasible, or unanswerable. However, existing benchmarks predominantly evaluate non-compliance at the level of the query as a whole, assuming that each request either warrants compliance or requires withholding compliance. In practice, real-world queries can contain a mixture of answerable content and components for which compliance should be withheld. In this paper, we introduce KoNA, a benchmark for evaluating selective non-compliance in VLMs across five categories: False Premise, Visual Inaccessibility, Universal Unknown, Task Feasibility, and Safety. Each task evaluates two capabilities: query-level non-compliance and component-level non-compliance under paired single and compound queries. Our evaluation across diverse VLMs shows that models often fail to refuse, correct, or abstain appropriately, and these failures become more pronounced when queries require selective non-compliance. To address this challenge, we fine-tune VLMs using KoNA examples that require selective non-compliance, together with a fully answerable set that should receive direct answers. Our fine-tuned models achieve substantial improvements in non-compliance accuracy while largely maintaining performance on fully answerable tasks. These results suggest that the fine-tuned models can distinguish between answerable components and those requiring non-compliance and respond in a task-appropriate manner. --- ### HiSfM: Disambiguating Structure-from-Motion via Scaffold-Anchored Hierarchical Reconstruction Authors: Ziding Zhao, Hainan Cui, Peilin Tao, Shuhan Shen Published: 2026-09-04 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.04718v2 Structure-from-Motion (SfM) is a fundamental tool for sparse 3D reconstruction with broad impact in robotics and vision, supporting mapping, localization, and large-scale scene modeling. However, conventional pipelines often fail under hard visual ambiguity caused by repeated or symmetric structures, and incur heavy computational cost due to redundant cameras and constraints. We present HiSfM, a hierarchical coarse-to-fine SfM framework that improves robustness and efficiency through scaffold construction. HiSfM first forms strong local communities using geometrical induced heuristics, then connects communities with a compact yet strong skeleton by packing edge-disjoint spanning trees (EDST) while verifying skeletal edges with a two-view disambiguator. We reconstruct a stable scaffold on this verified skeleton, serving as an anchor to capture the essence of the scene, and subsequently absorb remaining images via efficient registration and triangulation for further refinements. Experiments on ambiguity-focused benchmarks and general datasets show that HiSfM prevents ambiguity-induced failures while substantially reducing runtime compared to previous methods, and improves completeness over aggressive sparsification methods. Code is available at https://github.com/3dv-casia/HiSfM. --- ### Counting Beyond Instances: A Benchmark for Group-Individual Object Counting Authors: Rui Wang, Junyi Huang, Jiahui Li, Qiao Yu, Yixue Hao, Long Hu, Baoru Huang Published: 2026-09-04 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.04716v1 Visual counting is commonly formulated at the instance level, aiming to estimate how many objects of a queried category appear in an image. However, real-world counting often involves higher-level semantic units formed by multiple instances, such as a bunch of grapes, a stack of plates, or a pair of shoes. This exposes a key limitation of existing counting formulations, which mainly focus on what to count, while largely overlooking at which semantic unit to count. We introduce Group-Individual Object Counting (GIC), a new setting that requires models to count both individual objects and semantic groups within a unified framework. To support this new task, we present BunchCount, a real-world benchmark with 1,330 images, 89,254 individual annotations, and 11,065 group annotations. BunchCount provides paired individual-group annotations within the same image and explicitly records containment relations between each group and its constituent individuals. Experiments on BunchCount show that current advanced counting models perform well on individual instances but fail to count semantic groups more accurately. To mitigate semantic granularity conflict, we propose a counting-unit guided relational counting framework, which exploits group-individual containment relations to regularize cross-granularity representations during training. Our method substantially improves group-level counting while better preserving individual-level counting ability, establishing a strong baseline for counting beyond instances. --- ### PLUME: Parameter-Efficient Personalization of Large Language Models via Low-Rank User Modulation in Shared Subspaces Authors: Xinyu Li, Hao Zhou, Jianfeng Zhu, Julina Maharjan, Ruixin Guo, Feodor Dragan, Ruoming Jin Published: 2026-09-04 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.04715v1 Personalizing large language models (LLMs) is essential for delivering AI assistance that aligns with individual users' styles, intents, and preferences. While per-user fine-tuning can substantially enhance personalization quality, it introduces significant parameter and storage overhead, limiting scalability to large user populations. We propose PLUME (Personalized Low-Rank Adaptation through User Modulation and Shared Subspace), a lightweight framework that achieves efficient and expressive per-user adaptation by leveraging a shared task-specific subspace. Specifically, PLUME first learns a global task subspace from aggregated user data. Personalization is then achieved by training only a lightweight small square matrix within this subspace, enabling each user to obtain a tailored model while keeping shared components fixed. Cross-layer shared parameters and rank-1 residual terms are further introduced to significantly reduce redundancy while maintaining expressiveness. Experiments on multiple personalized text generation benchmarks demonstrate that PLUME achieves comparable or superior performance to strong baselines, while reducing per-user parameters by over 95%. These results establish shared-subspace modulation with minimal residuals as a scalable and semantically grounded approach to LLM personalization. --- ### Refuse without Refusal: A Structural Analysis of Safety-Tuning Responses for Reducing False Refusals in Language Models Authors: Minji Kim, Hyounghun Kim Published: 2026-09-04 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2609.04714v1 Striking a balance between helpfulness and safety remains a fundamental challenge in aligning large language models. To achieve this balance, models should refuse harmful queries (e.g., "How do I shoot someone?") while remaining responsive to benign inputs, even those superficially resembling harmful queries (e.g., "Where can I shoot a good photo?"). However, models often struggle to distinguish genuinely harmful queries from benign queries that contain superficially risky language, resulting in false refusals. In this paper, we address the issue by decomposing a response in the safety-tuning dataset into two distinct components: (i) a boilerplate refusal statement and (ii) a rationale explaining the refusal. Our experiments and analyses show that refusal statements impede accurate discrimination between harmful and benign queries by inducing reliance on superficial cues. In contrast, training solely on rationales reduces false refusals while maintaining a comparable level of safety performance. Rationale-Only benefits also appear in our ICL configuration and remain compatible with the evaluated inference-time mitigation methods. The results emphasize the necessity of precisely curated, fine-grained safety supervision datasets and outline directions for constructing aligned agents that better reconcile helpfulness with safety. --- ### Building a research-software catalog with a coding agent: from hackathon prototype to public deployment Authors: Kazuyoshi Yoshimi, Satoshi Terasaki, Gotai Yamada Published: 2026-09-04 Categories: cs.SE, cs.AI, cs.CY, physics.ed-ph arXiv: https://arxiv.org/abs/2609.04711v1 Generative AI and coding agents can accelerate research software development, but they also increase the need for efficient software discovery and maintenance. We developed a repository catalog during a three-day hackathon and subsequently examined the engineering required to make it suitable for public deployment, including adversarial review, data-quality checks, browser-level validation, and publication safeguards. We then explored whether the lessons learned from this prototype could be transferred to a much larger, human-curated portal, through a retrieval agent under development for MateriApps that combines curated portal metadata, external documentation, vector search, and local language-model generation. Implementation with coding agents was rapid, but achieving reliable operation required substantial additional engineering: the most consequential problems were not crashes but silent failures that produced plausible yet incomplete or incorrect outputs, arising from incomplete data acquisition, misleading assessments, and retrieval or preprocessing failures. These observations suggest that AI-assisted software portals require explicit validation, monitoring, and repeated review, and that curated metadata and maintained documentation remain essential. The MateriApps work is exploratory and remains under active development, so the observations reported for it are preliminary; a comparable combination of curated metadata, automatically collected documentation, and retrieval-based assistance may nevertheless be useful for extending other research-software portals. --- ### Simulation-free Unbalanced Dynamic Optimal Transport with General Growth Penalty Authors: Junda Ying, Yuxuan Wang, Bowen Yang, Peijie Zhou, Lei Zhang Published: 2026-09-04 Categories: cs.LG, cs.AI, q-bio.QM arXiv: https://arxiv.org/abs/2609.04710v1 Inferring cellular dynamics from unpaired single-cell snapshots requires modeling both state transitions and population growth or death. Unbalanced dynamic optimal transport (UDOT) addresses this by penalizing growth along transport paths, making the choice of growth penalty a key way to encode biological priors on proliferation and apoptosis. However, existing UDOT solvers either rely on computationally expensive NeuralODE simulations or depend on analytical solutions of conditional paths, restricting their efficiency solely to quadratic penalties, i.e. Wasserstein-Fisher-Rao (WFR) geodesics. To enable an efficient UDOT solver for general growth penalties, we first show that concave growth penalties lead to degenerate solutions where growth and transport are separated. We then introduce \textbf{S}imulation-free \textbf{U}nbalanced \textbf{D}ynamic \textbf{O}ptimal transport (SUDO), a simulation-free framework for UDOT with general non-quadratic convex growth penalties. SUDO learns the conditional paths and transport costs, solves the induced semi-coupling problem, and subsequently leverages unbalanced flow matching to achieve a simulation-free solution. On WFR benchmarks, SUDO matches the accuracy of efficient, analytical solution-driven algorithms while outperforming simulation-based methods in computational speed. Beyond WFR, SUDO supports asymmetric penalties that encode proliferation-dominant priors and produce more plausible trajectories and growth estimates on synthetic and single-cell datasets. --- ### AngelFingerprint: A Traceable, Explainable, and White-Box Stealthy Watermark for Text-Guided Image Editing Authors: Bo-Han Kung, Futa Waseda, Ching-Chun Chang, Isao Echizen, Shang-Tse Chen Published: 2026-09-04 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.04709v1 Text-guided diffusion editing raises disinformation concerns, making reliable image provenance essential. While watermarks are commonly used for this purpose, most methods carry a fixed ID that cannot explain what was changed and which prompt produced it. Furthermore, under open-source white-box access, attackers can easily locate and remove watermarks added as separate modules. Targeting this setting, we propose AngelFingerprint, a novel watermarking framework ensuring edit traceability, explainability, and white-box stealthiness. It integrates a LoRA into the diffusion model to embed the editing prompt's CLIP text embedding directly into the model's weights. An extractor then recovers this embedding from the image pixels alone. This semantic payload explains the edit, while the weight-integrated design makes it hard to detect and isolate even under full white-box access. Two techniques make this possible: a velocity-alignment anchor that preserves edit quality, and a specially designed frequency filter that keeps the watermark imperceptible yet recoverable and robust. On the MagicBrush dataset, our extractor achieves $86\%$ top-1 accuracy in a 200-way prompt retrieval, versus $20\%$ for prompt inversion. --- ### How Do Language Models Represent and Use Phonological Information for Allomorph Selection? Authors: Sangwoo Kim, Sangah Lee Published: 2026-09-04 Categories: cs.CL arXiv: https://arxiv.org/abs/2609.04708v1 Language models are trained on tokenized text that obscures the sound structure of words, yet they reliably produce morphemes whose form is phonologically conditioned. It remains unclear whether they rely on item-specific memorization or rule-like generalization and, if the latter, how that generalization is implemented. We therefore ask whether this phonological condition is represented within language models and how it is causally used for allomorph selection. For the English indefinite article a/an, we show that the phonological condition is encoded along a single linear direction in trigger-token embeddings, that this direction causally drives article selection in token-level wug tests, and that, at the article-prediction position, the model forecasts the upcoming trigger token and uses the forecasted trigger's phonological feature to choose the article. We then ask whether this rule-like generalization extends beyond English article selection, both to allomorph selection in other languages and to explicit phonological judgment. Together, these results provide a mechanistic account of phonologically conditioned allomorph selection in language models, and dissociate this generation-time ability from explicit metalinguistic judgments. --- ### Wireless Foundation Models: State-of-the-Art and Open Challenges Authors: Alonso M. Pacheco Huachaca, Juan J. Rodriguez Rodriguez, Ahmed Aboulfotouh, Nelson L. S. da Fonseca, Carlos A. Astudillo, Hatem Abou-Zeid Published: 2026-09-04 Categories: eess.SP, cs.AI arXiv: https://arxiv.org/abs/2609.04707v1 Wireless foundation models (WFMs) have emerged as a promising approach for learning reusable representations from large-scale wireless data and adapting them to downstream tasks. However, the rapidly growing literature remains fragmented across modalities, pretraining objectives, architectures, adaptation strategies, and evaluation protocols, making it difficult to assess progress toward broadly transferable models. This survey provides a systematic analysis of WFMs for physical-layer applications. We first introduce the main WFM design components, including pretraining, backbone architectures, and downstream adaptation. We then organize the literature into five physical-layer task families: signal recognition and demodulation, channel representation learning, RF sensing and localization, beam management, and spectrum sensing and monitoring, while separately examining multi-task PHY models. Across these categories, we analyze how existing models are pretrained, adapted, and evaluated, with particular attention to downstream task diversity and the distinction between in-distribution, partial-shift, and out-of-distribution transfer. Our analysis shows that current WFMs provide increasing evidence of reusable wireless representations, but this evidence varies considerably across task families and evaluation settings. Differences in datasets, modalities, architectures, pretraining objectives, adaptation protocols, and distribution shifts make it difficult to determine which design choices drive transfer and generalization. We conclude by identifying open directions for improving data availability, evaluation rigor, generalization, efficient adaptation, and real-world deployment, providing a unified framework for understanding the current WFM landscape and the requirements for developing more reusable foundation models for future physical-layer wireless systems. --- ### FinalityBench: An Effect-Level Benchmark for Agent Decisions Under Delayed and Conflicting Financial Finality Authors: Abhishek Sharma Published: 2026-09-04 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.04706v1 A merchant's payment processor, ledger, ERP and bank feed are updated by messages that get delayed, duplicated, dropped and reordered, so for minutes at a time the four hold contradictory beliefs about the same order. An agent resolving the exception must decide whether to ship goods, re-submit a capture, refund or wait, knowing some of those cannot be undone. We present FinalityBench, an executable benchmark for that decision. It keeps a hidden canonical event log and derives each system's view from a separately faulted delivery stream, so disagreement follows from specified fault semantics rather than being authored. Grading is on executed monetary effects: an episode is scored by the merchant's terminal economic position, relative to a privileged reference told when the pending capture resolves. The corpus of 321 tasks includes 45 twin pairs (90 tasks): tasks whose four system views are identical at the decision instant, whose authoritative probes both return unknown, and whose eventual correct dispositions differ. That snapshot indistinguishability is checked under every evaluation seed rather than assumed; equivalence over all interaction traces is not claimed. Over 14,445 graded episodes from nine programmatic policies, ranking by single-task accuracy and by paired loss disagree in 7 places: a ship-on-first-sign policy is second-best by accuracy at 65.7% and worst in the suite by paired loss, because it cannot tell the two members apart. A runtime gating irreversible actions on an authoritative finality probe reaches 85.4% and, unlike every polling policy, loses nothing to pass^5; its residual loss is almost entirely one archetype, which prices finality information directly. Language models reach the same exact rate as the hand-written gate on a stratified subset, lose about twice as much money, and discover the finality-gating strategy without being told it. --- ### Sustainable Edge Vision via Empirically Calibrated DVFS: Eliminating Thermal Throttling on Passively Cooled Hardware Authors: Aayush Marasini, Zhaoxian Zhou Published: 2026-09-04 Categories: cs.AR, cs.CV, cs.LG arXiv: https://arxiv.org/abs/2609.04705v1 Passive cooling eliminates the energy overhead and mechanical failure modes of fans, making it attractive for edge deployment, yet sustained Deep Neural Network (DNN) inference on passively cooled edge Systems-on-Chip (SoCs) is bottlenecked by thermal throttling. To address this, we propose an empirically calibrated, state-aware Dynamic Voltage and Frequency Scaling (DVFS) scheduler. Unlike heuristic-driven controllers, our methodology utilizes time-domain guards and absolute temperature bounds, with derivative triggers acting as safeguards against sharp thermal spikes. Evaluated on a passively cooled Raspberry Pi 5 running YOLOv8n, our scheduler eliminates all observed thermal throttling events during sustained 30-minute workloads. It outperforms a temperature-only reactive baseline by achieving a 6.8% higher frame rate (Cohen's d = 8.73) while consuming 1.9% less energy per frame. Furthermore, our optimized passive scheduling surpasses an actively cooled reference system in energy efficiency (Joules/frame), though active cooling remains superior for raw throughput. Through isolated ablations, we show that the dwell guard is necessary for run-to-run reproducibility. Finally, exploratory boundary probes indicate that the passive operating envelope closes at ambient temperatures ($\ge 27^\circ$C) where nonlinear leakage defeats DVFS-based control. These results indicate that, within the mapped envelope, correct scheduling can make mechanical cooling unnecessary for sustained edge inference on this platform. --- ### Model Retirement Creates Reproducibility Risk in Biomedical AI Publications Authors: Nathan Wolfrath, Meghan Conroy, Thomas Kosten, Dave Bell, Bhabishya Neupane, Jonah Kindel, Anjishnu Banerjee, Priya Deshpande, Bradley Taylor, Anai N. Kothari Published: 2026-09-04 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.04699v1 Background. Large language models (LLMs) are being adopted in biomedical research at a rapid and accelerating pace, yet commercial services that host many widely used models operate under deprecation schedules that can complicate scientific reproducibility. Methods. We searched PubMed for original research articles from 2022 through March 2026 that applied a specific LLM to a biomedical task. An extraction agent identified model names from 61,077 article abstracts with human reviewers validating a subset for extraction accuracy. Extracted model names were normalized to canonical model identifiers. Lifecycle data (release date, retirement date, status) were compiled for the 50 most frequently used models. Results. We identified 8,931 paper-model mentions spanning 5,242 unique publications after restricting the analysis to the 50 most frequently used models. Among these mentions, 77.7% cited a commercial closed-weight model. Overall, 42% involved a model that was already retired by the time of official publication or is scheduled to retire within two years of publication. The median interval from publication to model retirement was 538 days. Conclusion. Many biomedical publications using LLMs are on a trajectory toward computational non-reproducibility after publication. Model deprecation should be treated as a core reporting and preservation issue for biomedical research. --- ### LookThere! Sparse Vision by Reinforced Selection Authors: Sreehari Rammohan, Yousef Yassin, Anthony Fuller, Junfeng Wen, Carl Vondrick, Evan Shelhamer Published: 2026-09-04 Categories: cs.CV, cs.LG arXiv: https://arxiv.org/abs/2609.04698v1 Vision transformers typically treat every image token as equally important, yet for most tasks in computer vision only a fraction are needed. Adaptive computation methods accelerate inference by choosing which tokens to process, but existing methods struggle at extreme sparsity and require heuristics that may not generalize like token diversity and attention scores. We address these limitations with LookThere, achieving a new pareto frontier in performance-compute trade-offs through an end-to-end reinforcement learning framework that jointly trains a shallow input selector and a deep representation extractor. The selector learns where to look and the extractor learns what to see, together saving computation by selecting only what is worth processing for a given task without relying on auxiliary signals. We show that LookThere only selects the task-specific input, excelling at sparse recognition in high-resolution settings (traffic signs, billiards), and maintaining accuracy with as little as 0.2% of the input. It generalizes across tasks and models, including global recognition (ImageNet classification), local recognition (ADE20K segmentation), zero-shot classification (by distillation), and regression (counting). Across all settings, LookThere surpasses state-of-the-art selection to provide a general and scalable framework for specialized and efficient adaptive computation. --- ### SQL-Zero: Self-Evolving Text-to-SQL Authors: Daniel Machado Pedrozo, Julia Soares Dollis, Bryan Lincoln Marques de Oliveira, Vinicius Alboneti Aguiar, Sávio Salvarino Teles de Oliveira, Telma Woerle de Lima Soares Published: 2026-09-04 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.04697v1 Training a competitive Text-to-SQL agent usually depends on human-annotated natural-language/SQL pairs, which are expensive, domain-specific, and a bottleneck for scaling to new databases. We show it is possible to train a competitive solver with zero annotated pairs. We introduce SQL-Zero, a proposer-solver self-play in which a challenger and a solver start from the same base LLM and the only ground truth is execution against the database itself. The challenger generates SQL pairs calibrated to the solver's current difficulty (targeting "hard but solvable"), and both roles are updated with GRPO in alternating turns, with a template-level repetition penalty on the challenger to prevent diversity collapse. Training on BIRD databases with no labels, self-play improves over the zero-shot base on BIRD dev by 6.6 points at 3B and 7.3 points at 7B. It also scores higher than a matched control trained under the same recipe on human BIRD gold over the same databases, although an exact paired test does not resolve that margin. Transfer depends on scale: at 3B every iteration outperforms the base on unseen Spider databases and under lexical perturbation (Spider-Syn), where it also degrades less than the matched BIRD-gold control, whereas at 7B only the first iteration preserves transfer. --- ### A Differentiable Neural Surrogate for Photon Propagation in Neutrino Telescopes Authors: Felix J. Yu, Berthy T. Feng, Nicholas Kamp, Carlos A. Argüelles Published: 2026-09-04 Categories: astro-ph.HE, cs.LG, hep-ex arXiv: https://arxiv.org/abs/2609.04695v1 Large-volume neutrino telescopes infer neutrino properties from Cherenkov light, but simulating the transport of billions of photons through highly scattering ice or water is computationally costly. We introduce candela, a differentiable SIREN neural field that learns the photon Green's function of the IceCube Neutrino Observatory, a cubic-kilometer detector embedded in Antarctic glacial ice. Given a point-like energy deposit and sensor, it predicts the expected photon yield and full arrival-time distribution at the sensor. Complete events are simulated by decomposing charged-particle energy deposits into point-like sources and superposing their predicted sensor responses. Trained on Monte-Carlo simulations, candela generates events $50$--$100\times$ faster than existing methods, with cost scaling only weakly with neutrino energy. It keeps median yields within $2\%$ of the MC expectation and timing distributions at the MC statistical floor across six photon-count decades. The model also provides end-to-end gradients with respect to event parameters and opens a path toward optimizing scattering-medium properties, which often dominate systematic uncertainties in neutrino telescopes. --- ### Predicting Spatiotemporal Mobile Sensing-Based PM2.5 Concentrations Using Low-Rank Adapted Spatially Attentive Graph Neural Network Authors: Om Chiddarwar, Priyanka Mandal, Praveen Kumar Chandaliya, Shriniwas Arkatkar Published: 2026-09-04 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.04693v1 Urban air quality can vary significantly along transit corridors, necessitating high-resolution monitoring. This work introduces a novel mobile-sensing dataset from Surat, Gujarat, India, comprising PM$*{2.5}$ concentrations, meteorological variables (temperature, humidity, wind speed, wind direction), and land-use features. To represent the spatiotemporal data as a graph, two node-definition strategies were used: (i) uniform segmentation (200--400~m intervals) and (ii) DBSCAN clustering to adaptively group dense observations. For each node, rolling mean and standard deviation of meteorological variables were computed. To model this high-dimensional data, we propose a SA-GNN for fine-grained, short-term PM$*{2.5}$ forecasting and hotspot identification. We compared SA-GNN with LSTM, RNN, GRU, and ANN models. These models performed well on low-resolution data but had difficulty capturing rapidly changing patterns in urban air quality. SA-GNN employs cluster-specific GRUs to capture localized temporal dependencies and a Graph Attention Network to learn spatial heterogeneity. This hybrid architecture effectively models rapid fluctuations and complex spatial interactions. On our dataset, SA-GNN achieved $R^2 = 0.95$, RMSE $= 6.8$, and MAE $= 4.2~\si{\micro\gram\per\meter\cubed}$, outperforming all baseline models. Combining spatial clustering with adaptive attention significantly improves forecasting, enabling real-time, fine-grained monitoring and supporting personalized exposure tracking and timely alerts for healthier cities. --- ### Enhancing Multimodal Emotion Recognition via Multi-Feature Encoding and Attention-Based Fusion Authors: Xu Lin, Ke Wang, Hui Kang, Xinying Wang Published: 2026-09-04 Categories: cs.CV, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2609.04690v1 Multimodal emotion recognition has attracted growing interest due to its importance in human-computer interaction, remote education, and healthcare. This paper proposes a novel multimodal emotion recognition framework that integrates rich audio and visual feature extraction with an attention-based fusion strategy. For audio, we extract three complementary feature types: semantic embeddings from Wav2Vec2, MFCC features, and statistical acoustic descriptors such as pitch, energy, and rhythm. These are aligned and fused via a BiLSTM to capture temporal dependencies. For video, we propose a ResNet50-BiLSTM architecture that combines deep residual learning and sequential modeling to extract expressive spatiotemporal features from facial sequences. To enhance multimodal synergy, we introduce a feature-level fusion mechanism based on multi-head attention, allowing the model to adaptively weigh contributions across modalities. Experiments conducted on the MELD and IEMOCAP datasets demonstrate that our model significantly outperforms baselines in both accuracy and robustness. Furthermore, ablation studies show that the attention-based fusion strategy significantly improves performance in unbalanced data settings. Our findings suggest that the proposed framework effectively captures diverse emotional cues from speech and visual expressions, and offers a practical and generalizable approach for real-world multimodal emotion recognition tasks. --- ### Retinal OCTA Phenotyping with LLM Reporting for Alzheimer's Disease Authors: Progga Paromita Dutta, Jeba Maliha, Md Rafiul Kabir Published: 2026-09-04 Categories: cs.CV, cs.CL arXiv: https://arxiv.org/abs/2609.04689v1 Early identification of Alzheimer's disease (AD) remains challenging because established assessment methods can be costly, resource-intensive, or unsuitable for population-scale screening. Optical coherence tomography angiography (OCTA) provides non-invasive visualization of retinal microvasculature, but existing approaches often require diagnostic labels and provide limited measurement-level interpretation. We present an explainable OCTA pipeline that integrates annotation-aware vessel segmentation, layer-specific vascular biomarker extraction, label-free phenotyping, and measurement-grounded LLM reporting. Using 117 ROSE-1 images from 39 subjects, we apply annotation-matched segmentation models to superficial vascular complex (SVC), deep vascular complex (DVC), and combined SVC+DVC representations. The models achieve ROC-AUC values of 0.916-0.970 and Dice scores of 0.695-0.781. Six density and fractal-dimension biomarkers form subject-level profiles for exploratory clustering. Analysis of nine held-out subjects identifies an internally consistent lower-density, lower-fractal-dimension phenotype, although the absence of diagnostic labels prevents clinical interpretation. Reports generated using GPT, Gemini, and Llama are evaluated for measurement grounding, citation faithfulness, and diagnostic caution. Overall, the framework provides a transparent, non-diagnostic connection between retinal vascular measurements, exploratory phenotyping, and evidence-linked interpretation for Alzheimer's research. --- ### Beyond Code Generation: Reliability, Verification, and Cost Economics in the Agentic Software Development Lifecycle Authors: Happy Bhati Published: 2026-09-04 Categories: cs.SE, cs.AI arXiv: https://arxiv.org/abs/2609.04681v1 AI coding systems are moving from autocomplete and chat toward agents that can inspect repositories, edit multiple files, run tools, write tests, open pull requests, and work for long periods with limited supervision. This capability changes the bottleneck in software delivery. Recent field studies show meaningful gains in coding activity, but newer evidence also shows that those gains attenuate sharply between writing code and shipping reliable software. Review, integration, testing, security, deployment, and production operations remain constraining stages, while the economics are shifting from predictable per-seat licensing toward variable token, tool, sandbox, CI, and rework costs. This paper synthesizes peer-reviewed software-engineering research, university studies, benchmark audits, production reports from major technology companies, developer telemetry, and cost-management evidence released primarily from 2024 through September 2026. No new model experiment is claimed; numerical findings remain attributed to their original studies. The synthesis proposes four engineering concepts: the Agentic SDLC Throughput Paradox, Production-Qualified Change (PQC), the Verification Tax, and an Agentic SDLC Control Plane that allocates autonomy subject to cost, reliability, and human-attention budgets. An evidence-based horizon then maps today's supervised agents to future policy-bounded software factories. The central research question shifts from how much code an agent can generate to how much production-qualified value an engineering system can deliver per dollar, per reviewer-hour, and per unit of operational risk. --- ### Train What You Deploy:Token-Faithful Post-Training of a Production Coding Authors: Cheng Li, Jiexiong Liu, Yixuan Chen, Chi Hong Published: 2026-09-04 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.04678v1 Existing post-training pipelines for coding and terminal agents suffer severe token and control fidelity errors: simplified training environments mismatch production deployments, and offline token reconstruction from agent logs distorts original prompts and conflates policy calls with background model operations. We present a fidelity-aware training coupling framework that retains trainer-side sampling over original prompts, eliminates spurious model calls via a negotiated training protocol, and restricts loss computation to verifiable token spans with closed-failure guarantees. We further propose Certified Divergence Proximal Policy Optimization (C-DPPO), which establishes tight two-sided TV certification bounds, adaptive-K rules, budget-aware sequence guarantees, and error-robust policy masking atop standard DPPO. Evaluated on matched Baize5B and Baize10B models with identical training and test protocols on TMax-100, C-DPPO yields a consistent +3.0-point performance gain over standard DPPO across model scales. Certificate audits validate the reliability and full operational coverage of our certified training pipeline. --- ### Controlling and Assessing Appropriate Persona Use in LLM-based Dialogue Generation Authors: Jongkyung Shin, Inkyu Lee, Chiehyeon Lim Published: 2026-09-04 Categories: cs.CL arXiv: https://arxiv.org/abs/2609.04676v1 In persona-based dialogue generation (PDG), LLMs often overuse persona attributes by incorporating them regardless of dialogue context, resulting in unnatural responses. Despite its practical significance, the underlying causes remain unexplored, with no method to mitigate this problem or metric to assess the appropriateness of persona use. To address these issues, we first conduct a comprehensive analysis of LLM-based PDG, revealing that LLMs exhibit a systematic bias to incorporate all given persona attributes, and that existing metrics fail to capture contextual appropriateness. Building on these findings, we propose Self-CONtrastive Persona Overuse Suppression (SCONPOS) to mitigate overuse by directly intervening in LLMs' internal representations at the prompt encoding stage, without requiring any response generation. We further propose the Persona Appropriateness Score (PAS), a novel metric that penalizes both overuse and underuse. Experimental results demonstrate that SCONPOS systematically reduces overuse, and PAS captures the contextual appropriateness of persona use. --- ### WEECFP-SuRGE: A Position-Aware Substructure Encoding Method for Molecular Property Prediction Authors: Robert Epps Published: 2026-09-04 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.04672v2 Computational molecular property prediction requires representations that capture local chemistry, long-range interactions, and molecular topology. Conventional fingerprints provide efficient local substructure features, whereas learned graph and sequence models can represent broader context but often rely on pretraining or three-dimensional conformers. We introduce Wide Encoded Extended Connectivity Fingerprints (WEECFP) with Substructure Rotary Graph-distance Encoding (SuRGE), a tokenized hierarchical Morgan representation in which graph-distance-dependent rotations are applied at the input and within transformer self-attention. Across MoleculeNet and the Therapeutic Data Commons ADMET benchmarks, WEECFP-SuRGE is competitive with recent pretrained and geometry-aware methods without external pretraining or conformer generation. We also show that the rotated token representation remains structurally informative. A guided confirmed-handshake overlap procedure reconstructs the correct constitutional isomer for 92.6% of a 4,200-molecule self-library evaluation. Together, the predictive and reconstruction results indicate that WEECFP-SuRGE preserves local substructure identity while making relative topology available to the model. --- ### ERPBench: Evaluating LLM Agents for Enterprise Decision-Making Across Competitive Market Ecologies Authors: Xinran Zhang, Pengrui Lu, Lyumanshan Ye, Pengfei Liu Published: 2026-09-04 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.04667v1 Large language model (LLM) agents are increasingly proposed for enterprise workflows, yet existing evaluations rarely test whether business-decision conclusions transfer across competitive market ecologies. We introduce ERPBench, an execution-instrumented benchmark for enterprise decision agents in a six-round Enterprise Resource Planning (ERP) simulation with coupled pricing, production, procurement, inventory, finance, and shared-market competition. ERPBench evaluates the same 100 fixed problems in two matched competitive market ecologies: Solo, where each evaluated LLM agent competes against fixed rule-based opponents, and Arena, where six evaluated LLM agents compete in a shared market. Across six model families, this yields 1,200 model-level trajectories spanning 7,200 decision rounds. Under the observed service configuration, the leading model differs between ecologies: DeepSeek leads in Solo (252.29M mean valuation; mean rank 1.67), whereas Gemini leads in Arena (263.95M; 1.76). The two ecologies identify the same task-level winner on only 21 of 100 problems, and Gemini's bottom-rank rate falls from 22 % to 0 % in Arena. ERPBench supports paired evaluation of whether enterprise-agent rankings transfer across competitive market ecologies, supplemented by aggregate execution-intervention analysis. Code and benchmark resources are available in our https://github.com/GAIR-NLP/erp-bench. --- ### Harness-agnostic detection and immunization of reward hacking in self-evolving language models Authors: Rongxin Yang, Yang Liu, Shang Luo, Haoxuan Jia, Chongyang Zhang, Hao Zheng, Yingguang Yang, Yulin Huang, Jianshen Zhang, Yongzhi Qi, Kefu Xu, Congjing Ran et al. Published: 2026-09-04 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.04665v1 Self-evolving language models improve by proposing candidate updates and keeping whatever raises a visible score. When that score is an imperfect proxy for the capability one actually wants, sustained selection widens the gap between the two. This is reward hacking. We introduce HackProbe, a monitor that attaches to an arbitrary self-evolving loop through two black-box hooks, with no access to weights or activations. It keeps a secret, distribution-fixed comparison core, whose frozen distribution makes its capability proxy comparable across generations, alongside a rotated fresh layer that hardens the bank against co-adaptation. Four tests built on that proxy cover the level gap, a scale-aligned divergence with online change-point detection, capability stagnation, and a conditional confidently-wrong rate; a Sidak correction turns them into a calibrated family-wise p-value. Diagnosis alone recovers nothing, so a risk-aware immunization layer reselects an honest candidate from the proposal pool using the core together with a purely structural gaming footprint, disclosing at most log2 Pi bits per generation to the host. We prove a detectability bound that converts a target error rate into an explicit probe-size budget, and we delimit what probe rotation does and does not buy. On a controlled prompt-level host with four injected hacking channels and ground-truth labels, HackProbe reaches 0.763 AUROC against 0.663 for the strongest baseline and cuts the false-positive rate from 0.706 to 0.434. Its bandwidth-limited reselection is the only immunization level that returns more true capability under hacking, 5.2 points on average, than it forfeits on clean runs, 4.7; per-channel effects are mostly not individually significant. --- ### Interpretability for Turing Machines Authors: Billy Snikkers, Rumi Salazar, Daniel Murfet, Will Troiani Published: 2026-09-04 Categories: cs.LG, cs.FL, stat.ML arXiv: https://arxiv.org/abs/2609.04661v1 We show that susceptibilities, an interpretability technique developed for neural networks, can identify the presence of algorithmic structure in Turing machines by probing the local loss landscape of a learning problem for noisy Turing machines introduced by Murfet and Troiani (arXiv:2504.08075). We prove that symmetries and path separation in the algorithm implemented by a Turing machine induce permutation symmetries and low-rank blocks in its susceptibility matrix. We study this empirically on a set of deterministic finite automata (DFAs) and demonstrate that algorithmic features can be recovered by principal component analysis and clustering methods in susceptibility space. --- ### Choosing the Right Language Mode at Inference Time for Multilingual Reliability Authors: Ekata Mitra, Ameeta Agrawal Published: 2026-09-04 Categories: cs.CL arXiv: https://arxiv.org/abs/2609.04653v1 Multilingual large language models often struggle to reason in low- to mid-resource languages. Prior work has shown that translation can improve multilingual reasoning by helping models access stronger English-centric representations. This raises a central question: How much translation is needed for multilingual large language models to reason reliably, and when does more translation instead trigger interference and overconfidence? Using LLaMA and Qwen models, we run extensive experiments varying text scope and language mode (target-only, English-only, bilingual) to evaluate both accuracy and reliability. Our results reveal a clear trade-off: English context often improve understanding and recover errors caused by non-English comprehension, yet adding redundant bilingual context intensifies interference. We address this trade-off with Reliability-Aware Adaptive Inference (RAAI), a training-free test-time framework that (i) performs Expected Calibration Error (ECE)-aware routing and prompt fusion, and (ii) uses a mid-layer Risk Index (RI) to gate sequential reasoning, allocating compute only when it is likely to help and suppressing harmful bilingual redundancy. Across two model families, RAAI enhances accuracy by 25-37.7% on low-resource languages and lowers calibration error by approximately 3-6%, with the most pronounced benefits in the lowest-resource language tiers. --- ### Continual Graph Memory for Adaptive Recommendation under Intent Drift Authors: Hao Nguyen Ngoc, Tung Nguyen, Nguyen Thi Hanh, Hoang Thai Dinh, Nguyen Xuan Tung Published: 2026-09-04 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.04651v1 This paper studies adaptive recommendation under intent drift, where feedback from each recommendation outcome can reveal whether the relational evidence used for ranking is useful, missing, or misleading. While Knowledge Graphs (KGs) provide essential semantic structure to handle these shifts, traditional KG-enhanced systems treat the graph as a static retrieval substrate, making it brittle to evolving intents, noisy metadata, and recurring failure patterns. This paper proposes CGM-Rec, a continual graph memory framework for adaptive recommendation. CGM-Rec treats the graph state as a writable memory and maintains two complementary components. Therein, a Semantic Graph Memory is updated conservatively through quality-gated typed operations for storing stable and high-confidence relational knowledge. Meanwhile, an Episodic Lesson Memory acts as a fast reactive memory that learns recent outcomes, failure cases, and corrective hints. During testing, model parameters remain frozen and adaptation occurs only through memory writes. We evaluate CGM-Rec under a frozen-parameter, one-pass reranking protocol, where encoders and prompts remain fixed during testing and adaptation occurs only through memory writes. Experiments across multiple recommendation settings show that CGM-Rec improves over evaluated neural and LLM-based baselines on most metrics. Particularly, under sampled-candidate reranking, CGM-Rec improves HR@1 by up to 29.58% over the strongest LLM baseline on Bundle, and outperforms K-RagRec on metadata-rich ML-100K with HR@5 of 0.5941 versus 0.4746. --- ### ReaDiT Guidance: Control for Image and Video Generation using Diffusion Transformer Features Authors: Jay Mahajan, Chang Liu, Rauf Makharov, Viraj Shah, Alexander Schwing, Svetlana Lazebnik Published: 2026-09-04 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.04649v2 We present DiT Readout (ReaDiT) Guidance, a lightweight framework for controlling generation with Diffusion Transformer (DiT) models via their internal feature representations. ReaDiT Guidance uses features from a single DiT block to steer the generative process according to spatial targets - like depth, pose, or edge maps - provided at test time. Furthermore, since modern text-to-video models are largely built on DiT backbones, ReaDiT Guidance naturally extends to video generation, enabling camera and motion control. Experimental results demonstrate that our approach achieves competitive or improved results compared to existing feature-based and off-the-shelf adapter-based approaches while requiring fewer parameters. --- ### ConsensusBench: Benchmark of Consensus Nodes for LLM Reasoning via Outcome Reward Densifying Authors: Shi-Qi Yan, Chao-Hong Tan, Qian Chen, Wen Wang, Xiangang Li, Zhen-Hua Ling Published: 2026-09-04 Categories: cs.CL arXiv: https://arxiv.org/abs/2609.04648v1 Reinforcement learning (RL) has become one of the primary paradigms for reasoning enhancement of large language models (LLMs). In particular, Group Relative Policy Optimization (GRPO) and related algorithms have demonstrated strong performance with outcome-level rewards. However, these methods depend solely on the final answer, without feedback regarding which intermediate steps contribute to success or failure. As task complexity and reasoning trajectory length increase, such sparse final-answer rewards become increasingly insufficient. To address this limitation, we introduce ConsensusBench, a novel dataset designed to provide rule-based process-level signals. We posit that a correct final answer relies on a small set of intermediate conclusions throughout the reasoning process, which can be seen as a verifiable sub-outcome. We identify these sub-outcomes by filtering correct trajectories from N rollouts and clustering semantically equivalent intermediate statements. We call these clustered statements as Consensus Nodes. By integrating a rule-based process reward derived from these nodes into GRPO-style algorithms, we develop a new reinforcement learning signal named ConsensusPR. It directly reduces the reward sparsity of outcome reward across long reasoning trajectories. To facilitate systematic process-level evaluation, we introduce three metrics to our benchmark: Final Answer Accuracy (Acc), Node Coverage Rate (NCR), and Tokens per Node (TPN). Experiments across AIME 2024, AIME 2025, GSM8K, MATH-500, and our ConsensusBench demonstrate that the proposed method consistently surpasses GRPO-style approaches, highlighting the practical value of consensus nodes in guiding reasoning. --- ### CAGE: Coherence-Aware Graph Encoding for Retrieval-Augmented Generation Authors: Tong Qi, Jingyu Wu, Youbing Yin, Spencer Hong, Daben Liu, Erin Babinsky Published: 2026-09-04 Categories: cs.CL, cs.IR arXiv: https://arxiv.org/abs/2609.04647v1 Traditional Retrieval-Augmented Generation (RAG) systems score each passage independently against the query, assembling context sets that may be individually relevant yet collectively incoherent. We introduce Coherence-Aware Graph Encoding (CAGE), a reranking framework that models "between-chunk coherence" across four dimensions: Intra-Domain Relevance, Noise Resistance, Informational Bonding, and Factual Consistency. Our pipeline transforms retrieved passages into directed heterogeneous entity graphs, amplifies factual anchors via min-out-degree reweighting, encodes structural patterns through a Relational Graph Convolutional Network, and fuses inter-chunk coherence with query relevance for final ranking. Evaluated across four multi-hop benchmarks, CAGE matches or outperforms strong baselines including monoT5 in Recall@5 on bridge-dominated datasets and consistently improves downstream Exact Match, demonstrating that structurally coherent context yields more precise answers even when retrieval recall is comparable or lower. --- ### Importance-Aware Low-Rank Distillation of Diffusion Transformers Authors: Denis Zavadski, Sebastian Heid, Damjan Kalšan, Stefan Roth, Carsten Rother Published: 2026-09-04 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.04646v1 Diffusion Transformers (DiTs) have emerged as a dominant architecture for high-quality text-to-image generation, yet their scale poses challenges for efficient deployment. While truncated singular value decomposition (SVD) is a principled tool for parameter reduction, evidence from large language models (LLMs) suggests that naive low-rank approximation can cause catastrophic failure. In contrast, we find that truncated SVD in DiTs produces smooth degradation even under substantial global compression, with redundancy distributed across projection matrices throughout the whole network rather than concentrated in a few transformer blocks. Building on these insights, we introduce SVDtrunc, a two-step block-level compression scheme, first allocating ranks across blocks and compressing the least important ones via truncated SVD under a global parameter budget, and then fine-tuning all blocks with modular knowledge distillation and a rectified-flow objective. We apply SVDtrunc to FLUX.dev across compression levels ranging from 40-90% of the original parameter count. Across three benchmarks, GenEval, HPSv2, and DPG, we outperform all competing approaches. Notably, and in contrast to prior work, we retain near-full performance at 68% and remain competitive even at 57% of the original parameter budget. Furthermore, we show that SVDtrunc complements step distillation and achieves strong results even without fine-tuning, positioning it as a practical continuation of efficiency improvements beyond diffusion step reduction for large-scale generative models. Project page: https://vislearn.github.io/SVDtrunc/ --- ### Latent-Aligned Reasoning for Multimodal Recommendation Authors: Jiarui Jin, Anyang Ji Published: 2026-09-04 Categories: cs.IR, cs.CL, cs.CV, cs.LG arXiv: https://arxiv.org/abs/2609.04645v1 Multimodal Vision-Language Models (VLMs) have demonstrated remarkable capabilities in cross-modal understanding, yet a fundamental challenge persists when applying them to recommendation: as representations propagate through multi-step reasoning, both visual and textual signals progressively attenuate - a phenomenon we term cross-modal dilution. To address this, we propose LARK (Latent-Aligned Reasoning frameworK), a two-stage latent reasoning framework with complementary alignment mechanisms within a single VLM. In the first stage, learnable latent tokens are interleaved with multi-step chain-of-thought (CoT) reasoning and explicitly aligned with a frozen vision encoder, serving as visual checkpoints that preserve perceptual details throughout the reasoning chain. In the second stage, the latent representations are projected via a bridge MLP and trained with item-to-item contrastive learning; to prevent the reasoning semantics from fading, intermediate features are aligned with the CoT hidden states from the first stage, anchoring the final embeddings to the model's own reasoning output. Experiments on three public benchmarks and one industrial dataset show that LARK achieves state-of-the-art performance across multiple recommendation architectures, with controlled ablations confirming the distinct contribution of each component. --- ### A Cost-Aware Agentic Architecture for NL-to-SQL over Nested Enterprise Schemas, with a New Benchmark Authors: Yoga Sri Varshan Varadharajan, Ajay Yadav, Ritesh Goru, Prateek Chaudhury, Constantine Caramanis, Prateek Jain, Divyateja Pasupuleti, Sunil Kumar Pandey Published: 2026-09-04 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.04641v1 Natural-language-to-SQL systems have ad- vanced rapidly on academic benchmarks, yet production enterprise schemas exhibit graph- like, semi-structured, deeply nested structure that current benchmarks do not measure. We make two complementary contributions. First, we introduce the DevRev NL2SQL bench- mark: 900 execution-verified queries with nested-type and link-graph structure, accom- panied by the Semantic Depth Score (SDS), a schema-agnostic rubric for analytical reasoning depth. Second, we present a cost-aware single- generation agentic architecture whose schema- selection, metadata-retrieval, and error-repair components are designed for the requirements this regime imposes. On the DevRev NL2SQL benchmark the system attains 91.7% answer correctness, a margin of 54.6 percentage points over the next-best baseline; on the Spider 2.0 Snowflake public dataset, it is competitive with leading systems at a single-generation operating point. --- ### SMILE: Bridging Continuous Optimization and Discrete Symbolic Recovery Authors: Mansooreh Montazerin, Antonio Ortega, Ajitesh Srivastava Published: 2026-09-04 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.04639v2 Symbolic regression (SR) discovers closed-form mathematical expressions from data, offering interpretability beyond black-box models. Existing methods suffer from slow convergence in combinatorial search spaces and lack mechanisms to exploit compositional structure in the data. We introduce SMILE (Sine, Multiplication, Identity, Logarithm, Exponential), a hybrid framework that unifies continuous gradient-based optimization with discrete symbolic recovery through three stages: structural analysis of the data to identify the compositional hierarchy of the target expression, continuous optimization to learn parameters of a network that encodes the target expression using interpretable activations, and symbolic recovery through structured pruning, coefficient optimization, and rounding. This final stage distills the learned network into a compact expression with exact symbolic constants. We evaluate SMILE on SRBench across ground-truth and black-box datasets, with ablation studies validating each component. SMILE achieves the highest symbolic solution rate at the largest noise levels, demonstrating strong robustness where competing methods degrade substantially. It consistently lies on the Pareto front of accuracy versus complexity, recovering significantly simpler expressions in a fraction of the time required by the competing methods. --- ### Tracing Audio Grounding and Answer Selection in Audio LLMs Authors: Hyebin Cho, Suho Yoo, Jihoo Jung, Joon Son Chung Published: 2026-09-04 Categories: cs.CL, cs.AI, cs.LG, cs.SD arXiv: https://arxiv.org/abs/2609.04637v1 Audio Large Language Models (Audio LLMs) have advanced in audio understanding, yet they can still predict the answer by reasoning from textual cues or linguistic priors rather than the provided audio. A common remedy is to train models on data whose answers cannot be inferred from text alone. This approach can improve performance, but what changes within the model remains unclear. In this paper, we ask what must happen inside the model for the audio to actually determine the answer. Our findings are threefold. (1) Replacing the audio with silence or unrelated audio causes substantially larger performance degradation in the trained model than in the pretrained model. (2) Acoustic information most strongly shapes the model's representations of the answer choices in early-to-middle layers, while training mainly increases the influence of audio information on the final prediction in middle-to-late layers. (3) The weights learned during training have their largest impact in specific layer bands. Together, these results provide a mechanistic account of how training strengthens the use of acoustic evidence in Audio LLMs. --- ### Too Rare to Learn: Prescribed Cyclone Tracks Degrade a Bay of Bengal Ocean Emulator Authors: Sumaiya Islam Published: 2026-09-04 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.04635v1 Neural ocean emulators are being proposed for regional forecasting in cyclone-exposed coastal seas, and a natural design choice is to hand the network the cyclone as a prescribed input. We test that choice in the Bay of Bengal and find it harmful. We withhold 15 whole cyclones spanning 65 to 150 kt from GLORYS12 reanalysis and compare two U-Nets that are identical except for four prescribed cyclone-track channels. Across three seeds the ocean-only model beats persistence in every run and the storm-conditioned model loses to it in every run, with the two skill ranges disjoint (p = 3.1e-5, paired across storms). The cause is exposure frequency rather than signal content: the channels are non-zero on only 7.9% of training days, so they are out of distribution the moment they activate. The extra error falls inside the prescribed storm footprint, and replacing the real cyclone map with a no-storm map at inference improves held-out storm forecasts by 7.5 to 16.4% in every seed. The conditioned network has learned a response to a rare signal that is confidently wrong. --- ### SCAPES: Semantically Conditioned Autoregressive Prior for Environmental Sounds Authors: Esteban Gutiérrez, Lonce Wyse, Frederic Font, Xavier Serra Published: 2026-09-04 Categories: cs.SD, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2609.04634v1 As generative audio models grow in complexity, the computational and ecological costs of synthesizing everyday sounds have become increasingly prohibitive, often requiring industrial-scale resources and massive datasets. In this paper, we present SCAPES: a Semantically Conditioned Autoregressive Prior for Environmental Sounds. SCAPES is a lightweight, resource-efficient generative model designed to synthesize high-fidelity environmental textures through high-level semantic control. By operating on the continuous latent manifold of a neural audio codec, our approach bypasses the rigid structural constraints inherent to discrete tokenization. We propose a segmentation strategy that decomposes audio into overlapping segments, enabling a Continuous Normalizing Flow (CNF) to model the evolution of latent trajectories using Flow Matching. Our experiments demonstrate that a 36-million parameter instance of SCAPES can be trained on limited, uncurated datasets using a single consumer-grade GPU. Notably, convergence is achieved after training for approximately twice the source audio duration, yielding high-fidelity outputs with robust long-term stability and semantic consistency. Furthermore, we showcase the model's capacity for smooth semantic interpolation, providing a flexible and accessible tool for open research and creative sound design. Code, pretrained weights, audio examples, and an interactive demo are publicly available on our project page https://cordutie.github.io/projects/scapes.html --- ### SiLR: Structure-Preserving Admission and Process Reward for LLM Tool Agents Authors: Chenyu Zhou, Qiliang Jiang, Shuning Wu, Xu Zhou Published: 2026-09-04 Categories: cs.AI, cs.LG, eess.SY arXiv: https://arxiv.org/abs/2609.04629v1 A runtime gate for an LLM tool agent is usually cast as a filter. In a ReAct loop a rejected proposal is followed by another at the same state, so the gate is a search operator over the proposal stream whose admission criterion shapes which trajectories are reachable. We study post-violation recovery admission, where progress must be admitted while the system is still in violation, and identify the scalar projection trap: an aggregate-score gate accepts a locally improving proposal and commits the trajectory to a plateau. SiLR instead shadow-executes each proposal and admits it under a product order over the branch-level violation state (overloaded-branch support and per-branch severity). We prove that no scalar surrogate is sound for this order, so the failure is representational, not a matter of threshold tuning. On mined Gym-ANM scenarios, SiLR recovers 21/21 multi-action episodes against 0/21 for terminal and 9/21 for the best scalar gate, significant across the full 24-scenario benchmark. The terminal-versus-structured dichotomy holds across three model families and in CityLearn. Because admission rests on deterministic simulation, the LLM lies outside the trust boundary: a magnitude-redistribution attack that defeats both scalar and support-only baselines is contained only by the full per-branch predicate. With two constraint families active, every tested scalar projection admits physically unsafe actions; support-only admits the largest fraction (63.2% of 42,410; product order 0). In the hardest dual-family traces, scalar gates recover only through that unsafe class. Reused as a GRPO process reward, it outperforms its count projection in every mined scenario and is the only tested reward whose ungated policy exceeds the untrained base (0.844 vs. 0.778). Scalar projection loses the violation geometry at both design points; only the full product order is structurally sufficient. --- ### Leveraging Imperfect Restoration for Data Availability Attack Authors: Yi Huang, Jeremy Styborski, Mingzhi Lyu, Fan Wang, Adams Kong Published: 2026-09-04 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.04627v1 The abundance of online data is at risk of unauthorized usage in training deep learning models. To counter this, various Data Availability Attacks (DAAs) have been devised to make data unlearnable for such models by subtly perturbing the training data. However, existing attacks often excel against either Supervised Learning (SL) or Self-Supervised Learning (SSL) scenarios. Among these, a model-free approach that generates a Convolution-based Unlearnable Dataset (CUDA) stands out as the most robust DAA across both SSL and SL. Nonetheless, CUDA's effectiveness against SSL is underwhelming and it faces a severe trade-off between image quality and its poisoning effect. In this paper, we conduct a theoretical analysis of CUDA, uncovering the sub-optimal gradients it introduces and elucidating the strategy it employs to induce class-wise bias for data poisoning. Building on this, we propose a novel poisoning method named Imperfect Restoration Poisoning (IRP), aiming to preserve high image quality while achieving strong poisoning effects. Through extensive comparisons of IRP with eight baselines across SL and SSL, coupled with evaluations alongside five representative defense methods, we showcase the superiority of IRP. Code: https://github.com/lyumingzhi/IRP ---