# Research Papers — 2026-09 · bundle page 26 of 266 > Full-text markdown of papers #1251–1300 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/26 [← Previous bundle](https://aipod.dev/papers/archive/2026-09/page/25.md) [Next bundle →](https://aipod.dev/papers/archive/2026-09/page/27.md) ### RED: Reconstruction Evolution Dynamics for Generalizable AI-Generated Image Detection Authors: Wenpeng Mu, Junshan Jin, Tanfeng Sun, Xinghao Jiang, Qiang Xu Published: 2026-09-29 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.36822v1 The rapid evolution of image generators calls for forensic cues that generalize beyond known generation mechanisms. Existing detectors often rely on static image representations or endpoint reconstruction discrepancies, leaving the evolution of intermediate reconstruction stages underexplored. We observe that the relative token predictability of real and generated images can reverse across reconstruction scales, suggesting that intermediate stages may expose forensic evidence overlooked by endpoint comparisons. Motivated by this observation, we propose RED (Reconstruction Evolution Dynamics), a framework that captures transferable forensic cues from coarse-to-fine reconstruction evolution. To our knowledge, RED is the first framework to use scale-wise token predictability to guide forensic evidence aggregation across intermediate reconstruction states. It represents the reconstruction trajectory produced by a frozen multiscale VQ-VAE in the shared feature space of a frozen CLIP encoder. To connect the observed predictability variations with visual evidence, RED learns image-adaptive stage weights from scale-wise token negative log-likelihoods provided by a frozen VAR model. A cross-stage evidence aggregation module then jointly models the original-image representation and the weighted reconstruction features, capturing complementary forensic cues through interactions along the reconstruction trajectory. Experiments on six diverse benchmarks demonstrate that RED achieves the highest average accuracy of 92.5\% and average precision of 97.5\% among the evaluated methods. Further evaluations show strong robustness to common image degradations, supporting the value of reconstruction evolution for generalizable AI-generated image detection. The code will be made publicly available upon acceptance of this paper. --- ### CorrGRPO: Correlation-Normalized GRPO for Multi-Reward Learning Authors: Wenbin Hu, Huihao Jing, Haochen Shi, Yuxuan Liu, Haoran Li, Yangqiu Song Published: 2026-09-29 Categories: cs.LG, cs.CL arXiv: https://arxiv.org/abs/2609.36820v1 Group Relative Policy Optimization (GRPO) is widely used to train reasoning language models, where it computes advantages by centering and normalizing rewards across rollouts of the same prompt. For multiple rewards, GRPO sums the reward components and normalizes the total reward by its within-group standard deviation. The corresponding variance equals the sum of all pairwise reward covariances. For a fixed centered reward, larger aggregate covariance produces smaller advantages, and vice versa, allowing update magnitudes to adapt to reward dependence. However, correlated rewards with large scales can dominate this normalization and suppress signals from smaller-scale rewards. We propose Correlation-Normalized GRPO (CorrGRPO), which normalizes pairwise covariances into Pearson correlation coefficients. CorrGRPO keeps the centered total reward unchanged while balancing the influence of differently scaled rewards on the correlation-based normalization. This allows advantage magnitudes to adapt to reward correlations without the normalization being dominated by large-scale reward components. We compare CorrGRPO with GRPO and other variants on code generation, tool calling, and agent security, using models ranging from 0.5B to 8B parameters. These tasks all involve multiple rewards that can improve together or present tradeoffs. Results show improvements across three domains, including code generation, tool calling, and agent security. Our code is available at https://github.com/HKUST-KnowComp/CorrGRPO. --- ### Towards Better Training Signal: Advantage Clipped Policy Optimization Authors: Ruichuan Huang, Jinghan Liu, Congliang Chen Published: 2026-09-29 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.36816v1 Reinforcement learning (RL) has become a cornerstone for improving the reasoning capabilities of large language models (LLMs), but the need for on-policy data substantially limits training efficiency. Reusing off-policy data through importance sampling (IS) can improve efficiency but introduce considerable instability. Hence, algorithms such as PPO and GRPO widely adopt IS-ratio clipping to stabilize training. However, training stability and gradient estimate are mainly determined by the product of IS ratio and advantage. To further stabilize training, we propose ACPO, which clips the product of the IS ratio and the advantage, leading to more stable gradient estimates. We also establish a connection between ACPO and gradient clipping in policy mirror descent (PMD), which is a standard technique to stabilize optimization process, and prove the convergence of clipped-PMD under the standard RL setting. Experiments on widely used mathematical reasoning benchmarks show that ACPO consistently outperforms PPO and GRPO in both accuracy and training efficiency, delivering 4-6 percentage points gains on standard math benchmarks, with Qwen3-8B+PPO. Hence, ACPO is a practical and effective alternative to conventional IS-ratio clipping for RL post-training of LLMs. --- ### CurvSpec: Adaptive Multi-Curvature Learning for Partial Relevant Video Retrieval Authors: Zhen Liu, Letian Li, Jinpeng Wang, Shuzhao Xie, Yuzhi Huang, Jingyan Jiang, Zhi Wang Published: 2026-09-29 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.36815v1 Partially Relevant Video Retrieval (PRVR) seeks to retrieve untrim-med videos containing a moment that matches a text query, without temporal annotations. The relevant moment may last only seconds within a video spanning several minutes, creating an extremely low signal-to-noise ratio that makes PRVR more challenging than standard full-video retrieval. This task presents two intertwined challenges: (1) signal dilution, where coarse global representations blur the brief relevant signal into the dominant irrelevant surroundings;(2) curvature rigidity, where embedding all videos in the same fixed-geometry space distorts representations for videos that range from flat atomic events to deep compositional hierarchies. Existing PRVR methods have improved moment selection and cross-modal matching, but they still typically encode all videos in a single fixed-curvature retrieval space, limiting their ability to model diverse video structures. To address both challenges, we propose CurvSpec, a framework that learns content-adaptive curvature for video retrieval representations rather than imposing a fixed geometric prior. CurvSpec processes features through parallel Euclidean and hyperbolic attention layers, with independently learned curvatures assigned to the hyperbolic layers, and a content-aware fusion mechanism routes each input to its most suitable geometric regime. To further suppress signal dilution, CurvSpec represents each video with semantic centroids whose number is determined by the video's content complexity, projects them onto the learned manifold, and matches each query against its nearest centroid by geodesic distance. Experiments on ActivityNet Captions, TVR, and Charades-STA demonstrate state-of-the-art retrieval performance. --- ### RESCUE: Repairing Language Model Errors to Sparse Circuits via Reinforcement Learning Authors: Chuanpu Liu, Miao Yu, Yikai Cai, Yuanhe Zhang, Zhenhong Zhou, Li Sun, Zuming Jiang, Yufei Guo Published: 2026-09-29 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.36813v1 Large language models (LLMs) exhibit strong general capabilities that mechanistic interpretability has attributed to sparse computational circuits. However, existing circuit studies emphasize preserving functionality or explaining safety, leaving the mechanisms underlying failures across a broader range of tasks largely unexplored. Extending circuit analysis from abilities to errors, we explore the perspective that such failures may likewise arise from erroneous internal computations and that targeted tuning of the corresponding parameters can correct such errors while largely preserving other capabilities. Motivated by this insight, we introduce RESCUE (Reasoning-Error Sparse-Circuit Uncovering and Editing), a framework that localizes error-associated circuits and surgically repairs them for performance enhancement. General tasks typically involve multi-step reasoning and long-form generation, where early deviations can cause prefixes to drift from supervised references, leading SFT-based mask optimization to overlook circuits involved in generation-time errors. RESCUE therefore refines these masks through reinforcement learning with multiple masked-model rollouts, improving their relevance to observed task failures. Finally, RESCUE introduces a pruning technique and precisely fine-tunes error circuits to correct task failures, thereby translating error localization into a sparse and targeted model update. We validate RESCUE on heterogeneous repair sets across two domains: (1) mathematical reasoning, identifying a math error circuit of 1.40% density whose repair raises accuracy from 6.0% to 75.5%; and (2) medical QA, where a similarly compact 1.44% circuit improves repair-set accuracy from 0% to 81%. Our code is available at: https://github.com/chuanpupig/RESCUE. --- ### Diffusion Policy Improvement with Proposal-Conditioned Refinement Flows Authors: Junhyun Ha, Juho Lee, Byungwoo Park Published: 2026-09-29 Categories: cs.LG, cs.AI, cs.RO arXiv: https://arxiv.org/abs/2609.36812v1 Diffusion and flow policies can model complex behaviors in offline reinforcement learning (RL). However, penalizing their KL divergence from the behavior policy can discourage actions having high critic values with low behavior density. Directly refining behavior proposals may be an alternative, yet Gaussian or deterministic editors limit expressiveness to represent multiple separated modes for the same proposal. In this work, we introduce Proposal-Conditioned Refinement Flows (PReFlow), a policy extraction method combining critic-based proposal selection with a conditional refinement flow. To optimize proposal selection and refinement together, we formulate a KL-regularized objective whose optimum induces a Gibbs policy over final actions under a Gaussian-smoothed behavior prior. The refinement flow can represent multiple high value modes, while a proposal-centered Gaussian reference regulates large action changes. This Gaussian reference further enables us to make use of simulation-free, closed form adjoint matching targets from sampled endpoints and critic gradients, yielding a single velocity regression loss without a backward adjoint solve. On 50 OGBench tasks, PReFlow achieves competitive offline performance and the highest aggregate score among the compared methods after online fine-tuning, reaching 91\% after 500K environment steps. --- ### MeteoVerse: Unified Weather-Controllable Video World Model Authors: Renlong Wu, Guanqiao Wang, Xuan Shang, Yin Hanming, Xiaoxiao Sheng, Tianyu Huang, Hui Li, Wangmeng Zuo Published: 2026-09-29 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.36810v1 Video world models aim to predict future content from an observed scene while following prescribed camera motion. Real-world scene evolution is determined not only by changes in viewpoint and object dynamics, but also by environmental conditions such as weather, which can substantially alter scene appearance and visibility. Modeling such realistic weather evolution is challenging because the required weather modification depends jointly on the observed and desired weather states. Depending on their relation, the model may need to preserve, introduce, or remove a weather effect. Existing video world models typically leave this weather transition implicit, forcing the generation backbone to infer weather evolution together with scene dynamics and camera motion, which leads to imprecise weather control. To address this limitation, we propose MeteoVerse, a unified weather-controllable video world model that generates future videos from a single sunny or adverse-weather image, conditioned on a weather-free scene description, a target-weather instruction, and a camera trajectory. Rather than conditioning only on the desired weather, MeteoVerse explicitly estimates the observed and target weather states and represents the required weather transition. A transition-aware mixture of weather experts then translates this transition into category-specific residual weather features, unifying weather preservation, introduction, and removal while enabling fine-grained control over introduced weather intensity. We further construct the MeteoVerse dataset with over 50K real-world weather video clips, generated sunny counterparts, disentangled scene and weather descriptions, weather-intensity annotations, and camera trajectories. Extensive experiments demonstrate substantially improved weather controllability while retaining competitive scene consistency and camera-control performance. --- ### Geometry-Conditioned Fixed-Scaffold Encoders for Time-Warp Robust Sequence Retrieval Authors: Cassandra Yang, Yufan Tang Published: 2026-09-29 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.36809v1 Embedding-based retrieval is attractive for long sequence collections because each item can be encoded once and searched by nearest-neighbor ranking. The difficulty is that the objects being indexed are often observed under a noncanonical clock: cardiac cycles stretch with rate, speech changes with tempo, and sensor traces reach comparable states at different speeds. This paper studies a specific source of instability in patch-based encoders for this regime. If patch boundaries are chosen from signal geometry, then the tokenization can change under the same temporal deformation that the representation is expected to tolerate. We propose GeoPatch, a fixed-scaffold patch encoder that keeps token support independent of geometry and uses slope, curvature, acceleration, affine-residual, and confidence descriptors only as continuous conditioning variables. The design turns boundary variation into feature modulation: geometry can change the embedding through a controlled pathway, but it cannot change the number, order, or support of local tokens. We formalize this distinction through a mechanism-level stability analysis that separates boundary drift, affine timing variation, confidence-weighted geometry perturbation, and retrieval-margin effects. The same local tokens support global embedding retrieval and late-interaction scoring, so the scoring rule can be matched to the evaluation protocol. Across ECG, speech, and multivariate time-series retrieval tasks, GeoPatch improves early-rank retrieval under timing variation while exposing a clear trade-off between local surface matching and strict non-overlap retrieval. --- ### Spotter: Let the Embodied Model Lead, and the VLM Reflect for It Authors: Long Li, Qichao Zhao, Yue Yang, Fan Xu, Zhe Wang, Alan Wee-Chung Liew, Chao Qu, Heng Tao Shen, Shirui Pan Published: 2026-09-29 Categories: cs.RO, cs.AI arXiv: https://arxiv.org/abs/2609.36808v1 Current embodied models do not respond to their own failures, although what just went wrong could inform a small adjustment on the next attempt, the kind of reflection behind the gains of thinking in language models. We test whether they can repair a known error, which requires producing a correction and judging whether it is right. Stopped at a failure and allowed to retry, they seldom repair it through their own randomness or from a language description of the error, and best-of-N selection cannot pick the successful candidate after a failure. We attribute this to training only on successful demonstrations and to inputs too narrow to show what went wrong, and conclude that reflection must come from a vision-language model (VLM), which takes in far more information, such as the episode history and text, and is more general. Prior VLM-led work has the VLM plan every step and invoke the embodied model as a tool, placing the VLM on the critical path. We propose Spotter, which reverses the roles: the embodied model leads and executes continuously, while the VLM runs in parallel, monitors through a lightweight local screener, intervenes only when an error is detected, reflects on and corrects it, and returns control. We run Spotter with Qwen and with GPT as the VLM, and both improve the embodied models; with GPT, Spotter improves Cosmos Policy and $π_{0.5}$ by 5.6 and 7.5 percentage points on RoboCasa, and raises $π_{0.5}$ from 47.2% to 57.0% on the Hard setting of RoboTwin 2.0 and from 53% to 83% on a real robot. Because the VLM steps in only when an error is confirmed, a successful episode with Qwen takes only 13 to 16 s longer than with the embodied model alone and about 70% less time than with a VLM-led baseline using the same model. Our code is available at https://github.com/zqc3117/Spotter. --- ### CAD-Native Transformer Operators for AI-Aided Engineering Authors: Daniel Leibovici, Nikola Borislavov Kovachki, Dawon Ahn, Ruben Ohana, Ira J. S. Shokar, Abouzar Ghasemi, Semih Akkurt, Rishikesh Ranade, Neil Ashton, Jan Kautz, Jean Kossaifi Published: 2026-09-29 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.36806v1 Modern engineering systems, from automobiles to aircraft, are designed by using precise, continuous parametric computer-aided design (CAD) models. Evaluating design changes through numerical simulation requires meshing the continuous geometry, a computationally expensive and often brittle process that can require manual intervention and replaces the continuous representation with a discrete approximation. Most neural surrogates accelerate the simulation, but inherit this representation gap by relying on meshes, point clouds, voxels, or other sampled approximations of geometry. We introduce CANTO, a transformer neural operator that maps directly from continuous CAD geometry to physical fields, without meshing the input geometry. We develop a theoretical framework for learning operators from geometric manifolds to function spaces of physical fields, representing geometry through sequences of parametric patches. CANTO instantiates this framework by directly tokenizing non-uniform rational B-spline (NURBS) patches from their control points, knot vectors, and weights, and predicts continuous surface and volume fields at arbitrary query locations. We evaluate CANTO on four automotive and aircraft aerodynamics industry benchmarks: AhmedML, WindsorML, DrivAerML, and HiLiftAeroML. CANTO achieves state-of-the-art accuracy on most evaluated surface and volume prediction tasks, including a 19.8% reduction in surface-pressure relative $L_2$ error compared with AB-UPT on HiLiftAeroML. Differentiability with respect to CAD parameters further enables gradient-based inverse design of designs. On AhmedML, CANTO identifies designs with 4.4 to 20.4% lower drag than the best dataset designs satisfying the same volume and lift constraints, with the improvements verified using the same CFD setup used to generate the original dataset. --- ### UpliftMem: Learning Set-Level Uplift for Agent Memory Retrieval Authors: Mengkun Liang, Haoran Qiang, Guannan Liu, Junjie Wu Published: 2026-09-29 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.36805v1 Large language model (LLM) agents reuse external memory to guide new tasks, but effective retrieval requires learning which memory sets improve execution. Such learning relies on costly outcome feedback: ordinary retrieval observes only executed sets, while evaluating alternatives requires additional rollouts. We introduce \textsc{UpliftMem}, which learns memory retrieval from set-level execution uplift relative to the same executor without memory. A theoretical analysis of how retrieval preferences restrict feedback coverage motivates targeted probing of alternative memory sets. Probe selection follows an expected value of sample information (EVSI) criterion, derived in closed form under a correlated Gaussian model, to allocate limited training rollouts according to their expected improvement in local retrieval decisions. The shared scorer is trained with a frozen executor and selects memory sets without test-time probes. Across ALFWorld, WebShop, and BigCodeBench, \textsc{UpliftMem} achieves the best success rates among evaluated baselines on the main evaluation sets. Controlled fixed-store and matched probe budget evaluations further demonstrate improved memory-use decisions and more effective use of execution feedback. --- ### VAA-CSEC: Vote-guided Advantage Allocation for Chinese Semantic Error Correction Authors: Yitong Han, Nankai Lin, Juan Luo, Hongyan Wu, Lianxi Wang, Shengyi Jiang Published: 2026-09-29 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2609.36804v1 Chinese Semantic Error Correction (CSEC) targets semantic errors in Chinese text, which are typically more subtle and complex than spelling and grammatical errors but remain relatively underexplored. Existing LLM-based approaches face two recurring obstacles in this task: over-correction, and unclear interaction between Chain-of-Thought (CoT) reasoning and self-consistency decoding, such that the benefits brought by CoT cannot be reliably transferred to final corrections. We propose Vote-guided Advantage Allocation for CSEC (VAA-CSEC), a multi-stage framework that combines CoT distillation, Supervised Fine-Tuning (SFT), Reinforcement Learning (RL) and self-consistency decoding. During RL, we design a task-specific reward function that directly aligned with the minimal-editing principle of CSEC. We further introduce Group-Level Relative Policy Optimization (GLPO), which reallocates GRPO advantages according to the margin between individual rollout rewards and the vote-aggregated group reward, aligning the RL training objective with the self-consistency objective used at inference time. Experiments on CSED-C and NaSGEC-Exam show that VAA-CSEC outperforms all LLM-based baselines on CSED-C with an F0.5 of 47.72%, achieves the highest recall of 42.15% among all methods, and establishes a new state of the art of 41.55% F0.5 on NaSGEC-Exam. --- ### EGSD: Event-Grounded Self-Distillation for Streaming Video Understanding Authors: Yuwei Miao, Xuesheng Zhang, Wenhao Zou, Jixia Zhang, Jianwei Lv, Bo Yuan, Junfeng Wang, Shiao Xie Published: 2026-09-29 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.36803v1 Real-time video understanding requires incrementally maintaining a memory of streaming content, and optimizing this requires dense process signals. On-Policy Self-Distillation (OPSD), which lets one model serve as both teacher and student with the teacher receiving additional privileged information such as the question and ground-truth (GT) answer, can supply such token-level signals. However, applying it directly to streaming video raises two problems. (1) The student cannot be optimized end-to-end, where memory is written before the question arrives, yet the teacher scores it with the question-and-GT privilege, misaligning their preferences. (2) Effective-entity memory collapses, where the question-and-GT privilege makes the teacher favor only question-relevant entities, and token-mean averaging over a memory renders its signal invariant to how many entities that memory covers, both driving memory against the streaming need for diversity. To address these issues, we propose Event-Grounded Self-Distillation (EGSD), which characterizes streaming memory as an incremental update over verifiable Events (key visual entities, actions, and details) and targets the two problems on this basis. For problem (1), we adapt the OPSD signal into a multiplicative weight combined with the outcome reward; for problem (2), we re-weight the teacher with Events as privileged information to counter its question-relevance bias, and add an entity-coverage reward to supply the coverage preference the token-mean teacher lacks. Extensive experiments on mainstream online and offline benchmarks show EGSD achieves strong performance, reaching 79.8% on StreamingBench and 73.4% on the OVO-Bench Real-Time track, while memory analysis shows effective-entity recall rises 17.4% at only 6.8% more memory length. --- ### EasyPPO: Stabilizing the Critic Is Key Authors: Xuanyi Zhou, Qiuyang Mang, Huanzhi Mao, Dacheng Li, Wenhao Chai, Mayank Mishra, Yichuan Wang, Karthik Narasimhan, Alvin Cheung, Joseph E. Gonzalez Published: 2026-09-29 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.36802v1 A key strength of Proximal Policy Optimization (PPO) is its learned critic, which uses historical trajectories collected during reinforcement learning to estimate expected returns and reduce policy-gradient variance. However, we find that the critic is also a major source of instability in reinforcement learning for large language models (LLMs). We identify two critic failure modes that destabilize PPO. First, filtering truncated rollouts from both actor and critic shifts the policy objective to reward conditioned on completion, allowing truncation to increase even as conditional reward improves. Second, heterogeneous return noise can cause high-variance prompts to dominate critic updates in finite batches. We introduce EasyPPO to address these failures. Actor-only overlong filtering trains the critic on returns from both completed and truncated rollouts. Noise-normalized critic regression weights each prompt's critic loss by the inverse standard deviation of its sampled returns, balancing noise contributions across prompts. Moderately smaller critic mini-batches confine outlier influence to fewer rollouts during gradient clipping. Across continuous-reward coding on FrontierCS, binary-reward mathematical reasoning on AIME24, and multi-turn search on Search-R1, EasyPPO remains stable throughout the full training horizon and consistently outperforms vanilla PPO, VAPO, and HL-Gauss PPO. Its best validation scores show relative gains of 14.89%, 2.28%, and 9.47% over PPO, respectively. --- ### Scene Retargeting: Learning Object Placement with Analogical Transfer Authors: Minkwan Kim, Junho Kim, Seungmin Lee, Changwoon Choi, Young Min Kim Published: 2026-09-29 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.36801v1 Interactive simulations of embodied AI or spatial computing applications build on realistic 3D scenes that support daily activities. However, sparse, irregular layout structures impose scene-specific physical constraints, making it hard to define a generalizable framework for generating similar functional context. We formalize Scene Retargeting as stably transferring the semantically coherent spatial organization across layouts, rather than relying on textual descriptions or pairwise relationships. Our cluster-wise transfer flexibly handles mismatched object instances and adapts to distinctive floor plans. We optimize to preserve the rich semantic context of individual clusters by respecting the spatial distribution of foundation features. We can then impose physical constraints to refine wall contacts, pairwise alignment, or clear passageways and openings. Our framework outperforms state-of-the-art methods on layout generation on the 3D-FRONT dataset, and demonstrates downstream applications including real-to-sim transfer, analogical trajectory transfer, and multi-reference composition. --- ### AI as a Compiler: Compiling Triton kernels without the Triton compiler Authors: François Costa, Charly Castes, Thomas Bourgeat, Azalia Mirhoseini Published: 2026-09-29 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.36800v1 Compiler backends are expensive to build and maintain as programming models, workloads, and accelerators evolve. We investigate whether large language models can replace the conventional optimizing and lowering pipeline, a process that we call AI lowering. We study AI lowering from Triton to NVIDIA PTX: an LLM agent translates Triton kernels directly into PTX. We build an environment that evaluates candidate PTX, and an agentic harness in which an LLM translates Triton kernels into PTX. Across twelve common kernels on Ada, Hopper, and Blackwell GPUs and ten kernels from recent ML papers, AI lowering achieves 0.83x-3.34x the performance of autotuned Triton. The largest gains come from transformations that Triton's lowering pipeline does not perform, such as decoding packed binary weights directly into Tensor Core operands (3.34x on BitDelta), assigning each thread a complete softmax row in tensor memory (1.37x on FlashAttention), and reusing overlapping convolution windows (up to 2.23x). These results rely on a robust evaluation harness with comprehensive verification support. We build on Volta, an existing PTX verifier, and substantially extend it to support modern GPU architectures by introducing support for Blackwell's tcgen05 Tensor Core interface. This requires modeling three architectural features: managed tensor memory, descriptor-based operand layouts, and asynchronous execution coordinated through commits, waits, memory barriers, and proxy fences. We discuss the challenges involved in formalizing them, as well as the current limitations. Our results suggest an emerging future in which AI compilers replace custom-written intermediate representations and checkers, reducing the time and engineering effort required to bring up software for new general-purpose and custom chips. --- ### Seeing What Should Be Heard: Diagnosing and Repairing Cross-Modal Shortcuts in Omni-Modal LLMs Authors: Yueran Ma, Ronghao Lin Published: 2026-09-29 Categories: cs.CV, cs.CL, cs.LG arXiv: https://arxiv.org/abs/2609.36798v1 Omni-modal large language models (LLMs) are expected to answer a question using the modality it explicitly refers to. However, existing training paradigms rarely verify whether models actually follow this modality, because multimodal inputs from the same sample often provide redundant evidence for the same answer. In this work, we uncover a pervasive cross-modal shortcut in omni-modal LLMs: when asked an audio-related question, models rely on the image as much as on the audio, and sometimes even more. To systematically diagnose this behavior, we introduce the Factorized Modality Diagnostic, which independently swaps audio and images between samples to isolate each modality's causal contribution. Across two model families in different settings, we find that this shortcut persists throughout supervised fine-tuning and reinforcement learning post-training, while judge-based RL may further amplify such reliance on irrelevant visual information. Based on this finding, we propose DMC-Repair, which trains models on the same kind of cross-modal swapped samples while assigning supervision according to the modality specified by the question. This prevents models from exploiting the spurious correspondence between modalities within the same clip. Experiments demonstrate that DMC-Repair reduces the image-induced share of the answer effect by 59.9%, effectively suppressing the cross-modal shortcut without compromising audio-question answering performance. The reduction in shortcut reliance generalizes across two model families and zero-shot to an unseen dataset and an unseen benchmark, and persists through subsequent post-training. Code is available at https://anonymous.4open.science/r/DMC-Repair. --- ### Where Does Randomness Matter in Neural Cellular Automata? Authors: Fei Zuo, Jiaqi Shi, Yujing Liu Published: 2026-09-29 Categories: cs.LG, cs.NE arXiv: https://arxiv.org/abs/2609.36797v1 Stochastic cell updates are often used throughout the life of a neural cellular automaton (NCA), from backpropagation through time to final rollout. This leaves two questions entangled: does update randomness help learn a useful rule, and must that randomness remain at execution? We separate training and evaluation update modes in controlled Growing NCA experiments, then vary the states shown during training. Under the standard constant-rate persist recipe, asynchronous training passes the short-horizon quality test in 10/10 runs, compared with 3/10 synchronous runs. All ten asynchronous models also retain the target for 4,096 steps under deterministic evaluation. For a scalar translation-invariant lattice, we derive an exact mean-square criterion: random masking can damp mean modes, but it also injects variance, and a mean-only test misclassifies four non-marginal settings. Finally, among 30 models that all pass the same reconstruction test, eight of ten grow-trained models become off-target at 4,096 steps, while all persist and regenerate models retain the target; damage recovery separates persist from regenerate. The results distinguish optimization reliability, execution mode, and task-specific behavior instead of treating them as one stability property. --- ### RAE-PPG: Duration-Grounded Retain-and-Extend Pretraining for PPG Foundation Models Authors: Suyeong Lee, Hochang Lee, Seokyong Sheem, Daekyum Kim Published: 2026-09-29 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.36794v1 Signal features derived from photoplethysmography (PPG) require different signal durations to characterize. Existing PPG foundation models treat duration as a pretraining or evaluation condition rather than using the different durations required by PPG features to organize self-supervision. We hypothesize that self-supervision should expand with signal duration, allowing a single encoder to progressively acquire additional features while preserving and reusing earlier learning. We introduce Retain-and-Extend PPG (RAE-PPG), which trains a single Transformer encoder successively on 10 s, 30 s, and 240 s inputs, adding supervision for signal features supported by each longer observation. The encoder is partitioned into duration-specific parameter groups, allowing later stages to reuse earlier groups while updating only the group assigned to the current stage. Selected earlier targets are reused to supervise later stages, encouraging the corresponding features to remain accessible in longer-input representations. Direct decoding from the final encoder shows that earlier features remain recoverable from longer-input representations, while later-stage features show higher mean decoding performance at their introduction durations. Controlled comparisons further show that prior-stage learning provides a better basis for learning newly introduced features at both transitions. Across 18 tasks from eight datasets, the final frozen encoder achieves the best observed score on 12 tasks compared with five existing PPG foundation models. --- ### Harnessing Large Language Models to Compile Task-Relevant Context into Bayesian Optimisation Authors: Zhongwei Yu, Sourabh Roy, Bin Cao, Xue Yan, Anjie Liu, Jun Wang Published: 2026-09-29 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.36788v1 Incorporating rich task-relevant context, such as domain knowledge and external observations, is a key capability yet remains challenging for Bayesian optimisation (BO). Recently, practitioners have started to use large language models (LLMs) to generate and execute BO programs through coding harnesses. In such emerging practices, the posterior belief is shaped not only by Bayesian inference but also by LLM-generated model and data artefacts, offering a flexible route for task context to enter BO as executable code. To study whether and how LLMs can be harnessed to compile diverse contextual signals for BO, we formulate LLM-compiled BO as generalised-context decision making. We propose HarBO, a BO-specialised harness that compiles generalised context into the core artefacts of standard BO through a validated multi-stage workflow. Our theory analyses the regret under imperfect compilation and the effect of adding new context. Across synthetic functions and real-world benchmarks, we find that LLM harnesses can effectively compile context into standard BO, achieving competitive performance with specialised LLM-embedding-based and direct LLM-in-the-loop BO methods. General coding harnesses can be effective in familiar domains such as hyperparameter optimisation, but fall short in unfamiliar, context-rich domains. Together, these results establish LLM harnesses as a promising, but not automatically reliable, route for making rich task context usable in BO. --- ### Decoding Affective Nuances: Enhancing MLLMs via Hierarchical Emotion Reasoning and Contrastive Discriminative Pruning Authors: Cheng Ye, Weidong Chen, Zhaobo Qi, Beier Zhu, Zhendong Mao Published: 2026-09-29 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.36782v1 While multimodal large language models (MLLMs) have demonstrated exceptional capabilities in objective understanding tasks, their performance in affective reasoning still falls significantly short of human standards. We attribute it to a central capability gap: MLLMs are difficult to reliably distinguish semantically proximal emotions based on fine-grained visual evidence, which could be decoupled as two limitations: 1) Insufficient Attribution. The global reasoning paradigm of conventional MLLMs severely dilutes fine-grained emotion cues, where subtle emotional states are usually implicitly encoded, thereby generating emotional misjudgments in complex scenarios. 2) Insufficient Discrimination. Existing methods could only identify regions generally associated with emotions, which fails to distinguish discriminative regions between semantically similar emotions, leading to ambiguous emotion judgements. To overcome these limitations, we present a training-free inference-time optimization framework, named Decoding Affective Nuances (DAN). Specifically, we propose a Hierarchical Emotional Reasoning Chain (HERC) that enhances the insufficient attribution by harmonizing fine-grained scene/object-level cues and performing a soft-gated reasoning. Furthermore, to discriminate between semantically proximal emotions, we design a Contrastive Discriminative Visual Pruning (CDVP), which isolates discriminative visual tokens to reason the final emotion category by computing the absolute discrepancy between the attention distributions of similar emotions. Performances on several benchmarks demonstrate that DAN significantly improves discrimination for affective nuances without consuming additional training resources, especially achieving +10.47% improvements with Qwen3-VL-8B-Instruct on WebEmo25 dataset that contains 25 fine-grained emotion categories. --- ### Aperture: Merge-Consistent Rotary States for Compressed Tokens Authors: Yuhao Du, Shunian Chen Published: 2026-09-29 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.36781v1 Token compression combines content from several positions, yet rotary position embeddings usually assign the merged token one coordinate. We ask what positional information must survive later merges. Aperture stores Fourier moments of the token's weighted support at the model's rotary frequencies. We prove that these moments have minimal real dimension among continuous states sufficient for the selected expected rotary interactions. Represented mass makes updates additive; attention normalisation remains a separate readout choice. Uniform intervals give a centre rotation times a sinc gain. We characterise when centres determine interval widths and construct matched examples where they do not. Numerical checks verify the weighted-support implementation. In trained temporal readers, compression transfer varies with gain calibration and feature placement. In a prespecified native video question-answering comparison, stored support reaches $65.63\%$ accuracy versus $67.12\%$ for the deployed merging rule. These results separate exact positional preservation under compression from downstream benefit. --- ### DRHeC: Differentiable Rendering for Hand-Eye Calibration with RGB-Based Gradients Authors: Xiaotian Zhang, Yusheng Wang, Naoya Kagawa, Noritaka Takamura, Keiji Okuhara, Hiroyasu Baba, Jun Ota Published: 2026-09-29 Categories: cs.RO, cs.CV, cs.GR arXiv: https://arxiv.org/abs/2609.36779v1 Accurate hand-eye calibration is crucial for precision manipulation. Traditional methods rely on markers, with their precision dependent on marker accuracy and observability. In contrast, markerless methods, such as learning-based approaches, use deep neural networks to directly extract keypoints or features from images, enabling the computation of hand-eye transformation with a single image and without the need for physical markers. Recently, differentiable rendering-based methods for hand-eye calibration have leveraged physical models to render binary masks and compare them with observations, enabling hand-eye calibration without fiducial markers in the calibration stage and providing interpretable optimization. While the state-of-the-art differentiable rendering methods achieve remarkable accuracy, the use of binary masks can result in the loss of internal profile details, reducing precision. Additionally, these methods can also suffer from unstable optimization and local minima. In this study, we propose a novel RGB-based differentiable rendering framework that provides richer geometric and appearance cues by incorporating color and mask geometric features, thereby improving calibration accuracy and optimization stability. Additionally, we propose a mask-guided image-to-image translation method to ensure explicit preservation of color and geometric consistency throughout the translation. Our approach is validated through both simulation and real-world experiments, with results demonstrating strong accuracy and robustness and clear improvements over existing differentiable rendering methods. Our method achieves a grasping success rate of 88.9% and insertion success rate of 57.4% on the UR5e real-world experiment, outperforming the state-of-the-art differentiable rendering hand-eye calibration method EasyHeC by 46.3 and 48.1 percentage points, respectively. --- ### Code4Scene: Benchmarking Coding Agents for Constructing and Editing 3D Scenes Authors: Xiaokang Ye, Siddhant Hitesh Mantri, Zimeng Chen, Edward Zhang, Zhaoxu Zheng, Yuanheng Li, Yizhao Chen, Tianyang Huang, Lianhui Qin Published: 2026-09-29 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.36777v1 Frontier coding agents can now write and execute code that authors 3D environments, but whether they reliably understand 3D structure and precisely control scene state remains unclear. The generated 3D scene is a persistent, executable artifact: a convincing render can hide incorrect spatial relations, intersecting objects, or unintended modifications. We introduce Code4Scene, a benchmark of 190 Unreal Engine cases built from human-assembled scenes that evaluates coding agents on two complementary settings under a shared execution interface. Construction tests scene-level spatial reasoning from open-ended language specifications, where many realizations are valid; editing tests precise control of scene state, where the agent must recover the target scene from reference images while preserving everything else. Rather than scoring code or rendered views, Code4Scene evaluates the generated engine-native scene for task fulfillment, artifact integrity, and static physical validity, with edits additionally compared against withheld ground truth. Across 14 coding-agent configurations on the 95-case public set, construction and editing performance are strongly correlated but not interchangeable (Spearman $ρ= 0.78$): Claude Fable 5.1 leads construction, Gemini 3.8 Flash leads editing, and GPT-6 Astra narrowly leads overall. Spatial Composition is the weakest construction category for every agent, while editing remains imprecise: the best Repair F1 is only 0.527, and 35.8% of edits that fully recover the target still introduce unintended changes elsewhere in the scene. These results expose a gap between plausible 3D generation and reliable spatial reasoning and state control. --- ### Causal-EVC: Breaking Emotional Spurious Causality via Spatiotemporal Grounding and Counterfactual Intervention Authors: Cheng Ye, Weidong Chen, Peipei Song, Zhendong Mao Published: 2026-09-29 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.36776v1 Emotional Video Captioning aims to generate factually accurate and emotionally empathetic descriptions. While recent methods have recognized the importance of visual causes to guide emotion perception and caption generation, they fundamentally rely on simple attention matching, which inevitably suffers from {causal redundancy and spurious correlations} in co-occurrence bias (e.g., misclassifying ``sadness'' as ``joy'' on a sunny beach), leading to severe shortcut learning from confusing backgrounds. Furthermore, existing evaluations fail to verify whether models have genuinely mastered causal reasoning or merely exploited background confounders. To address these limitations, we first construct {EVC-CauseGround}, a comprehensive benchmark with dense spatio-temporal causal annotations. Crucially, it introduces a carefully selected {Causal-Faithfulness Subset} to explicitly quantify genuine emotion-cause attribution. Second, we propose {Causal-EVC}, an emotion-grounding captioning framework, which introduces a Motion-guided Causal Spatiotemporal Localization module to precisely decouple causal triggers from background confounders. Besides, we introduce an Interpretable Sparse Emotion Routing module. By synthesizing counterfactual representations and formulating a novel counterfactual contrastive objective, we enforce the model to anchor its emotion predictions strictly on authentic causal triggers instead of confusing background. Extensive experiments show that Causal-EVC not only achieves the best performance on semantic metrics but also exhibits significant advantages in the causal-faithfulness subset, which demonstrates that our model could mine emotional cues from genuine visual causes and mitigate co-occurrence bias for interpretable multimodal emotion understanding. --- ### DSPO: Diversity-aware Subjective Policy Optimization for Robust Emotional Reasoning Authors: Cheng Ye, Weidong Chen, Bingyan Xu, Zhendong Mao Published: 2026-09-29 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.36775v1 Reinforcement Learning has significantly advanced the complex reasoning capabilities of MLLMs. However, prevailing RL algorithms suffer a severe failure in emotion reasoning tasks. These methods heavily rely on deterministic hard-label supervision and point-wise isolated evaluation, creating a fundamental gap with the inherently subjective and continuously distributed nature of human emotions. Furthermore, unlike explicit physical objects, emotional states are deeply implicit within visual cues. This abstract nature exacerbates visual hallucinations in MLLMs, leading to plausible yet ungrounded emotional evidence. To address these limitations, we propose Diversity-Aware Subjective Policy Optimization (DSPO), a reinforcement learning framework that jointly promotes subjective affective coverage and visual grounding. First, we construct a context-grounded emotional distribution prior in the VAD space by combining the lexical prior of the annotated emotion with image-specific contextual information. Based on this prior, we introduce a Distribution-Aligned Emotional Diversity Reward (DEDR), which measures the leave-one-out marginal contribution of each candidate emotion within a rollout. DEDR rewards candidates whose inclusion brings the predicted affective set closer to the context-grounded prior, thereby preserving plausible subjective interpretations without encouraging unconstrained dispersion. We further develop Counterfactual Visual Intervention Gating (CVIG), which masks the visual region highlighted in the reasoning process and uses the resulting candidate-wise probability changes to reduce the weights of interpretations unsupported by visual evidence. Extensive experiments demonstrate that DSPO achieves state-of-the-art performance across multiple public benchmarks, especially on the cross-domain performance, i.e., improving +10.8\% on average cross-domain accuracy than EMO-R3. --- ### NeuroDyn-EEG: An Interpretable Pre-trained Model for EEG Based on Neural Dynamics Authors: Yi Cui, Tong Zhao, Jiaxin Lei, Chuyi Yang, Yifan Cui, Ling Zhang, Yuxiang Yan, Bo Hong Published: 2026-09-29 Categories: cs.NE arXiv: https://arxiv.org/abs/2609.36773v1 Clinical scalp electroencephalography (EEG) offers a noninvasive window into neural dynamics of neuropsychiatric disorders. However, discriminative deep models often lack anatomically indexed physiological interpretability. We propose NeuroDyn-EEG, a pretraining framework integrating generative priors from neural dynamics. It couples an extended Jansen-Rit neural mass model, leadfield-based source projection, and simulation-based parameter inversion. Trained on synthetic parameter-EEG pairs within physiological ranges, NeuroDyn-EEG estimates 11 regional parameter families across 90 AAL regions plus one global parameter from standard 19-channel EEG, using only ~2.43M trainable parameters. We evaluate the framework across three levels. First, controlled simulations demonstrate robust parameter recovery under diverse noise conditions, while real resting-state EEG evaluations confirm spectral and phase consistency in an inverse-forward closed loop. Second, on four clinical benchmarks (AD65, PD31, Figshare MDD, and TUAB), NeuroDyn-EEG achieves competitive classification performance, securing the highest BACC, AUROC, and AUCPR on PD31 and MDD, and highest BACC on AD65. Third, post hoc regional analyses reveal disease-specific alterations: local synaptic connectivity C_1 involves the most altered regions in AD65, whereas the firing threshold theta ranks first in MDD, offering testable mechanistic hypotheses. Overall, NeuroDyn-EEG maps scalp EEG to anatomically indexed dynamical parameters, bridging representation learning and mechanistic neurophysiology. Code: https://github.com/Gnosis-Neurodynamics/NeuroDyn-EEG. --- ### Beyond Conditional Independence: Root Cause Analysis with Deep Causal Models Authors: Md Musfiqur Rahman, Kenneth Lee, Ziwei Jiang, Padmaja Jonnalagedda, Ruocheng Guo, Murat Kocaoglu Published: 2026-09-29 Categories: cs.LG, cs.AI, stat.ML arXiv: https://arxiv.org/abs/2609.36771v1 Root cause analysis (RCA) is a critical problem in many real-world scenarios. RCA enables the identification of faulty or failing mechanisms in a system by comparing anomalous observations with corresponding reference (i.e., regular) observations. However, existing approaches rely either on heuristic methods or on conditional independence tests with a strong unconfoundedness assumption, and thus fail to exploit other complicated distributional constraints in the presence of latent variables. To relax these assumptions, we model the underlying system as a causal model and the anomalous system as a change in the structural functions of the same causal model. Specifically, to handle unobserved confounders, we establish an implicit connection between distributional constraint testing and root cause analysis. To adapt our approach to data generated from arbitrary causal models, we employ the deep causal model (DCM) framework, in which we design the causal model using neural networks. Finally, we illustrate how our method, RCA-DCM, can utilize different levels of partial graphical knowledge to perform RCA. We evaluate RCA-DCM against state-of-the-art baselines on simulated datasets, a physics-based causal chamber and two micro-service applications. RCA-DCM improves top-1 accuracy over the strongest baseline on both Sock Shop (0.880 vs. 0.752) and Online Boutique (0.776 vs. 0.712), and when the true root cause in the causal chamber is unobserved and acts as a latent confounder, it recovers the exact root-cause set more often than any competing method (perfect recovery rate (PRR) 0.846 vs. 0.731). --- ### Emergent Specialization in Populations of Self-Supervised Collaborative Vision Experts Without a Shared Gate or Cross-Agent Gradients Authors: Aram Davtyan, Pablo Acuaviva, Sebastian Stapf, Paolo Favaro Published: 2026-09-29 Categories: cs.AI, cs.CV, cs.LG arXiv: https://arxiv.org/abs/2609.36770v1 Can a population of neural networks develop a useful division of labor without a shared gate or gradients between agents? We study a setting where each network has its own weights, trains independently on the same heterogeneous data, and can ask another agent for help through a forward pass. Unlike mixtures of experts, where a jointly trained gate assigns inputs to experts, specialization here must emerge without central control. We test this in a small scale proxy for predictive visual pretraining. Initially identical agents are finetuned on an unlabeled mixture of six visual domains using masked prediction of frozen DINOv3 features. We measure specialization by asking whether the best agent for an input aligns with its latent domain, and utilization by asking whether responsibility is distributed across agents. We progressively remove central control, ending with DISCO (DIStributed COllaboration) where each agent locally selects a helper, reads its internal state through a gradient free channel, and rewards its router only for the improvement that help provides. Specialization emerges and is useful. Randomly routed populations underperform a single generalist, while semantically routed populations outperform it, showing that specialization rather than population size drives the gain. Specialization persists without a central router, and gradient free communication lets nonexperts exploit emergent expertise. In DISCO, a random agent helped by the expert matches the solo generalist, while experts surpass it, including on data outside the specialization mixture. Local routers select the emergent expert for 98% of inputs. These effects persist across population size, model capacity, data imbalance, and finetuning seeds, providing measurable evidence for the dynamics needed by decentralized predictive pretraining. --- ### When Can Prefixes Compile LoRA? Exact Resource-Capped Tests for Frozen Attention Authors: Joyanta Jyoti Mondal, Ibne Farabi Shihab Published: 2026-09-29 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.36766v1 Can a fixed continuous prefix replace a given low-rank adapter while the attention head stays frozen? In this research, we show that the answer depends on the adapter's target through three conditions. First, observability: at one causal readout, every independent key--value prefix sees the content only through the query, attention partition, and value numerator, so a target that differs on two inputs with equal summaries incurs an error floor at every prefix length; norm caps extend this floor to nearly equal summaries. Second, realizability: at a common query, any prefix reduces exactly to two aggregate variables, and the norm-capped optimum is an attained second-order-cone program, also after a fixed output projection; it places two equal-norm rank-one value updates on opposite sides of compilability. Third, implementation: under affine query exposure, $2r$ signed slots approximate a rank-$r$ value update, but their values grow as $O(ε^{-3/2})$, and the construction passes all 400 tolerance checks in float64 yet only 38 in bfloat16. A first-layer GPT-2 readout with fixed token and position meets the common-query condition without clamping activations; at three such heads, the capped optimum leaves 18.4\% to 74.2\% of the projected adapter effect uncompiled, with a head-dependent value--query ordering. All claims concern local approximation at one head, not whole-network equivalence. --- ### Graph-Spectral Flow Matching for Multivariate Time Series Anomaly Detection Authors: Zepeng Zhang, Jhony H. Giraldo, Wenbin Wang, Olga Fink Published: 2026-09-29 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.36765v1 Multivariate time series anomaly detection typically relies on evaluating discrepancies between observations and outputs produced by models trained on normal data. An alternative perspective is to characterize the distribution of normal data through the generative dynamics, i.e., the velocity field, of flow matching models. However, standard flow matching typically adopts linear probability paths that overlook dependencies among variables, leading to a misalignment with the structured data distribution. To address this issue, we propose GRASP, a flow matching framework with a graph-spectral path for multivariate time series anomaly detection. GRASP incorporates graph structure into the probability path by minimizing a fixed-endpoint action that combines kinetic energy with graph Dirichlet energy. This formulation yields a closed-form path based on graph-frequency-dependent hyperbolic interpolation. A velocity predictor trained on normal data then detects anomalies using weighted velocity discrepancies aggregated across source samples, flow times, and graph frequencies. Theoretically, we establish that GRASP is invariant to the choice of Laplacian eigenbasis and decompose its expected oracle anomaly score into bounded endpoint uncertainty and graph-frequency-weighted Fisher discrepancy. Experiments on four benchmarks demonstrate the superior anomaly detection performance of GRASP and validate the effectiveness of its graph-spectral path and weighting mechanism. --- ### Federated Clustering with Unknown Local and Global Cluster Cardinalities Authors: Mitushi Goyal, Tarun S., Riddhanya Senapathi, Arun Raman Published: 2026-09-29 Categories: cs.LG, cs.DC, stat.ML arXiv: https://arxiv.org/abs/2609.36762v1 Federated clustering methods that do not require the global number of clusters $K$ still assume that each client knows its local number $K_g$. This assumption is hard to justify when clients know no more about their data than the server does, as in fault diagnosis across independently operated industrial sites. We propose a two-phase framework in which neither count is known: each client first estimates $K_g$ from its own data, and an aggregator that requires local counts, such as FedGEM, then uses these estimates in place of the true values. For the first phase we introduce Adaptive Split--Merge (ASM), which grows a spherical Gaussian mixture by BIC-driven splitting and then merges excess components. ASM uses no labels, selects its hyperparameters on held-out client data only, and makes no assumption about how clusters are shared across clients. We derive a closed-form split criterion whose critical cluster size falls with anisotropy and rises with dimension, and show empirically that over-fragmentation grows with the number of points per cluster, which federation divides among clients. Across eight datasets, ASM with FedGEM attains a mean ARI of 0.333, against 0.256 for the next best label-free estimator and 0.361 when the true local counts are supplied. It also gives the most reliable global estimates of $K$ and is robust when client size is decoupled from local cardinality. --- ### QuantMLA: Function-Aligned Dual-Path Quantization for Low-Bit MLA KV Caching Authors: Zunhai Su, Yuxuan Sun, Jianchao Tan, Tao Zhang, Ruihan Hu, Yuchen Xie, Xunliang Cai, Ngai Wong Published: 2026-09-29 Categories: cs.LG, cs.CL arXiv: https://arxiv.org/abs/2609.36760v1 Multi-Head Latent Attention (MLA) enables expressive multi-head attention with compact caches for its content and decoupled RoPE paths, yet cache memory still scales linearly with context length and batch size. In this work, we establish a systematic model of MLA's dual-path quantization errors, characterizing their distinct effects on attention-output distortion and explaining the pronounced amplification of RoPE-path errors. Guided by this analysis, we introduce QuantMLA, a function-aligned framework for low-bit dual-path quantization. We derive path-specific transformation spaces that preserve full-precision computation while remaining fully fusible into model parameters offline, eliminating online transformation overhead. Within these spaces, QuantMLA learns path-specific transformations with function-aligned objectives: attention-output reconstruction captures the content path's coupled matching and aggregation errors, while positional QK reconstruction preserves the RoPE-induced component of the attention logits and admits a theoretical bound on output distortion. Across four MLA model families, QuantMLA enables, to our knowledge, the first reported joint INT4 caching of the content and RoPE caches with minimal accuracy degradation. Further compressing the content cache to INT2 while retaining the RoPE key cache at INT4 maintains competitive performance on challenging reasoning and code benchmarks. We develop a native low-bit MLA attention kernel that integrates unpacking and dequantization directly into attention computation. The physical cache layout provides 3.59x compression at 128K context, while a cache-pressure serving workload achieves 5.168x higher whole-job output throughput than BF16. The code will be released upon acceptance. --- ### Dual-Mode Low-Rank Learner with Bridge-Prototype Ensemble for Vision-Language Class-Incremental Learning Authors: Chiyuan He, Zihuan Qiu, Fanman Meng, Chao Wang, Liangjiang Chen, Linfeng Xu, Qingbo Wu, Hongliang Li Published: 2026-09-29 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2609.36759v1 Benefiting from transferable visual-textual alignment, CLIP has been widely adopted for class-incremental learning (CIL). However, existing learners either repeatedly update components shared across tasks, leading to knowledge overwriting, or overly isolate new-task updates, hindering the reuse of CLIP's transferable knowledge and limiting plasticity. Moreover, the text-based or bimodal classifier designs still fail to effectively integrate complementary information from the visual and textual modalities. To address these challenges, we introduce DuLBE, which couples dual-mode low-rank learning with a bridge-prototype ensemble classifier for exemplar-free CIL. DuLBE allocates two visual low-rank update modes according to the gradient demand and uses gradient routing to coordinate them: a compact and rewritable shared mode is selected from historically occupied visual directions to reuse transferable knowledge, while residual modes provide low-interference channels for task-specific variations. Building on the resulting stable inter-modal structure, we further construct geodesic bridges between visual prototypes and text embeddings on the unit hypersphere, and ensemble reliable bridge prototypes to compensate for the modality-gap limitations of textual decision boundaries. Extensive experiments under multiple settings show that DuLBE achieves state-of-the-art CIL performance while retaining the high parameter efficiency of low-rank tuning. --- ### FastVR: Efficient Streaming Video Restoration with One-Step Diffusion Authors: Xiaoxu Chen, Qin Yang, Haoran Bai, Sibin Deng, Ying Chen Published: 2026-09-29 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.36757v1 Diffusion-based video restoration recovers realistic details, but its practical deployment is limited by two efficiency bottlenecks: costly VAE encoding and decoding, and the quadratic cost of full self-attention in diffusion transformers (DiTs). This paper presents FastVR, a streaming video restoration framework built on a one-step diffusion model, which delivers strong restoration quality and temporal consistency while processing 1080p video at 11 FPS on a single H20 GPU. To improve inference efficiency, FastVR combines a lightweight VAE with chunk-wise causal attention, which substantially reduces the computational cost. During training, it further adopts velocity consistency regularization and continuous trajectory learning, which improve restoration quality. Extensive experiments show that FastVR is more efficient than the evaluated diffusion baselines while achieving state-of-the-art performance on synthetic and real-world benchmarks. We hope that this work supports further progress in the community. --- ### NesTok: Nested Self-Aligned 1D Tokenizer for Autoregressive Image Generation Authors: Jiawei Zhang, Shuhao Liu, Rong Huang, Yuancheng Li, Zhihui Li, Xiaojun Chang, Changlin Li Published: 2026-09-29 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2609.36756v1 One-dimensional (1D) variable-length visual tokenizers enable adaptive compression by varying the number of tokens, allowing downstream autoregressive (AR) models to flexibly trade off generation quality against computational cost using a single tokenizer. However, existing approaches based on nested dropout often fail to fully exploit the representational capacity of the tokenizer, resulting in suboptimal performance in both image reconstruction and generation. In this work, we introduce NesTok, a nested self-alignment framework tailored to dynamic visual tokenizers. NesTok introduces cross-length training, which jointly optimizes reconstruction across token lengths while using the full-length sequence to guide shorter counterparts, enabling shorter token sequences to approach the reconstruction quality of full-length sequences. On ImageNet, NesTok improves substantially over standard training and achieves an rFID score of 0.98. On downstream image generation, it achieves the state-of-the-art gFID score of 1.46 on ImageNet 256$\times$256 among existing variable-length autoregressive image generation methods. Code will be available at https://github.com/jaiwei804/NesTok. --- ### Drag as Evidence: Motion-Grounded Latent Recomposition for Drag-Based Editing Authors: Xinyu Pu, Hongsong Wang, Jie Gui, Pan Zhou Published: 2026-09-29 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.36755v1 Modern image editors excel at semantic manipulation and visual synthesis, yet remain limited in precise spatial control, motivating the development of drag-based editing. However, existing drag-based methods often struggle to balance drag accuracy with natural, plausible, and intent-aligned generation. We propose MoRe-Drag, a motion-grounded drag-based editing method. Our key insight is to treat pixel-space warping as coarse motion evidence, and to inject this evidence into the generative sampling trajectory. Specifically, MoRe-Drag performs region-aware latent recomposition over refinement, inpainting, and anchor regions, coupled with stage-adaptive conditioning that progressively shifts from motion-grounded structure formation to semantic refinement. We further support an instruction-free interface by adapting the MLLM-based text encoder for drag-aware instruction inference. Experiments on DragBench-SR and DragBench-DR show that MoRe-Drag substantially improves drag precision over strong base editors and achieves superior drag accuracy among SOTA drag-based methods, while delivering strong semantic consistency and visually realistic results. Code and dataset will be publicly released. --- ### Does a prosody-trained representation help beyond trainable fusion? A parameter-matched study with frozen HuBERT Authors: Ki Woong Moon, Daniel Brenner Published: 2026-09-29 Categories: eess.AS, cs.CL arXiv: https://arxiv.org/abs/2609.36754v1 Explicit prosodic cues may help automatic speech recognition (ASR) of spontaneous speech, but auxiliary representations typically require additional trainable components, making it unclear whether gains come from the auxiliary information or the fusion mechanism. We address this using a frozen HuBERT backbone and a 64-dimensional representation trained to predict log F0, voicing, Delta log F0, log energy, and spectral tilt. We compare a frozen-backbone recognizer (Baseline), trainable fusion with zero auxiliary input (Null), and the same fusion supplied with the learned representation (Learned). Across Buckeye, Switchboard, and AMI IHM, Null reduces WER by 0.71-1.45 points over Baseline, whereas Learned differs from Null by +0.07, -0.09, and +0.00 points, with no significant differences. However, removing or mismatching the representation at inference increases Learned WER. Thus, Learned depends on the representation yet shows no measurable incremental WER benefit over the parameter-matched control. --- ### cktFormer: Transformer-Based Approach for Automated Analog Circuit Design Authors: Pasindu Dodampegama, Praveen Wijesinghe, Naveen Basnayake, Keshawa Jayasundara, Tharindu Bandaragoda Published: 2026-09-29 Categories: cs.LG, cs.AR arXiv: https://arxiv.org/abs/2609.36752v1 Circuit design is a complex and iterative process that requires expertise in electronic engineering. It involves selecting components while meeting performance constraints, such as power efficiency, cost-effectiveness, and signal integrity. However, manual design is time-consuming and prone to errors. Although other stages of the manufacturing pipeline have benefited from AI-driven optimizations, circuit design remains a bottleneck, limiting overall productivity. Generative AI and machine learning offer the potential to automate and improve this stage, boosting efficiency and accuracy. To address this, we introduce a dual transformer architecture that bridges the gap between AI and circuit design by leveraging attention mechanisms to model complex, non-sequential circuit relationships. Our approach structures netlist data into graph-based representations, enabling effective learning of circuit topology and component interactions. The system consists of two interlinked models: a node prediction model that proposes components and an edge prediction model that infers valid connections. This collaborative and decoupled design captures both component-level semantics and global structural coherence. In our experiments, this architecture outperforms recent models such as AnalogGenie and cktGNN in the validity of generated circuits. By addressing key limitations in existing methods, our work advances automation in electronics engineering and contributes a benchmark for AI-driven circuit synthesis. --- ### Group-Marginalized Self-Rewarding RL Drives Zero-Label Self-Evolving Authors: Yiming Wang, Yikang Liu, Qingyuan Tian, Xingyu Chen, Zhuosheng Zhang, Zhaopeng Tu, Rui Wang Published: 2026-09-29 Categories: cs.LG, cs.AI, cs.CL arXiv: https://arxiv.org/abs/2609.36750v1 Self-rewarding reinforcement learning (RL) enables large language models (LLMs) to self-evolve without human labels. Existing ensemble-based methods construct reward references from rollout groups and assign rewards accordingly. However, a response's reward representation also depends on its randomly sampled group context, i.e., the other responses in its group. Using only one group-context realization may miss desired reward signals and provide unreliable guidance for policy optimization. To address this issue, we propose Group-Marginalized Advantage Estimation (GMAE), which aggregates reward realizations across possible contexts into a response-level distribution and estimates expected advantages. Experiments across eight benchmarks and four base models demonstrate strong performance and cross-domain generalization. GMAE also exhibits stable learning, low extra cost, and good applicability across training datasets and RL backbones. --- ### Generalizable Lifelong Model Editing via Preference Optimization Authors: Dahyun Jung, Suhyune Son, Heuiseok Lim Published: 2026-09-29 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.36748v1 Knowledge editing enables rapid updates of specific factual knowledge in large language models (LLMs) without full retraining. However, more realistic scenarios call for a lifelong framework that handles continual updates rather than one-off modifications. In such settings, existing editing methods often overfit to target prompts, significantly degrading both the generalization of the edited knowledge and the model's general capabilities. To address this issue, we propose GLIME (Generalizable Lifelong Model Editing), which combines knowledge editing with preference optimization over generation behavior. GLIME further incorporates replay-based editing and a gradient constraint to preserve previously edited knowledge. Experimental results show that GLIME significantly improves knowledge generalization in lifelong editing settings while maintaining both editing performance and general capabilities. --- ### EASE: Behavior-Adaptive Skill Curation for Self-Evolving Agents Authors: Zhen Xiong, Qiaoyu Tan Published: 2026-09-29 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.36746v1 Agent skills provide a lightweight mechanism for self-evolving agents to accumulate reusable procedural knowledge without updating model parameters. However, existing learned skill curators typically optimize curation without explicitly modeling downstream executor behavior. We show that this can cause systematic cross-executor degradation: curators trained with different executors perform best when paired with their own training executor, indicating that effective skill curation is executor-dependent. We formulate behavior-adaptive skill curation and introduce EASE, a framework that learns a single curator that adapts its decisions to different executor behaviors. EASE maintains an online behavioral profile of recent execution patterns and conditions the curator on this profile, the current trajectory, and retrieved skills to add, modify, or remove skills from an evolving repository. We train the shared curator jointly across multiple frozen executors with reinforcement learning, using retrieval-aware and behavior-aware temporal attribution to focus optimization on curation actions with observable downstream influence. Across ALFWorld, ScienceWorld, and WebShop, with executors ranging from Qwen3-8B/32B and GPT-OSS-120B to unseen Kimi K2.6, DeepSeek V4 Flash, and Gemini 3.5 Flash, EASE outperforms strong skill- and memory-based baselines without per-executor finetuning. EASE also maintains 34.5--41.0% fewer skills, improves skill retrieval by 36.3--38.7% and measured edit utility by 51.8--60.0%, and reduces deployment-time inference tokens by 9.1--14.5%. These results establish behavior-adaptive skill curation as an effective principle for building self-evolving agents. --- ### SIPO: Unifying Reinforcement Learning with On-Policy Self-Distillation Authors: Zhenrui Yue, Huimin Zeng, Yueqi Wang, Yaokun Liu, Fengran Mo, Jinghan Zhang, Mung Yao Jia, Gyuseok Lee, Yang Zhang, Na Wei, Dong Wang Published: 2026-09-29 Categories: cs.AI, cs.CL arXiv: https://arxiv.org/abs/2609.36742v1 Reinforcement learning with verifiable rewards (RLVR) has become a standard paradigm for improving large language models (LLMs) on various tasks, yet its sparse outcome rewards lack token-level credit assignment for intermediate steps. To address this, on-policy self-distillation (OPSD) leverages a self-teacher with privileged context to provide additional dense learning signals. However, because the self-teacher is often overconfident and imposes excessive penalties on long reasoning trajectories, OPSD frequently struggles in practice. To mitigate this, we propose self-instructing policy optimization (SIPO) with a contrastive self-teacher to provide dense credit. At each iteration, SIPO samples multiple rollouts per prompt from the current policy, scores them with environment rewards, and constructs two teacher contexts for each rollout by pairing the reference answer with mistakes made within the group. The model then re-evaluates its own responses under both contexts, using the difference between the two teacher log-probabilities as token-level feedback, so that biases shared by both contexts are expected to largely cancel. The resulting objective yields a token-level advantage for every rollout: the reward still sets the main direction of each update while the self-teacher redistributes credit across tokens. Even in groups where every rollout fails and group-relative advantages vanish, SIPO still provides a learning signal. By preserving direct optimization of the task reward while providing dense, token-level feedback, this approach bridges reinforcement learning and on-policy self-distillation. Extensive experiments across multiple reasoning and code-generation benchmarks demonstrate that SIPO outperforms both RLVR and OPSD baselines without an external teacher or additional generation. --- ### Distinguish or Homogenize: Last-Chance Policy Identification and Risk-Budgeted Recovery under Irreversible Resource Depletion Authors: Yibo Guo, Xiaodan Wang Published: 2026-09-29 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.36741v1 Under irreversible resource depletion, an agent can spend resources to distinguish among latent fault models, or to change the system state so that the remaining models admit a common acceptable continuation--at which point further diagnosis becomes unnecessary. This distinguish-or-homogenize principle identifies a path that existing frameworks for identification, planning, and diagnosis do not make explicit: prior formulations treat the mapping from fault models to acceptable policies as a given, whereas LCPI makes it a function of the agent's own actions. We formalize this principle through Last-Chance Policy Identification (LCPI), where correctness is evaluated at the state the agent reaches rather than at the initial state. The Last Identifiable Margin (LIM) marks the feasibility boundary between distinguishing and homogenizing. For deterministic diagnostic graphs we provide the Exact-LIM recursion; for noisy finite-horizon recovery we propose Risk-Budgeted Compatibility Planning (RBCP), which searches a compatibility-aware frontier under a hard worst-case failure constraint. Across incident recovery on abstract microservice topologies and latent-damage navigation in MiniGrid, RBCP improves risk-feasible recovery while satisfying the failure budget. A sham control--cost-matched actions that preserve model incompatibility--eliminates the gain entirely, confirming that the benefit comes from changing which policies are acceptable for which models, not from extra search or additional budget. --- ### Efficient Offline Learning of Ranking Policies via Top-$k$ Policy Decomposition Authors: Ren Kishimoto, Koichi Tanaka, Haruka Kiyohara, Yusuke Narita, Yasuo Yamamoto, Nobuyuki Shimizu, Yuta Saito Published: 2026-09-29 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.36740v1 Many recommender systems such as for e-commerce and news platforms aim to provide users with rankings they are likely to interact with. Off-Policy Learning (OPL) of ranking policies enables us to learn new ranking policies using only historical logged data. However, ranking settings make OPL remarkably challenging because their action spaces consist of permutations of unique items, being extremely large. Existing methods primarily use either policy- or regression-based approaches. The policy-based approach, which typically uses importance-weighted policy gradients, can suffer from high variance due to large action spaces. The regression-based approach, on the other hand, estimates the expected reward using conventional machine learning methods, avoiding variance issues but potentially suffering from severe bias. To circumvent these issues of existing methods, we propose a new OPL method for ranking, named Ranking Policy Optimization via Top-$k$ Policy Decomposition (R-POD), which combines the policy- and regression-based approaches in an effective fashion. Specifically, R-POD decomposes a ranking policy into a first-stage policy for selecting top-$k$ actions and a second-stage policy for choosing the bottom actions given the top-$k$ actions. It learns the first-stage policy using a new policy gradient estimator and the second-stage policy via the regression-based approach. This method can substantially reduce variance, since it applies importance weighting only to the top-$k$ actions. We also demonstrate that our policy-gradient estimator for the first-stage policy is unbiased under a conditional pairwise correctness condition, which only requires that the expected reward differences of pairs of rankings sharing the same top-$k$ actions can be estimated correctly. --- ### Frontier Autolab: Organizational Memory, Adversarial Dissent and Temporal Leakage in Multi-Agent LLM Firms Across Fifty Years of Technological Change Authors: Bravish Ghosh Published: 2026-09-29 Categories: cs.CE, cs.AI arXiv: https://arxiv.org/abs/2609.36739v1 Multi-agent LLM systems are increasingly structured like organizations, with roles, critics and shared memory, yet they are evaluated on tasks that last minutes. We ask how such an organization behaves when the ground it stands on keeps moving. Frontier Autolab is a long-horizon testbed in which one simulated firm, voiced by sixteen role personas and a dedicated Red Team, must re-found itself in nine technology eras from 1990 to 2040. Each era is temporally gated: the firm decides from a dated briefing, a historian-judge then reveals what happened and scores the decision on a five-dimension rubric, and lessons enter a persistent Playbook. Six eras are scored against history, one against the live market and two are open forecasts. Across four trajectories (36 era decisions, 180 subscores) we find a consistent foresight-commitment gap: in all 24 historically scored eras the judge rated the firm's recognition of the coming shift above its choice of where to build (mean gap 1.9 points on a 10-point scale), because boards chose the layer their existing assets could reach. Organizational design shaped long-run character. A Red Team armed with numeric kill gates produced fifty years of gated pilots and no product, and the rubric rated this firm highest; firms whose memory stored market-structure lessons pivoted every era, while a firm whose memory stored only validation procedure kept one method throughout. We also show why such results are hard to trust. Scores rise across eras in every run while the judge's own hindsight subscore falls (within-run r = -0.58), so apparent learning is confounded with recall of history, and we trace further distortions to self-judging, briefing selection and score aggregation. We release all records and an API harness, and specify fictional and post-cutoff eras that would turn the testbed into a benchmark. --- ### Backpropagated Output Momentum: Relocating Optimizer History from Parameters to Task Space Authors: Yuchen Li, Zongqi Fan, Nguyen H. Tran, Ken-Tye Yong Published: 2026-09-29 Categories: cs.LG, cs.CL arXiv: https://arxiv.org/abs/2609.36738v1 Optimizer momentum is usually stored as a parameter-sized moving average of past gradients, which makes history costly and fixes each past signal in the coordinates in which it was computed. We introduce Backpropagated Output Momentum (BOM), which instead stores a compact moving average of prediction errors at the model output and reprojects that history through the current network at every step. A batch-level analysis characterizes the information retained and omitted by this relocation, while the implementation preserves the current supervised gradient and can replace the first-moment component of several adaptive optimizers. As a plug-in for momentum-based optimizers, including ones that already compress their state, BOM reduces parameter-shaped optimizer state by 49.7-99.8% in three compositions and, averaged over three language backbones, paired step time by 4.0%. It also improves mean validation performance across language and vision fine-tuning, by 1.42 points in the primary five-task comparison. Language and vision pretraining studies, together with matched mechanism controls, further test the construction across output spaces and model scales. --- ### Reconstructing the Vocal Tract with Differentiable Acoustic Simulation Authors: Eric Ming Chen, Jin Woo Lee, Vincent Sitzmann Published: 2026-09-29 Categories: cs.SD, cs.CL, eess.AS arXiv: https://arxiv.org/abs/2609.36737v1 The vocal tract is the region of the human body responsible for filtering one's voice to create speech. In this paper, we present a differentiable and GPU accelerated acoustic simulator for the vocal tract. The differentiable simulator synthesizes speech by propagating sound along an acoustic tube model of the vocal tract, and via its gradients, can solve the inverse problem: reconstructing the shape of the vocal tract solely from the sound it produces. Although the inverse mapping between geometry and sound is notoriously non-convex, we discover that gradient descent succeeds with three technical contributions: (1) we design a frequency domain formulation of the vocal tract's fluid dynamics that is 70x more GPU parallelizable than finite differences in time, (2) we integrate a differentiable model for turbulence to synthesize consonants, and (3) similar to prior work in implicit neural representations (INRs) and neural fields, we find that parameterizing the geometry with a neural network accelerates convergence and escapes local minima that trap discrete representations. Because the simulator is differentiable, it is readily integrated with other deep learning pipelines to enable novel linguistics and medical imaging applications. (1) We demonstrate self-supervised autoencoding of vocal tract shapes across 11 languages, and (2) we couple our simulator with a generative model of MRI (magnetic resonance imaging) images to reconstruct one's moving vocal tract from only their speech without paired data. --- ### From Neurons to Conversation: Speech Brain-Computer Interfaces Authors: Moein Khajehnejad, Forough Habibollahi, Tommaso Boccato, Margarida Sousa, Michal Olak, Francesco Jamal Sheiban, Matteo Ferrante Published: 2026-09-29 Categories: cs.HC, cs.CL, q-bio.NC arXiv: https://arxiv.org/abs/2609.36736v1 Speech brain-computer interfaces (BCIs) aim to restore communication by transforming neural activity related to speech, language, or communicative intent into external outputs such as text, synthesized voice, or avatar control. Recent advances in intracortical and electrocorticographic recording, deep sequence models, and language-model-assisted decoding have enabled rapid progress, including high-performance attempted-speech decoding and increasingly naturalistic speech synthesis. Yet these achievements also reveal that speech BCIs are not simply neural-to-text decoders. They are adaptive clinical systems in which neural representations, recording hardware, decoding architectures, language priors, feedback, and user learning interact over time. Here, we synthesize speech BCI research from a system-level perspective. We first examine the neural substrates of speech and language, emphasizing their hierarchical, distributed, temporally structured, and non-stationary organization. We then examine recording and decoding choices, closed-loop adaptation, evaluation, clinical translation, and ethics. Across these domains, we highlight recurring trade-offs between signal resolution and invasiveness, low-level motor and high-level semantic targets, decoder accuracy and user agency, and language-model fluency and faithful neural evidence. We argue the next generation of speech BCIs should be evaluated not only by offline accuracy, but also by robustness across sessions, calibration burden, latency, uncertainty, usability, and safeguards against unintended decoding. By reframing speech BCIs as adaptive, user-centred systems, we outline the interdisciplinary priorities spanning speech neuroscience, neural engineering, machine learning, clinical practice, and neuroethics needed to move from proof-of-concept decoding toward reliable, expressive, and controllable communication neuroprostheses. --- ### BiFE: Search-Efficient Discovery of CPU-Only Branching Policies via LLM-based Bi-Fidelity Evolution Authors: Ce Zhang, Bin Zhang, Zhiwei Xu, Hao Chen, Xinyue Lu, Shanwei Fan, Yingxuan Teng, Guoliang Fan Published: 2026-09-29 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.36735v1 In branch-and-bound (B&B) for mixed-integer linear programming (MILP), branching variable selection critically impacts efficiency. Existing neural branching policies often require GPU inference, while CPU-efficient symbolic expressions lack the representational capacity for complex logic. Large Language Model (LLM)-generated code provides a flexible search space for designing lightweight branching rules with diverse algorithmic logic. To discover effective rules within LLM-based evolutionary frameworks, a core challenge arises: full B&B evaluation on real instances is prohibitively expensive, whereas offline imitation learning suffers from distribution shift. To address this, we introduce a Bi-Fidelity Evolutionary framework (BiFE). It employs low-fidelity imitation scores as a rapid pre-screener and selectively applies high-fidelity on-instance evaluation only to elite candidates, effectively balancing search efficiency with performance reliability. Experiments validate both the search efficiency of BiFE and the competitiveness of its discovered rules, which outperform the SCIP solver and other baselines on CPUs, and even surpass certain GPU-based neural policies. ---