# Research Papers — 2026-09 · bundle page 46 of 266 > Full-text markdown of papers #2251–2300 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/46 [← Previous bundle](https://aipod.dev/papers/archive/2026-09/page/45.md) [Next bundle →](https://aipod.dev/papers/archive/2026-09/page/47.md) ### Compute Time Scaling with Recursive Models for Combinatorial Optimization Authors: Zhengxi Zhang, Paul Swoboda Published: 2026-09-28 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.34585v1 We propose Tiny Recursive Models for Combinatorial Optimization (\ours{}), a general neural method for combinatorial optimization that scales both depth (how often we recursively invoke our network) and width (how much we sample in parallel). Both are fundamental for combinatorial optimization: hard instances demand a large amount of compute, while a small network is essential to avoid overfitting and capture the algorithmic essence of optimization. In particular, our method consists of a graph-aware tiny recursive model that iterates on a latent state with adaptive halting and needs only a lightweight problem-specific decoder. Compared with previous heatmap-based general neural solvers, it achieves a better balance between solution quality and inference speed on both the Traveling Salesman Problem~(TSP) and the Maximum Independent Set~(MIS) problem, and remains competitive with hybrid methods that combine neural components with heuristics specific to each problem. With the same backbone architecture for both tasks, \ours{} outperforms every diffusion-based solver on TSP from 500 to 10,000 cities at a lower inference cost, and on the standard Erdős--Rényi-(700-800) MIS benchmark it surpasses all neural solvers except those that only work well on MIS. We then explore self-relabeling for self-supervised training. We periodically replace the current set of training labels with the model's own better solutions, as an alternative training signal. Self-relabeling can, while forgoing supervision from near-optimal solutions, still result in on-par quality. --- ### In-game Toxic Detection: Bi-directional Representations with Attention Residuals Authors: Yuanzhe Jia Published: 2026-09-28 Categories: cs.CL, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2609.34584v1 In-game toxic language has emerged as a critical concern in the gaming industry and community. While several frameworks and models for online game toxicity analysis have been proposed, detecting toxicity in player chat utterances remains a formidable challenge: stemming not only from the extremely short length of such utterances but also from the heavy reliance on game slang, abbreviations, and domain-specific jargon, which generic language models are poorly suited to recognize. This paper presents a shared task for in-game toxic language detection built upon real-world in-game chat data, and proposes the best-preforming model for the toxic language slot filling: Bi-directional Representations with Attention Residuals (BRAR). Experimental results demonstrate that BRAR effectively captures the global context and outperforms the existing baselines on slot filling. --- ### SpeechCritic: Learning a Diagnostic Speech Judge from Limited Human Preferences Authors: Mingyue Huo, Shivam Mehta, Bhavin Jawade, Yinghong Lan, Haoqi Li Published: 2026-09-28 Categories: cs.AI, cs.SD arXiv: https://arxiv.org/abs/2609.34582v1 Human speech conveys rich perceptual information, such as emotion and speaker identity, yet most automatic speech quality judges reduce it to a single naturalness score. We study diagnostic speech judges: given two candidates, a diagnostic judge decides which is better, along which perceptual dimensions (e.g., timbre, emotion, timing) they differ, and which audible cues support its decision. Learning such judges is challenging: expert annotation is costly, and simply prompting a frontier audio-language model to produce labels is unreliable: our probing reveals substantial errors and unstable instruction following. We introduce SpeechCritic, which learns a diagnostic judge in a reference-conditioned cross-lingual setting from only about 300 human-labeled comparisons. Rather than replacing the frontier model, SpeechCritic calibrates it with these labels: for each dimension, it selects the acoustic measurements that agree with human judgments, maps them to A/Tie/B probabilities, and passes these to the model as non-binding hints alongside the audio. Compared with the same model labeling without hints, this raises dimension-level agreement with humans by 6.3 points and cuts the mismatch with human Tie rates by 10.4 points. We then train a 7B judge on this supervision and find that different training signals shape different judge behaviors: SFT establishes the task, OPD transfers the teacher's dimension-level strengths and weaknesses, and RL helps most on clear-cut comparisons where human raters agree. Notably, human listeners also find that RL makes rationales cite more specific, localized acoustic cues, although it never directly rewards rationale text. Finally, we show that the pipeline is language-pair agnostic by instantiating it on both English-Japanese and English-Spanish. Together, these results demonstrate a path from limited human preferences to a diagnostic speech judge. --- ### Counterfactual Attention Policy Distillation for Temporal Video Grounding Authors: Shaobo Ju, Haiyang Yu, Xuecheng Wu, Qiong Wu, Jiacong Wang, Fan Shi, Jun Peng, Yiyi Zhou Published: 2026-09-28 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.34581v2 Temporal video grounding is a key capability of advanced Multimodal Large Language Models (MLLMs) for the thorough understanding of video events, which is however often limited by repeated actions and visually similar contexts in long videos. In this paper, we study this issue from the perspective of On-policy distillation (OPD) and propose a new training regime for MLLMs termed Counterfactual Attention Policy Distillation (CAPD). In particular, OPD is a viable solution for MLLMs via providing dense teacher supervision on student-generated trajectories. But its next-token based teacher-student distillation is hard to identify the specific video segments supporting each predicted timestamp, which is critical for temporal grounding. In this case, CAPD measures how masking each temporal group changes the teacher's output distribution. The resulting counterfactual influence calibrates the teacher's attention and weights token-level distillation, allowing the student to learn the temporal evidence that affects boundary prediction. To validate CAPD, we trained it on Qwen3-VL-8B-Instruct using only 2,500 samples for one epoch, and evaluated it on the TimeLens and multiple general video benchmarks. Experimental results show that CAPD improves average recall by 12.0% relative to GRPO on TimeLens while preserving general video understanding, achieving comparable accuracy to the base model. --- ### GenNVS: Geometry-enhanced Novel View Synthesis via Disentangled 3D Prior Authors: Yajiao Xiong, Youyu Luan, Xiaoyu Zhou, Yongtao Wang Published: 2026-09-28 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2609.34579v2 Single-image novel view synthesis remains challenging because the underlying 3D geometry is highly ambiguous. Recent diffusion-based approaches produce plausible results, but they often struggle to preserve the geometric structure and spatial coherence of foreground objects. We present GenNVS, a framework for geometry-enhanced novel view synthesis via a disentangled 3D prior. Specifically, GenNVS models foreground objects and the background with 3D Gaussian Splatting and aligns them through a coarse-to-fine geometric optimization process to form a unified 3D scene. This scene conditions a video diffusion model through the proposed Dual-Stream Masking mechanism, which guides synthesis by jointly exploiting rendered validity masks and geometry-aware warping. Experimental results show that GenNVS performs favorably against recent methods in both visual quality and geometric accuracy, while naturally supporting flexible scene editing. --- ### Calibrated Uncertainty for Informative Path Planning in Aquatic Environmental Monitoring Authors: Samuel Yanes Luis, Alejandro Casado Pérez, Alejandro Mendoza Barrionuevo, Dame Seck Diop, Sergio Toral Marín, Saniel Gutiérrez Reina Published: 2026-09-28 Categories: cs.AI, cs.IR arXiv: https://arxiv.org/abs/2609.34577v1 Informative Path Planning for scalar field reconstruction uses predictive uncertainty to direct sensing vehicles toward maximally informative locations. Gaussian Processes provide this signal but their stationary isotropic kernels are misspecified for non-homogeneous phenomena such as oil spills, producing miscalibrated estimates that degrade planning. We investigate whether replacing the Gaussian Process with a well-calibrated Deep Ensemble improves path planning outcomes, and whether uncertainty quality interacts with the choice of planning algorithm. Five strategies ($ε$-Greedy, Value Greedy, Uncertainty Greedy, Monte Carlo Tree Search, and Receding Horizon Orienteering) share a common Deep Ensemble backbone trained on physics-based oil spill simulations. On held-out stochastic spill scenarios, the Deep Ensemble reduces normalised reconstruction error by $83\%$ relative to the Gaussian Process baseline. Crucially, well-calibrated uncertainty amplifies the importance of the planning strategy: the performance gap between algorithms is negligible under miscalibrated models but becomes substantial under the ensemble, where multi-step lookahead planners outperform greedy selection by up to $32\%$ in reconstruction error and achieve IoU above $0.85$. Monte Carlo Tree Search is the recommended planner, matching Orienteering in reconstruction quality at an order-of-magnitude lower computational cost. --- ### Diffusion Subgoal Planning for Long-Horizon Offline Goal-Conditioned Reinforcement Learning Authors: Hengrui Zhang, Yuhu Cheng, C. L. Philip Chen, Xuesong Wang Published: 2026-09-28 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.34575v1 Offline goal-conditioned reinforcement learning (GCRL) learns goal-directed policies from reward-free data, but in long-horizon tasks, goal-conditioned value functions often provide unstable guidance due to sparse rewards and discounting. Hierarchical methods partially mitigate this issue via subgoal decomposition; however, high-level decision-making still relies on noise-sensitive value estimates, leading to unstable behavior in complex environments. We address this limitation by proposing \textbf{D}iffusion \textbf{S}ubgoal \textbf{P}lanning (\textbf{DSP}), a diffusion-based framework for high-level subgoal generation. DSP casts high-level planning as guided generative inference over goal-conditioned subgoals and learns both conditional and unconditional flows, enabling classifier-free guidance to introduce a goal-directed bias at inference time. By removing explicit value-based guidance from high-level planning, DSP generates reachable and goal-directed subgoals through a generative model while retaining hierarchical execution. Experiments on offline GCRL benchmarks demonstrate that DSP outperforms prior methods on a range of navigation and manipulation tasks, with particularly strong performance in maze environments that require multi-step subgoal planning. --- ### Nudgeability: Reasoning Models Follow Confidence Signals Without Tracking Their Own Competence Authors: Rohit Saxena, Utkarsh Upadhyay Published: 2026-09-28 Categories: cs.AI, cs.CL, cs.LG arXiv: https://arxiv.org/abs/2609.34572v1 Reasoning language models that can call tools must decide during inference whether to answer unaided or delegate. Any self-reflection mechanism for this must answer three questions: where the reflective signal comes from (verbal reports, output distributions, hidden states, a separate predictor), how it is presented to the model (numerical prediction, confidence token, prompt injection), and whether it changes the model's subsequent action. We isolate the third question. At a fixed point in otherwise identical reasoning trajectories, we insert a single first-person sentence expressing either confidence or doubt; the model then continues reasoning and chooses whether to answer directly or call a tool. Comparing these counterfactual continuations measures the causal effect of the reflective signal on delegation. We call this behavioral response Nudgeability and measure it along two dimensions: sensitivity, how strongly confidence and doubt change delegation rates, and targeting, whether delegation increases for problems the model cannot solve unaided and decreases for those it can. Across nine small-to-medium open-weight reasoning models from three families (Qwen, Gemma, and GLM) and two tasks, models are consistently sensitive: doubt increases delegation and confidence decreases it, with a median confidence-to-doubt swing of 20.6 percentage points, and 53 to 70 points for the larger provider-served models. This responsiveness is poorly targeted: a median 42% of induced flips are well-targeted, only a +2 percentage-point lift over a random-selection baseline. Confidence language is thus a strong control surface for delegation, but current models use it only weakly in accordance with their actual competence. Nudgeability offers a simple, post-training-free way to evaluate both sensitivity and targeting as endogenous self-reflection mechanisms mature. --- ### PersonaManifold: Revealing and Exploiting Curved Geometry in LLM Persona Representations Authors: Rui Xu, Yinghui Xu, Libo Wu Published: 2026-09-28 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.34571v1 Controlling persona in large language models (LLMs) at inference time is important for role-playing, personalized dialogue, and social simulation. Recent methods extract persona vectors from the model's activation space and apply Euclidean operations---addition, scaling, and linear interpolation---under the linear representation hypothesis. However, these methods themselves report systematic failures: non-orthogonal trait dimensions, asymmetric ceiling and resistance effects, and significant deviations in multi-trait composition, suggesting that the linear isotropic assumption does not hold. We propose PersonaManifold, a framework that models persona representations as points on a curved, low-dimensional Riemannian submanifold in activation space. We estimate the manifold's intrinsic geometry---local metric tensors, geodesic distances, and Ollivier-Ricci curvature---and introduce geodesic steering, which interpolates between personas along manifold geodesics rather than Euclidean straight lines. We also propose the Behavioral Similarity Triplet (BST) benchmark, which automatically generates situational questions grounded in six established psychological constructs and defines persona similarity through behavioral responses rather than self-report questionnaires. Experiments on three open-source LLMs show that persona activations form a manifold with heterogeneous curvature, geodesic distance predicts behavioral similarity more accurately than Euclidean alternatives with independent contributions from anisotropy and curvature, and geodesic steering produces more coherent intermediate personas on both our BST benchmark and external evaluations, with the advantage concentrated in high-deviation regions where the manifold deviates most from flatness. --- ### Recent Advances in Agentic Agri-Robotic Phenotyping: A Perspective Review from Fragmented Multimodal Sensing to Unified PhenoAgent Intelligence Authors: Muhammad Owais, Ehtesham Iqbal, Samee Ullah Khan, Muhammad Umraiz, Yusra Abdulrahman, Irfan Hussain Published: 2026-09-28 Categories: cs.CV, cs.ET arXiv: https://arxiv.org/abs/2609.34567v1 This review examines the evolution of plant phenotyping from conventional manual trait measurement to high-throughput, robotic, and artificial intelligence-driven crop monitoring. Despite significant advances in imaging, autonomous platforms, multimodal sensing, and deep learning, current phenotyping systems remain fragmented across sensing modalities, crop traits, growth stages, environments, and management objectives. We therefore frame phenotyping as an integrated \emph{seed-soil-plant-environment-management} (SSPEM) intelligence problem, where crop performance reflects interactions among seed quality, root-zone conditions, plant development, environmental exposure, and management actions. The review synthesizes conventional, high-throughput, robotic, and AI-driven phenotyping approaches, highlighting their capabilities and persistent limitations in temporal integration, multimodal reasoning, biological interpretation, and actionable decision support. Building on this analysis, we introduce a conceptual PhenoAgent framework that extends phenotyping beyond the estimation of isolated traits to evidence-based crop-state interpretation, uncertainty-aware reasoning, and management-oriented support. The PhenoAgent concept primarily brings together scattered advances in phenotyping to deliver insights ranging from detailed to high-level, such as what is happening in the crop, why it might be occurring, what evidence is missing, and what actions or additional measurements should be considered. We also discuss challenges in dataset scarcity, annotation, benchmarking, model generalization, and explainability. By linking multimodal phenotyping with agentic AI and closed-loop decision support, this review outlines a path to interpretable, scalable, and deployment-oriented crop intelligence. --- ### Retracing Hodgkin and Huxley: State Recovery Does Not Certify Mechanism Authors: Peiyu Zang, Jiayi Hao, Yongqiang Cai Published: 2026-09-28 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.34566v1 Predicting observed dynamics does not establish recovery of the underlying physical mechanism. Can machine learning retrace the hidden-state reasoning behind the Hodgkin-Huxley (HH) model? We train structured latent models on simulated current and voltage, withholding gate identities and trajectories from training and model selection. We then test response prediction, state recovery, protocol transfer, and agreement with HH dynamics. Prediction error and its cross-seed spread both drop sharply at three latent dimensions under the tested protocols, while gate recovery under new protocols improves through five to six coordinates. State recovery depends on which observations the chart uses. Observed voltage improves current-clamp decoding relative to freely predicted voltage. Under voltage clamp, adding latent state to command voltage raises m-state $R^2$ from 0.976 to above 0.99, yet the transported field disagrees with HH on identical smooth samples. Known invertible HH coordinates achieve high fast-m field agreement under the same audit procedure. An exact HH identity decomposes the discrepancy into time-scale-weighted state error and a residual in the transported field; these terms can cancel or reinforce. These findings concern the tested models and charts. They support evaluating state and dynamics recovery separately, including chart inputs and transported-field agreement across interventions. --- ### FlowState: Execution State as Memory for Long-Horizon LLM Agents Authors: Minghao Li, Bangyan Li, Zifan Wang, Yulong Li, Hu Xu, Gan Zhang, Jingtong Wu, Wenqiang Xu Published: 2026-09-28 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.34565v1 Long-horizon tasks require LLM agents to continually draw on information from earlier interactions. However, retaining the full history increases context costs, while compressing it risks losing details needed later, and the relevance of historical information often becomes apparent as the task progresses. To address these challenges, we propose FlowState, which treats execution state as memory that can be retained and revisited across requests, unifying current decision-making with the reuse of historical information. FlowState preserves semantically typed state nodes, their relations, and references to raw tool observations, separating persistent retention from on-demand access. Within a single execution loop, Incremental State Update (ISU) maintains the current state based on new inputs and feedback, while Progressive State Access (PSA) progressively reveals historical states and supporting evidence as needed during reasoning. Together, these mechanisms enable agents to reassess prior decisions in light of new information and guide subsequent actions. Compared with a full-context baseline using the same DeepSeek-V4-Flash model, FlowState improves the average success rate on MemoryArena and the average pass rate on $τ^3$-Bench by 4.55 and 13.95 percentage points, respectively, while reducing total token consumption by 43.2% and 40.6%. These results demonstrate the performance and efficiency advantages of FlowState on long-horizon tasks. --- ### GLF-Q: Global-Local Feature-based Quantization for Vision Transformers Authors: Peilin Sun, Guang Liang, Jin Tong, Jianxin Wu Published: 2026-09-28 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.34564v1 Post-training quantization (PTQ) efficiently compresses Vision Transformers (ViTs) without retraining, yet suffers severe accuracy degradation at low bit-widths. Existing optimization-based PTQ methods guide block reconstruction via either soft logits or second-order Hessian proxies. Logit supervision is prone to overfitting on limited calibration data, while Hessian approximations incur structural truncation errors. To address these limitations, we propose \textbf{GLF-Q}, a novel PTQ framework guided by Global-Local Feature alignment. GLF-Q propagates quantized block outputs through downstream full-precision layers to align penultimate-layer representations under local output regularization, providing downstream feature supervision without explicitly approximating the Hessian or using a Taylor expansion. Furthermore, offline Hadamard transformations are introduced with zero runtime overhead to disperse activation outliers across channels, effectively contracting dynamic ranges and reducing quantization errors. Meanwhile, optimizing this loss via a Straight-Through Estimator (STE) achieves rapid convergence, bypassing continuous relaxation rounding formulations such as AdaRound. Extensive experiments across representative ViT architectures demonstrate that GLF-Q with standard uniform quantizers substantially outperforms state-of-the-art methods under 3-bit quantization on image classification. In addition, GLF-Q exhibits strong out-of-domain calibration robustness and achieves speedups under 8-bit GPU deployment. --- ### Rethinking Latent Visual Reasoning: Grounding Latent Reasoning in Visual Evidence Authors: Xi Xiao, Tianchen Zhao, Youngeun Kim, Zhuowei Li, Linghan Xu, Jiaye Wu, Zheng Zhang, Xiang Xu, Xuanbai Chen, Farhan Tejani, Jakub Zablocki, Julia Xu et al. Published: 2026-09-28 Categories: cs.CV, cs.CL, cs.LG arXiv: https://arxiv.org/abs/2609.34563v1 Latent visual reasoning (LVR) enables multimodal large language models (MLLMs) to perform intermediate computation in continuous latent tokens rather than expressing every reasoning step in words. However, unlike textual CoT, latent reasoning is not directly observable, making it difficult to supervise what latent tokens learn. In this work, we first conduct a thorough analysis of latent-token behavior and identify a latent evidence-credit gap: latent tokens respond only weakly to image perturbations that alter the correct answer. We hypothesize that this issue stems from the lack of explicit supervision during GRPO training. These findings suggest that a final-answer reward provides too little guidance on what visual evidence to preserve or how credit should be assigned across latent tokens. To bridge this gap, we propose ReaLVR, which brings visual-evidence supervision to the model's own free-running latent trajectories. ReaLVR contrasts correct and model-generated wrong answers to determine where stronger supervision is needed, and relevant and mismatched visual evidence to specify what to preserve. Across three model families, ReaLVR consistently outperforms evaluated LVR baselines, achieving the highest five-task average of 63.7% on Qwen2.5-VL-7B. Crucially, we are the first to scale visual reasoning in latent space, showing that our framework continues to deliver robust improvements at frontier model scales up to 235B. Further analyses show more question-sensitive latent-token positions, stronger alignment with relevant visual regions, and greater fixed-context dependence on the most attended latent tokens. --- ### Single-Layer MeMo as a Randomized Hamming-Kernel Classifier Authors: Alessandro Straziota Published: 2026-09-28 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.34562v1 MeMo (Zanzotto et al., 2025) is a recent language-model architecture that stores associations between token contexts and next tokens in a correlation matrix memory. In this work, we study its single-layer form and show that its ideal retrieval rule is a multiclass classifier based on the positional Hamming kernel. The MeMo architecture represents both the sequence features and the output labels with Gaussian random codes. Its score is therefore a doubly randomized sketch of the ideal classifier. Under independent input and output codebooks, we bound the errors introduced by context sketching and output decoding, characterize their dependence on model and data parameters, and give a margin-based guarantee for recovering the ideal prediction. Controlled simulations support the trends predicted by the analysis. On a restricted WikiText-2 next-token task, we compare single-layer MeMo with classical baselines and show that it can offer a useful trade-off among predictive accuracy, memory, and throughput, particularly on a GPU, where its matrix operations can be parallelized. --- ### Brain-Conditioned Action Policies for Neural Motor Decoding Authors: Luyao Jin, Running Zhao, Huan Zhao, Vincent C. K. Cheung, Wei-Hsin Liao Published: 2026-09-28 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.34561v1 Motor brain-computer interfaces (BCIs) aim to decode motor intention, enabling people with paralysis to control external devices. Neural motor decoding typically learns task-specific mappings from neural activity to kinematics, yet remains constrained by scarce paired neural-action data. We propose BrainVLA, a framework that enables neural motor decoding by drawing on a pretrained vision-language-action (VLA) model through language-mediated alignment. BrainVLA mitigates reliance on scarce paired neural-action data by leveraging VLA policies. We first construct VLA-compatible datasets including paired neural activity, action signals, language instructions, and rendered visual observations. Then, we adapt the OpenVLA-OFT policy to the target action spaces through LoRA fine-tuning. To establish an effective interface through which neural activity can convey motor intention to adapted VLA policies and guide action generation, we train a neural encoder via neural-language alignment, using language representations as semantic targets to capture latent motor intent from neural activity. The resulting neural representations serve as an endogenous intention signal to guide VLA policies to generate executable actions, while visual observations provide complementary information about the evolving task state. BrainVLA is evaluated on two neural motor datasets with different action dimensionalities using causal rollout decoding. It outperforms the evaluated baselines in cross-session decoding $R^2$ and task success rate, while demonstrating high training data efficiency. These results establish a route for neural motor decoding to draw on large-scale robotic priors through brain-conditioned VLA policies. --- ### ACPruner: Visual Token Pruning as Biased Attention Coverage Maximization in LVLMs Authors: Xu Li, Yuxuan Liang, Yi Zheng, Zhe Liu, Xiaolei Chen, Haotian Chen, Rui Zhu, Fan Shi, Xiangyang Xue Published: 2026-09-28 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.34558v1 Large Vision-Language Models (LVLMs) face significant computational inefficiencies caused by the large number of visual tokens. Existing visual token pruning methods mainly focus on either retaining individually important tokens or selecting mutually diverse ones. In this work, we revisit visual token pruning from a coverage perspective and formulate it as a biased attention coverage maximization problem. The key idea is to select a compact token subset whose encoder-side outgoing attention can jointly cover the image while assigning higher coverage priority to more informative regions. From this perspective, we propose ACPruner, a training-free visual token pruning framework for efficient LVLM inference. ACPruner first estimates token importance by combining intra-modal saliency and inter-modal relevance, then derives token-wise coverage from attention patterns within the vision encoder, and finally performs greedy selection to maximize the proposed coverage objective. Extensive experiments across multiple LVLM backbones, including LLaVA-1.5-7B/13B, LLaVA-NeXT-7B/13B, Qwen2.5-VL-7B, and LLaVA-OneVision-7B, show that ACPruner consistently achieves strong performance retention while delivering substantial end-to-end inference speedups. --- ### SkillRubric: Co-Evolving Actor Guidance and Evaluator Rubrics for Multimodal Agents Authors: Bingqing Jiang, Guoxi Zhang, Jasper Wang, Auric Wang, Bingning Wang, Tianyi Lin, Zichao Yu, Yujin Han, Ziye Ma, Difan Zou Published: 2026-09-28 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.34557v1 Recent work incorporates reusable skills distilled from past interactions into multimodal agent training, providing procedural guidance for long-horizon planning and tool use. However, policy optimization in these methods remains driven primarily by sparse outcome rewards, providing little supervision for intermediate decisions. Rubric-based rewards address this limitation through explicit intermediate criteria, but reliable rubrics are difficult to construct at scale and often disconnected from the procedure followed by the actor. We observe that a well-structured skill naturally specifies both how to act and what successful execution should achieve. Based on this insight, we introduce SkillRubric, which represents each skill through aligned actor-facing guidance and an evaluator-facing rubric. A multimodal verifier evaluates skill-defined goals using screenshots and tool outputs, assigning completion and progress rewards to the responsible turns. We further introduce an alternating co-evolution scheme that validates guidance revisions through paired rollouts under a frozen policy and rubric revisions offline under fixed guidance. Experiments across diverse multimodal agent benchmarks demonstrate consistent performance gains, while controlled paired rollouts further show that evolved skills provide more effective guidance for planning and tool use than their preceding versions. --- ### RoPE is Dead, Long Live RoPE: Towards Scalable Data-aware Positional Encodings Authors: Jarod Lévy, Mathurin Videau, Jad Yehya, Jean-Rémi King, Stéphane d'Ascoli, Thomas Moreau Published: 2026-09-28 Categories: cs.CL, cs.LG arXiv: https://arxiv.org/abs/2609.34556v1 Transformers process tokens without any inherent notion of order, making positional encoding a fundamental requirement rather than an architectural refinement. Rotary Position Embedding (RoPE) has become the default positional encoding in modern language models, yet it is heavily biased toward nearby tokens. Existing alternatives have been evaluated under different settings, leaving the literature fragmented and without a clear replacement. We bring structure to this landscape by examining a specific weakness of RoPE: its slow frequency bands, whose wavelengths exceed the training context and expose models to unseen angles during extrapolation. We therefore introduce Data aware RoPE (DaRoPE), which preserves standard RoPE on the fast bands but replaces absolute position on the slow bands with bounded coordinates learned from contextual representations. Therefore, the slow-band geometry depends on the data rather than only on positional distance. We compare representative encodings under matched conditions across synthetic tasks, symbolic music, genomics, neural signals, and language models spanning 124M to 50B parameters. Across these experiments, DaRoPE leads on non-text benchmarks, mitigates recency bias, while remaining best or on par in language modeling and length extrapolation. Moreover, the learned coordinates also make the mechanism interpretable, revealing how attention layers leverage contextual information beyond token distance. Together, these results support DaRoPE as the best overall default among the evaluated methods, when there is no domain-specific reasons to prefer another. --- ### PulseInfer: I/O-Centric Sparse KV Cache Offloading for Efficient Long-Context LLM Decoding Authors: Qiuyang Zhang, Kai Zhou, Kai Lu, Haocheng Lu, Jian Zhou, Yuanpeng Su, Kun Bao, Jiguang Wan, Fei Wu Published: 2026-09-28 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.34555v1 Long-context LLM serving is increasingly bottlenecked by decode, where large KV caches limit batch size and underutilize GPUs. Sparse KV cache offloading expands effective capacity by storing most historical KV blocks in CPU DRAM and recalling only selected blocks on demand. However, we find that existing offloading systems shift the bottleneck to CPU-GPU recall I/O: recall volume varies widely across layers, decode steps and requests, while headwise sparse selection fragments recalls into many small PCIe transfers. This paper presents PulseInfer, an I/O-centric sparse KV cache offloading system. PulseInfer hides variable recall latency with interruptible layer-wise scheduling, adapts offloading decisions with IO-Adaptive Offloading Admission, and coalesces fragmented transfers using SoloHead sparse selection and a gather-scatter I/O engine. Implemented on SGLang, PulseInfer improves decode throughput by up to 4.7x over SGLang and 2.6x over the best existing offloading baseline, while reducing TPOT by up to 76% and preserving near-lossless accuracy. --- ### Verifying Neural Networks with Reinforcement Learning Authors: Hai Duong, Thanh Le, ThanhVu Nguyen Published: 2026-09-28 Categories: cs.LG, cs.SE arXiv: https://arxiv.org/abs/2609.34553v1 Formal verification can play a key role in ensuring the reliability of Deep Neural Networks (DNNs) deployed in safety-critical systems. Modern DNN verifiers employ a branch-and-bound framework, which alternates between branching (splitting into smaller subproblems) and bounding (pruning subproblems) to efficiently explore the verification space. However, existing branching heuristics make greedy decisions based on static scoring functions. They do not anticipate long-term efficiency or leverage the growing availability of verification data to improve performance. This work introduces RSB, a reinforcement learning framework that learns to refine baseline branching heuristics. It trains an actor-critic architecture to maximize cumulative future rewards rather than immediate scores. The actor generates attention weights from observations of raw neuron features and learned graph embeddings, which rescale baseline heuristic scores to guide neuron branching. Evaluation on 600 challenging instances demonstrates that RSB consistently outperforms state-of-the-art branching heuristics, solving 11% more instances while reducing branch exploration by 50%. --- ### On Parameter Symmetries and Conservation Laws in Gradient Flow Authors: Khang Nguyen, Guido Montúfar Published: 2026-09-28 Categories: cs.LG, math.OC arXiv: https://arxiv.org/abs/2609.34549v1 Parameter space symmetries and conservation laws play an important role in understanding the loss landscapes and implicit biases of neural networks. Inspired by Noether's theorem in physics, prior works have sought to derive conservation laws under gradient flow from parameter symmetries, but the scope and limitations of this connection remain unclear. We develop a unified geometric framework that clarifies the precise relationship between the two notions, including the conditions under which symmetries correspond to conservation laws. We introduce a notion of compositional identifiability and use it to establish a general inheritance principle for complete characterizations of symmetries and conservation laws in multilayer networks. We apply the framework to multi-head and grouped-query attention, polynomial neural networks, and square deep linear networks. --- ### SGG-ReflAct: Sub-Goal Guided ReflAct with Structured Planning for Reliable Long-Horizon Reasoning Authors: Jaeho Jung, Sung Hoon Jung Published: 2026-09-28 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.34548v1 Recent advances in reasoning backbones have empowered large language model (LLM)agentstotackle complex, multi-step tasks. However, as reasoning horizons grow, inconsistent internal beliefs induce intermediate errors that cause agents to drift from their goals. This limitation also persists in REFLACT, which reflects only on the end-goal at each step without explicitly considering intermediate sub goals. To address this problem, we propose SGG-ReflAct (Sub-Goal Guided Re flAct), a reasoning backbone that integrates sub-goals generated through a single path LLM planner into the reflection process. We further extend this framework to BeamSGG-ReflAct, which replaces the single-path planner with a beam search based LLM planner for structured plan exploration. We run experiments on ALF World, ScienceWorld, and Jericho with multiple LLM models. SGG-ReflAct out performs REFLACT in nearly all settings, achieving best success rate gains of 14.9 percentage points on ALFWorld and 8.0 percentage points on ScienceWorld with Llama-3.1-8B-Instruct. Our experimental analysis shows that SGG-ReflAct re duces hallucinated actions and achieves its largest gains on procedurally ordered tasks. Furthermore, experimental results with BeamSGG-ReflAct show that the backbone's effectiveness depends on plan quality: explicitly specifying the re quired operations recovers gains that plan searching alone cannot achieve. These results demonstrate that SGG-ReflAct offers a practical and highly effective rea soning backbone, enabling LLM agents to achieve reliable performance in com plex, long-horizon tasks through easy integration. --- ### ActionLens: Diagnosing Spatial-Temporal Binding Failures in Vision-Language Models Authors: Gueter Josmy Faure, Min-Hung Chen, Hao Ping Wang, Timothée Lardy, Hung-Ting Su, Winston H. Hsu Published: 2026-09-28 Categories: cs.CV, cs.AI, cs.CL arXiv: https://arxiv.org/abs/2609.34547v1 Video-capable vision-language models score above 80\% on popular benchmarks yet struggle with spatial-temporal binding: associating the right action with the right person at the right moment. We introduce ActionLens, a diagnostic benchmark of 6,701 multiple-choice video questions spanning five targeted diagnostics: transition detection, actor-specific identification, concurrent action binding, directed interaction reasoning, and gaze detection. Ground-truth answers are derived deterministically from 1.58 million per-second, per-person annotations. Fourteen rounds of human quality engineering raised answer clarity from 53% to above 90% human accuracy. Across 20 VLMs, the full-set leader scores 68.8%; on the human-reviewed subset, it scores 65.9% versus 91.0% for the pooled human reference. Gaze detection remains near chance against 89.6% human accuracy. On actor disambiguation, reference-interface controls show that relational descriptions recover 5.55--13.25 points over static coordinates, confirming a substantial numeric-parsing penalty; yet visual boxes still lead every model by 1.15--6.50 points, exposing a residual unboxed actor-resolution gap. A binding-trap analysis shows models systematically select the wrong actor's action. ActionLens provides diagnostic measurements of these distinct failure modes across model families and scales for direct comparison. We release all data, code, and evaluation scripts at https://anonymous.4open.science/r/lmms-eval-2276 --- ### Remember Before You're Asked: MemDream for Self-Probing Memory Evolution Authors: Mingfei Lu, Mengjia Wu, Runsong Jia, Zhe Luo, Yi Zhang Published: 2026-09-28 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.34545v1 Memory is essential for enabling LLM-based agents to maintain coherent, personalized behavior over long-horizon interactions. However, existing memory systems share a fundamental limitation: they never proactively test their own memory, repairing it only after real queries expose weaknesses. This reactive paradigm means every retrieval failure corresponds to a real interaction in which the cost has already been paid. We propose MemDream, a framework that enables self-probing memory evolution for LLM agents. Our framework periodically enters offline dream cycles where three specialized agents (Dreamer, Analyst, Consolidator) collaboratively probe, diagnose, and repair the memory graph before failures occur. A policy trained via Group Relative Policy Optimization learns which repair operations produce durable retrieval improvements, while a soft decay mechanism provides reversible forgetting driven by the same anticipatory signal. Experiments on LoCoMo and MemoryAgentBench demonstrate that MemDream improves answer F1 by 4.5 points on LoCoMo and achieves a 9.1-point higher overall score on MAB over the strongest reactive-evolution baselines. --- ### APOLO: Automatic Prompt Optimization for Ontology Learning Authors: Huu Tan Mai, Roman Kochnev, Cuong Xuan Chu, Lukas Lange, Heiko Paulheim, Daria Stepanova Published: 2026-09-28 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.34540v1 Ontology Learning (OL) from text has advanced with the emergence of Large Language Models (LLMs), but it remains challenging due to the limited availability of annotated training data and the difficulty of adapting LLMs to perform OL effectively. We address this via APOLO - Automatic Prompt Optimization for Ontology Learning, by casting OL as an explicit prompt optimization problem over LLM modules. To obtain training data, we employ a multi-agent system that generates text-ontology pairs from existing expert-curated ontologies. We then propose two ontology learner architectures: a greedy and an autoregressive learner, and optimize both using GEPA, a greedy evolutionary prompt optimizer built on DSPy. Experiments on two ontologies - a biomedical (DOID) and a plant ontology (PO) show consistent improvements after optimization across nearly all model and mode combinations, with autoregressive learners achieving the largest gains. Our results demonstrate that prompt optimization is a viable and lightweight alternative to fine-tuning for OL, and that the autoregressive formulation better captures ontological structure than the greedy approach. --- ### TSGate: Timestep-Aware Gated Attention for Diffusion Transformers Authors: Boyu Zhang, Yifan Liu, Shuxia Lin, Qingjian Ni, Yinfei Xu, Xu Yang Published: 2026-09-28 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.34539v1 Diffusion Transformers (DiTs) have emerged as the dominant architecture for high-fidelity image and video generation. Recent DiT systems increasingly use structured prompts for training, improving caption quality and prompt adherence. However, their generation quality can degrade severely under out-of-domain (OOD) prompts, including the free-form descriptions supplied by users at inference time. Although LLM-based rewriting can convert these prompts into structured formats, it does not guarantee that the rewritten prompts align with the training distribution. Our analysis links this degradation to attention sinks and reduced early-step image-to-text attention and shows that sink suppression alone is insufficient to restore generation quality. Despite effective sink suppression, models trained with standard gated attention exhibit reduced early-step image-to-text attention and suboptimal generation quality. Based on these insights, we propose Timestep-Aware Gated Attention (TSGate), which injects a timestep-conditioned bias into the gate signal so that gating behavior adapts across denoising steps. Extensive experiments show that TSGate consistently outperforms both the baseline and standard gated attention across multiple benchmarks, improving the raw-prompt DPG score by 9.5% over the baseline. --- ### Shallow Queries, Mature Values: Depth-Asynchronous Self-Speculation for Looped Transformers Authors: Guanghao Li, Zihan Su, Hao Yu, Jinyang Jiang, Tao Ren, Zehao Li, Feng Lu, Ming Tang, Chun Yuan Published: 2026-09-28 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2609.34538v1 Looped Transformers reuse a shared block across recurrent depths, making autoregressive decoding expensive because every generated token requires many sequential recurrent passes. Self-speculative decoders reduce this cost by drafting at an early depth and verifying at full depth, but typically bind draft computation to prefix representations from the same recurrent depth. We find that queries and keys approach their final-depth representations earlier than values, and controlled prefix-channel interventions show that mature values substantially improve shallow draft predictions. Motivated by this asymmetry, we introduce Depth-Asynchronous Self-Speculation (DAS), which decouples the depth of draft computation from the depth of verified-prefix representations it reads. Its Mature-V primitive lets shallow queries retrieve full-depth prefix values without additional recurrent computation. We further develop DAS-Wave, which combines depth-asynchronous prefix reads with carried parallel refinement, progressive block growth, and an independent full-depth verifier. Across four recurrent-model checkpoints and mathematics and code workloads, DAS-Wave achieves 4.00--6.96$\times$ mean throughput speedup over paired full-depth autoregressive decoding in the same inference stack. These results identify prefix-information depth as an effective design axis for recurrent self-speculation. --- ### The Marathon of Scientific Reasoning: Robustness of Scientific Agents to Perturbations in Multi-Turn Interactions Authors: Xiaoting Lyu, Xinbo Ma, Yufei Han, Hangwei Qian, Ziyang Lin, Bin Wang, Bin Wang, Wei Wang Published: 2026-09-28 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.34537v1 Large language model (LLM)-based scientific agents are increasingly used for scientific problem solving, yet their robustness to imperfections arising during multi-turn interactions remains poorly understood. We introduce \textsc{SciARP} (\textbf{Sci}entific \textbf{A}gent \textbf{R}obustness to \textbf{P}erturbations), a benchmark for evaluating scientific agents under scientifically plausible perturbations throughout multi-turn problem solving. \textsc{SciARP} transforms 620 scientific problems into interdependent tasks of 3--13 turns and defines 13 perturbation types spanning problem understanding, evidence processing, reasoning, and conclusion formation. Clean and perturbed versions of each task are independently executed under matched settings, producing paired live trajectories for evaluating both task success and process reliability. Experiments across eight LLMs from four model families reveal three key robustness characteristics. First, different classes of scientific perturbations exhibit distinct robustness profiles and can decouple task progression from scientific reliability: agents may continue advancing through the task even after their information or reasoning has become unreliable. Second, stronger clean-task performance does not necessarily translate into stronger robustness, as models with higher clean-task accuracy can exhibit larger degradation under perturbation. Third, perturbation effects exhibit strong temporal dynamics: they may remain latent for multiple turns before emerging and subsequently propagate through downstream dependencies. Together, these findings show that current scientific agents remain insufficiently robust to scientifically plausible perturbations, with failures often remaining undetected, propagating, and resisting recovery. --- ### Papers Without Code: Availability of GitHub Repositories Linked in *CL Publications Authors: Selina Meyer, Michael Roth Published: 2026-09-28 Categories: cs.CL arXiv: https://arxiv.org/abs/2609.34534v1 Source code and data published at computational linguistics (*CL) venues are increasingly being shared via GitHub. While this generally is a favourable development for the accessibility and potential reusability of research artifacts in natural language processing (NLP), the long-term availability of such repositories has not been evaluated. In this squib, we discuss the availability of repositories linked in papers published in the Computational Linguistics (CL) journal as well as at ACL and its co-located events over the past ten years. Contrary to our expectations, we find that GitHub repositories linked in more recent ACL publications are unavailable at similar rates as in older publications, in parts due to an increase in empty and placeholder repositories. Similar trends hold for other *CL venues, but not for platforms other than GitHub. --- ### Preference-Guided Adaptation for Open-Vocabulary Semantic Segmentation via Prompt Disagreement Authors: Hyun-Kurl Jang, Jihun Kim, Kuk-Jin Yoon Published: 2026-09-28 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.34528v1 Open-vocabulary semantic segmentation (OVSS) enables pixel-level prediction over arbitrary text-specified vocabularies and has shown strong generalization on common benchmarks. However, OVSS performance often degrades in specialized domains such as medical imaging, remote sensing, and industrial inspection, where dense pixel-level masks for adaptation are costly to obtain and require domain-specific expertise. We propose a preference-guided adaptation framework that replaces dense mask supervision with binary preferences. We observe that different prompt templates produce systematically different segmentations for the same image, a phenomenon we call prompt disagreement, and we repurpose it as a built-in source of preference supervision. Building on this, we mine localized preference queries from regions of high cross-template uncertainty, and adapt the OVSS model with Region-Localized Preference Optimization (RLPO) together with consistency regularization that stabilizes updates outside the queried region. Across extensive experiments on the MESS benchmark, the proposed method achieves consistent gains across diverse OVSS backbones without any pixel-level annotation, and remains effective under noisy preferences. Our code is available at https://github.com/blue-531/pref-ovss. --- ### RRG-SLAM: Real-time Reflection-aware Gaussian SLAM for Indoor Scenes Authors: Yong Liu, Keyang Ye, Zhexi Peng, Ruixian Mei, Kun Zhou, Tianjia Shao Published: 2026-09-28 Categories: cs.CV, cs.GR arXiv: https://arxiv.org/abs/2609.34527v1 We introduce the first real-time reflection-aware Gaussian SLAM system for indoor scenes. The system features a reflection-aware TSDF-Gaussian hybrid representation that explicitly separates diffuse scene appearance from reflection components. The base scene is modeled by a TSDF volume and a set of base Gaussians capturing geometry and diffuse appearance, while planar reflections are represented by reflection Gaussian groups associated with detected reflective planes. The rendering is performed in three passes: TSDF raycasting first yields surface color, depth, plane IDs and reflection masks; base Gaussians are then rendered order-independently with depth culling and combined with the TSDF output to form the base image; finally, under the guidance of the plane ID map, reflection Gaussians from different reflection groups are rasterized only into their corresponding planar regions to generate the reflection image, which is subsequently composited with the base image via the reflection mask to produce the final output. For online reconstruction, our system first estimates the camera pose through reflection-aware tracking to suppress interference of reflection-dominated regions. It then identifies reflective planes using geometric, semantic, and temporal cues, and fuses the observations into the augmented TSDF volume with reflection-aware attributes. Afterwards the base and reflection Gaussians are initialized, optimized, and pruned online to maintain both reconstruction quality and efficiency. Experiments on a variety of datasets show that our method outperforms existing SLAM systems in reconstruction quality, tracking robustness, and novel-view rendering for indoor environments with reflections, while preserving real-time performance. --- ### PairPref: When Should Memory Guide the Answer? A Benchmark for Contextual Preference Use Authors: Mingfei Lu, Mengjia Wu, Yi Zhang Published: 2026-09-28 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.34526v1 Memory-augmented assistants use retrieved preferences to guide their responses. A small change in the situation can change whether a preference is appropriate while barely affecting its retrieval similarity. Memory benchmarks typically test whether systems store and retrieve preferences, with less attention to when those preferences should apply. We introduce PairPref, a benchmark of contextual preference use. Each pair changes only the situation, keeping the preference, request, and four candidate replies fixed. The preference remains valid in both situations. In the selection track, models must choose the reply that applies the preference only where appropriate. In the free-generation track, they must decide when to apply it without seeing candidate replies. Both tracks use the same 1,227 pairs across 45 preferences and eight situation categories. We evaluate eight models, most of which achieve selection scores ($Δ$) of 51 to 65 points. In free generation, however, both responses are appropriate for their respective situations in only 3.6\% to 18.3\% of pairs. Models continue to apply the preference in both situations even with fewer retrieved memories, alternative presentation formats, and a stricter prompt. These results show that models still struggle to judge when user preferences apply and respond accordingly. --- ### SAGE: Subspace Alignment for Classifier-Free Guidance in Mixture-of-Experts Diffusion Models Authors: Boyu Zhang, Yangming Cheng, Ning Zhang, Pengfei Liu, Weijie Li, Yifan Gao, Hangyu Li, Litong Gong Published: 2026-09-28 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.34525v1 Diffusion Transformers with Mixture-of-Experts (MoE) routing are a leading recipe for scaling generative models. Classifier-Free Guidance (CFG) is essential for generation quality, yet excessively high guidance scales trigger collapse. We identify a previously unreported failure mode in their combination: the two CFG branches route independently, so their realized activations occupy different subspaces. The unconditional write then leaves the conditional subspace, and CFG amplifies that residual linearly in the guidance scale. We propose SAGE, a training-time regularizer that aligns unconditional MoE activations to the conditional subspace without restricting routing diversity, at zero inference cost. Toy experiments show that SAGE dramatically suppresses extreme drift by 9.2x. When scaled to a 1B-parameter text-to-image model, SAGE significantly improves generation quality, delivering a 9.3% boost in peak DPG-Bench performance. Extensive experiments demonstrate that SAGE consistently outperforms the baseline. --- ### Evidence Before Accuracy: A MRI-PET Fusion Network for Alzheimer Disease Classification with Causal Regional Validation Authors: Saeid Firouzi Daghigh, Saeed Ayat Published: 2026-09-28 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.34520v1 Deep learning models for Alzheimer disease (AD) classification routinely report near-perfect discrimination, yet few are shown to rest on AD-relevant neurobiology rather than on dataset artifacts, subject-level leakage, or non-brain image content. We present a fusion network combining T1 MRI and FDG PET across axial, coronal, and sagittal planes, trained on ADNI consists of 554 paired subjects. The fusion model reaches AUC 0.962, accuracy 0.909, and F1 0.891, competitive with recent 3D CNN and multimodal transformer systems at substantially lower cost. We first quantify how much modality, plane and slice geometry matter. A validation-only search over slice centres and neighbour spacings moves AUC by 0.180 for MRI and 0.078 for PET, selecting narrow spacing for MRI and wide spacing for PET, with the chosen coronal centres falling on the hippocampal body and on the posterior cingulate respectively. The contribution, however, is the evidence layer built around that number. Shortcut controls collapse the model to AUC 0.622 (silhouette), 0.608 (exterior), and 0.500 (blank), and a label-permutation null yields 0.456. Forward region-of-interest (ROI) ablation shows that masking medial temporal cortex in MRI and the posterior default-mode network (DMN) in PET produces the largest shift in the AD logit, while area-matched controls remain indistinguishable from that null. Reverse ROI ablation shows that the medial temporal lobe alone retains 89.2% of above-chance discrimination in MRI and the posterior DMN alone retains 79.0% in PET. A quantitative comparison of attribution methods shows occlusion sensitivity reaching 3.5-5.0* enrichment inside a priori AD regions against 0.10-0.43* in controls. Ablation and attribution independently establish a biologically correct double dissociation: hippocampal evidence is carried by MRI, posterior cingulate evidence by PET. --- ### EOPSA: Efficient On-Policy Self-Distilled Safety Alignment Authors: Qirui Liu, Yichen Sun, Yan Wang, Yu Mi, Wei Cao, Yue Shen, Zhixuan Chu, Kui Ren Published: 2026-09-28 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.34519v1 On-Policy Self-Distillation (OPSD) has emerged as a promising paradigm for safety alignment, delivering dense, token-level supervision by distilling from a teacher conditioned on refusal-oriented privileged prompts. However, we reveal that this paradigm suffers from critical inefficiencies that degrade both training efficiency and general reasoning capabilities. Specifically, we diagnose two fundamental bottlenecks: (1) supervisory collapse over extended rollouts, where the teacher's corrective efficacy degrades precipitously as the student's generation prefix lengthens, injecting noisy gradients into late-stage tokens; and (2) gradient dilution from stylistic shifts, where the distillation objective is dominated by safety-irrelevant stylistic discrepancies induced by privileged prompting, washing out genuine safety signals and impairing base reasoning. To resolve these issues, we propose Efficient On-Policy Self-Distilled Safety Alignment (EOPSA), which concentrates computational and gradient budgets exclusively on reliably supervised, safety-critical tokens. EOPSA incorporates two coordinated mechanisms: (i) Adaptive Rollout Scheduling, which dynamically bounds the generation horizon guided by a novel Teacher Rescue Rate (TRR) metric to operate strictly within reliable supervision regimes; and (ii) Selective Distillation, which filters out safety-neutral tokens to restrict gradient updates exclusively to safety-pivotal transitions. Extensive evaluations across reasoning models up to 32B parameters demonstrate that EOPSA slashes rollout computation by $\sim$50% and backpropagates through merely $\sim$2% of tokens, consistently outperforming full-token distillation baselines in both safety compliance and reasoning retention. --- ### SEAD: A State-Based Perspective on Attack and Defense in Tool-Using Agents Authors: Xinjie Shen, Junran Wang, Rongzhe Wei, Pan Li Published: 2026-09-28 Categories: cs.CR, cs.AI arXiv: https://arxiv.org/abs/2609.34518v1 Language-model agents increasingly use tools to act on external systems. Earlier actions can alter files, permissions, database records, or other state, making a later routine-looking action harmful. Yet the visible interaction may not reveal the underlying state needed to assess that action. We formulate attack and defense as partially observed state control in SEAD, deriving their design requirements from this shared execution process. Because attackers supply instructions while the target chooses concrete actions, DART decomposes harmful goals into locally plausible steps and uses feedback from actual tool execution to guide trajectory search. The defender must decide before execution with incomplete state evidence. SAGE can therefore investigate relevant state through read-only queries before allowing or blocking each action, including those proposed after a block. We construct an environment-verifiable dataset integrating controlled initial states, replayable tool environments, and task-specific executable checks. Across four target models, DART improves semantic attack success by 18.8--35.9 percentage points over the competing baseline, with consistent gains under executable verification. On recorded trajectories, SAGE preserves 95.79% of benign trajectories while intercepting 92.73% of harmful paths by the harm-enabling boundary. In online attack-defense evaluation, it reduces DART's executable attack success from 48.0% to 4.0%. SAGE remains effective across four attack methods and generalizes to out-of-domain environments. Our code and data is available at https://github.com/EverywhereSafety/SEAD. --- ### How to Tame a Multi-Headed Hydra? Adaptive Multi-Category Safety Steering for Large Language Models Authors: Chenxi Wang, Ruiyang Huang, Li Huang, Yifan Wu Published: 2026-09-28 Categories: cs.CR, cs.CL, cs.LG arXiv: https://arxiv.org/abs/2609.34514v2 As large language models (LLMs) become increasingly widespread, preventing unsafe responses to harmful prompts is essential for their safe deployment. Activation steering offers an approach to improving LLM safety by modifying internal activations during inference without updating model parameters. However, a single prompt can involve multiple harm categories, and steering toward safety in one category may leave harmful content from another unaddressed. Despite advances in adaptive steering, existing methods do not explicitly coordinate steering direction and strength when multiple harm categories co-occur within a single prompt. To address this problem, we propose CAM-Steer, a Category-Adaptive Multi-category Safety Steering framework. Specifically, it estimates the risk associated with each harm category by comparing the current hidden state with safe and unsafe prototypes. The estimated risks are then used to combine the safety directions for different harm categories into a single steering direction and to determine the strength of the intervention. Finally, it rotates the hidden state along the composed steering direction, with the rotation angle determined by the estimated risks, while preserving the hidden-state norm. Experiments across three LLM backbones and seven harm categories show that CAM-Steer outperforms the evaluated baselines in average defense success rate, including when categories co-occur. Further analyses support its component designs and informative risk scores, with negligible inference overhead. --- ### XFlow: A Workflow Model for Instruction-Guided Lesion Segmentation in Chest X-rays Authors: Geon Choi, Hangyul Yoon, Hyunki Park, Sang Hoon Seo, Edward Choi Published: 2026-09-28 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.34513v1 Existing text-guided segmentation models in the medical domain cover only a narrow set of anatomical structures and lesions in chest X-rays (CXRs), and most of them assume that the queried target is always present in the image. Instruction-guided lesion segmentation (ILS) was introduced to overcome these limitations by segmenting diverse lesion types from simple user instructions while also recognizing when the queried lesion is absent, and ROSALIA was proposed as the first model for this task. However, the masks produced by ROSALIA remain of limited quality, often carrying scattered noise. Moreover, ROSALIA predicts the mask in a single shot, which differs fundamentally from how radiologists perceive and delineate lesions in practice. A radiologist first surveys the entire thorax, then localizes the approximate region of abnormality, and only then refines the lesion contour. Motivated by this coarse-to-fine, multi-level perception process, we present XFlow, a workflow model for ILS that combines box-based localization with multi-turn point refinement. XFlow detects the lungs, decides whether the queried finding is present in each of them, and prompts a fine-tuned SAM with the lesion box for an initial mask. It then corrects that mask through point prompts until its boundary follows the lesion, leaving every intermediate decision visible. Our experiments show that XFlow achieves the best segmentation quality on both internal and external evaluation. Notably, it surpasses ROSALIA in segmentation quality even when the two are trained on the same lesion annotations. Code and model weights will be made publicly available. --- ### HALO: Enhancing Time Series Generation via Hyperspherical Latents and Masked AutoregRessive Modeling Authors: Chunyi Hou, Xiangfei Qiu, Hanyin Cheng, Yutong Li, Bin Yang Published: 2026-09-28 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.34511v2 Most existing time series generators rely on a two-stage modeling paradigm: the first stage learns discrete latent representations of time series; the second stage performs autoregressive modeling on these discrete latents through next token prediction. However, this paradigm suffers from two stage-specific limitations: the first stage can lead to information loss when discretizing continuous time series, while the second stage is prone to error accumulation during autoregressive generation. To address these limitations, our core idea is to perform generative modeling in a continuous latent space with a more efficient autoregressive framework. We propose HALO, which enhances time series generation via Hyperspherical Latents and Masked Autoregressive modeling to achieve this goal by tackling two key bottlenecks: (1) variance and scale heterogeneity of continuous latent representations; (2) the difficulty of balancing generation efficiency with temporal correlation modeling. HALO first introduces a hyperspherical VAE that constrains continuous latents to a fixed-radius hyperspherical shell, effectively stabilizing the numerical fluctuations of continuous latent representations. Secondly, we develop a masked autoregressive model that balances parallel decoding and temporal correlation learning, substantially reducing the number of inference steps required for generation and improving generation stability. Our extensive experiments demonstrate that HALO achieves state-of-the-art generation performance while offering significantly improved inference efficiency over existing advanced baselines. --- ### Can AI Make Money in Crypto? Measuring the Gap from Backtests to Real Markets Authors: Xingtong Yu, Jiarun Zhou, Guanlin Ding, Wenkang Wei, Jiarui Liu, Chang Zhou, Fangzhou Ge, Chenyi Xu, Xikun Zhang, Renqiang Luo, Jie Zhang, Hong Cheng et al. Published: 2026-09-28 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.34510v1 AI-based trading methods have rapidly evolved from machine learning and reinforcement learning to large language models (LLMs) and trading agents, yet their performance is still predominantly assessed through historical backtesting. Such evaluations provide limited evidence of whether a method can generalize to unseen future markets or whether its backtested performance can be sustained in realistic trading frictions (e.g., latency, slippage, liquidity constraints, and market impact). We present a unified benchmark that evaluates representative machine learning, reinforcement learning, LLM-based, and agent-based trading methods in cryptocurrency markets through three progressively more realistic stages: historical backtesting, prospective exchange-based paper trading, and real-money live trading. These stages jointly increase temporal realism by moving from historical to unseen future markets, and execution realism by moving from offline simulation toward live trading. This protocol enables us to quantify the backtest-to-realization gap, identify when performance begins to deteriorate, and compare how this gap differs across major classes of AI trading methods. We further provide a unified open-source system supporting all three evaluation stages, together with a public platform that continuously updates benchmark results. Code is available at https://github.com/Starlien95/Awesome-TradingAI. --- ### Low-Confidence Remasking Traps Flexibility: Realizing Arbitrary-Order Potential for Diverse Rollouts in Diffusion LLMs Authors: Moongyu Jeon, Dongjae Jeon, Bumjun Kim, Mingyu Kim, Albert No Published: 2026-09-28 Categories: cs.LG, cs.CL arXiv: https://arxiv.org/abs/2609.34509v1 Masked diffusion language models support arbitrary-order generation, suggesting a natural way to produce diverse outputs. However, recent work argues that this flexibility reduces diversity by delaying high-uncertainty tokens that can lead to different generation paths. We trace this diversity loss not to arbitrary-order generation itself, but largely to low-confidence remasking (LCR), a widely used decoding rule. At each step, LCR samples a token at every masked position but commits only the sampled token with the highest probability, filtering out the rest. We show that this mechanism can exponentially suppress lower-probability tokens as more positions compete, and observe the same suppression in LLaDA. In contrast, top-probability position selection (TPP), which has often been conflated with LCR under the shared label confidence-based decoding, avoids this diversity loss. TPP first selects the position whose most likely token has the highest probability, then samples directly from that position's distribution. Replacing LCR with TPP restores diversity and yields Pass@$k$ comparable to left-to-right decoding, suggesting that the reported diversity loss stems largely from LCR's filtering rather than from generating high-confidence positions first. To further exploit order flexibility, we introduce Entropy-Guided Initialization (EGI), which samples the first token at the highest-entropy position and then follows TPP. This simple modification further improves rollout diversity and solution coverage beyond left-to-right decoding, with gains extending to downstream policy optimization, highlighting the potential of arbitrary-order generation for diverse rollouts. --- ### KiT: A Foundation Model for Financial Time-Series Forecasting using DiffusionTransformers Authors: Boyu Zhang, Haorui Li Published: 2026-09-28 Categories: cs.LG, cs.CV arXiv: https://arxiv.org/abs/2609.34507v1 Financial candlestick forecasting is fundamental to quantitative investment, yet it remains exceptionally challenging due to extremely low signal-to-noise ratios and vast heterogeneity across markets and instruments. Existing approaches have largely attempted to introduce deep learning to capture hidden temporal features, but most adopt an auto-regressive formulation, which leads to error accumulation during inference. Meanwhile, general-purpose time-series foundation models are not tailored to the unique structure of k-line data and yield unsatisfactory performance on downstream candlestick forecasting tasks. To tackle these problems, we introduce KiT, a K-line Diffusion Transformer foundation model, and reformulate future prediction as conditional path generation via flow matching: given a historical context window, the model generates an ensemble of plausible future OHLCV trajectories. We pre-train KiT at multiple parameter scales on billions of candlestick bars spanning multiple markets and timescales. Across three markets and seven resolutions, KiT attains a mean return RankIC of 0.057 and a mean volatility RankIC of 0.66, leading at every timescale and outperforming both task-specific financial forecasters and general time-series foundation models. Code will be available at: https://github.com/Luciferbobo/KiT. --- ### Does Model Uncertainty Track Human Ambiguity? Evidence from Multi-Annotator Vision Benchmarks Authors: Manya Singh, Arjun Pakrashi Published: 2026-09-28 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.34506v1 Human-model alignment is critical for trustworthy AI-assisted decision-making systems. Yet, most work evaluates model predictions against single ground-truth labels, overlooking that humans themselves often disagree on labels, a signal of genuine ambiguity. We investigate whether models struggle on the same instances that humans find difficult. We measure this on two vision datasets (FER+ and CIFAR-10H) where multiple human annotations per image capture human disagreement patterns. We evaluate eight pretrained models across three architectures (ResNet, EfficientNet, MobileNetV3) in two parts: first, whether model uncertainty (softmax confidence, entropy) correlates with human disagreement, and second, whether predictive multiplicity measures (inter-model disagreement, Jensen-Shannon divergence) do. We find that it does not: alignment is weak in both dimensions. At the discrete label level, 50.4% of CIFAR-10H images and 33.5% of FER+ images receive multiple valid classifications from humans, while the models converge on only one. These instances represent a critical failure case where humans perceive ambiguity and would request expert review, yet models decide confidently. At the continuous score level, single-model uncertainty correlates weakly with human disagreement ($ρ= 0.24--0.55$), and predictive multiplicity provides only modest improvement. Widely-used uncertainty quantification methods do not reliably identify instances humans find ambiguous. Model uncertainty should not be treated as a trustworthy signal by default for decision-making in high-stakes scenarios. --- ### Distribution-Conditioned Task Routing for Class-Incremental Learning Authors: Longhuan Xu, Zhipeng Zhou, Wei Ji, Chunyan Miao, Peilin Zhao, Lijun Zhang Published: 2026-09-28 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.34503v1 Parameter-efficient adaptation enables continual learners to acquire task-specific knowledge through compact model updates while maintaining strong within-task performance. However, class-incremental inference requires each input to be classified among all classes seen so far without access to its task identity. For learners equipped with task-specific parameter-efficient modules, this introduces a critical task-routing challenge beyond catastrophic forgetting. We study post-hoc task routing without retraining the learner or introducing a separately trained router. Such training-free inference-time calibration remains comparatively underexplored in parameter-efficient class-incremental learning. We identify three sources of routing error (feature-level, task-level, and class-level misalignment) and propose Feature Distribution Calibration (FDC). Its three components address these misalignments: Task Subspace Filtering (TSF) suppresses feature components outside each task's principal subspace, Residual Likelihood Calibration (RLC) evaluates the typicality of its subspace residual, and Prototype Affinity Calibration (PAC) measures compatibility with the task's class prototypes. Experiments demonstrate plug-and-play applicability to eight parameter-efficient class-incremental methods using a shared encoder. With one component configuration selected per method across all five benchmarks, FDC improves final accuracy in all 40 method-dataset pairs by 4.39 percentage points on average. Enabling all components improves 35 of the 40 pairs, with an average gain of 4.45 points. When applied to a simple baseline, FDC achieves strong overall performance. --- ### SubjectAnchor: Subject-Aware Memory-to-Video for Multi-Shot Storytelling Authors: Xinyu Wang, Huafeng Shi, Zian Li, Yan Zhou, Xiaoqiang Liu, Yue Ma, Pengfei Wan Published: 2026-09-28 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.34502v1 We present SubjectAnchor, a Subject-Aware Memory-to-Video paradigm for multi-shot storytelling in which the current shot is generated by conditioning on explicit visual memories extracted from previous shots. The objective is to preserve subject identity and scene consistency across cuts while retaining the controllability of shot-wise prompting. Built on Wan2.2-I2V-A14B, SubjectAnchor contains three key components: subject-related memory construction, subject-aware temporal rotary position encoding, and memory-aware attention partition. For each target shot, the method constructs a compact memory bank by tracing each required subject to its historical appearance and retrieving the most relevant precomputed keyframes. These memory frames are encoded into the model input as explicit visual conditions, while different subjects are assigned to separated negative temporal slots to reduce identity interference. In addition, memory-aware attention partition regulates the interaction between memory tokens and generated content within a shared backbone. This formulation preserves the appearance anchoring of explicit visual memory while remaining compatible with script-driven shot-by-shot generation. Experiments show that SubjectAnchor improves cross-shot identity consistency over representative memory-based and holistic baselines while maintaining competitive visual quality. --- ### Can Attack Difficulty Be Characterized Before Optimization? A Study of Pre-optimization Difficulty in Person-Vanishing Attacks Authors: Jingyao Xu, Dongdong Wang, Siyang Lu Published: 2026-09-28 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.34501v1 Adversarial attacks against object detectors are traditionally studied from an optimization perspective, where attack difficulty is regarded as an outcome observed only after adversarial optimization. This raises a fundamental question: \emph{can the relative attack difficulty of different inputs be characterized before optimization begins?} In this paper, we investigate this question for person-vanishing attacks by introducing the concept of pre-optimization attack difficulty, which captures intrinsic differences in optimization effort across input images. To estimate this latent difficulty before optimization, we propose Quad-CLEVER, an efficient geometry-based estimator derived from a quadratic approximation of the local person-vanishing margin along the most attack-relevant direction. Extensive experiments across multiple attack algorithms demonstrate that Quad-CLEVER consistently correlates with the observed optimization cost, providing empirical evidence that attack difficulty exhibits a predictable pre-optimization structure. Building upon this finding, we further propose a difficulty-aware attack framework that leverages the estimated difficulty to adaptively allocate optimization budgets for a base attack under a fixed computational budget. On BDD100K, the proposed framework improves the image-level attack success rate by up to 5.78$\%$ while reducing the average optimization cost by up to 11.42 iterations. On the more challenging EventPed dataset, it saves 2.25 optimization iterations while maintaining comparable attack performance. These results demonstrate that attack difficulty can be meaningfully estimated before optimization and that exploiting such estimates enables more computationally efficient adversarial attacks. --- ### Scalable GNN-based Knowledge Graph Representation Learning with Efficient Message Passing Authors: Huu Tan Mai, Cuong Xuan Chu, Heiko Paulheim, Daria Stepanova Published: 2026-09-28 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2609.34499v1 Graph neural networks (GNNs) excel at representation learning on Knowledge Graphs (KGs), achieving stateof-the-art performance on tasks like link prediction or entity classification. However, their high computational complexity, inherent to their user-defined message passing (MP) algorithm, still prohibits their widespread adoption, especially for large KGs. Current efforts to mitigate the scalability bottlenecks of GNNs on KGs, such as subgraph sampling, are often task- and model-specific, and do not reliably guarantee lossless (if applicable) runtime/space reductions. To address this, we extend Relational Sparse Matrix Multiplication (RSPMM), originally designed to losslessly lower the space complexity of composition-based MP with pointwise composition functions, to support more expressive functions (e.g., 2x2 block-diagonal matrix multiplication, Givens rotation, circular correlation). Our method delivers significant task-independent reductions in runtime and space for current GNNs on KGs and facilitates efficient re-implementations of GNNs that maintain near state-of-the-art performance on challenging KG tasks, for a fraction of computational costs. --- ### QAMM: Adjoint MeanFlow Matching for Few-Step Offline Reinforcement Learning Authors: Yuehu Gong, Shutong Ding, Mokai Pan, Yimiao Zhou, Jiashu Hou, Ye Shi, Yanwei Fu Published: 2026-09-28 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.34497v1 Flow policies can model rich action distributions, but their iterative sampling limits decision speed. Adjoint matching uses the critic's action gradient to improve a flow policy without backpropagating through its sampling trajectory, yet its supervision is defined for instantaneous velocities. We propose QAMM, a method that turns the critic-derived adjoint signal into supervision for MeanFlow's average velocity. The resulting policy learns finite-interval transport directly and generates actions with few network evaluations. We derive the adjoint MeanFlow target, specify its gradient boundaries, and train it with an offline actor-critic. On ten HumanoidMaze tasks, QAMM produces effective two-call policies and achieves competitive performance against strong flow-policy baselines. These results show that adjoint-based Q optimization can be combined with average-velocity learning to obtain expressive offline policies with few-step action generation. --- ### ConCAD: Constraint-Aware Image-to-CAD Generation with Dual-Granularity Rewards Authors: Chenxi Zhai, Xi Cheng, Hang Cheng, Zhicheng Guan, Mingyu Fan, Yanzhe Tang, Pingfa Feng, Long Zeng Published: 2026-09-28 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.34494v1 Image-to-CAD generation seeks executable parametric programs that recover both the geometry and design intent of a reference object. Existing systems are commonly evaluated by validity and shape overlap, although two solids with similar volume can encode different CAD relations. We introduce ConCAD, a constraint-aware image-to-CAD framework optimized via Group Relative Policy Optimization (GRPO) with rewards at two complementary granularities: a code-level constraint reward and an execution-level geometric reward. This complementary design disambiguates structurally distinct yet volumetrically similar shapes while ensuring valid 3D geometry. To verify that these rewards recover geometry and design intent, we introduce a B-rep geometric constraint satisfaction rate (G-CSR), which analytically extracts and evaluates geometric constraints from boundary representations. Experiments on the DeepCAD and Zero2CAD demonstrate that ConCAD achieves the best IoU and Chamfer Distance over competitive baselines, while also outperforming them on G-CSR, validating its superior recovery of both geometric fidelity and parametric design intent. ---