# Research Papers — 2026-09 · bundle page 14 of 266 > Full-text markdown of papers #651–700 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/14 [← Previous bundle](https://aipod.dev/papers/archive/2026-09/page/13.md) [Next bundle →](https://aipod.dev/papers/archive/2026-09/page/15.md) ### Lasting Effects of Abstract Pretraining Beyond Perplexity Authors: Zachary Shinnick, Hemanth Saratchandran, Damien Teney, Anton van den Hengel Published: 2026-09-30 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.38764v1 Language models are typically pretrained from random initialization. Recent work challenges this convention, showing that a brief warm-up on abstract, algorithmically generated data can provide a better starting point for subsequent learning of natural language. In this paper, we show that in small language models, such a warm-up improves specific capabilities that are not reflected in language-modeling perplexity. Our warm-up uses an abstract stack-manipulation task that requires compositional and state-tracking capabilities. Allocating as little as 1% of pretraining tokens to this data improves multi-hop question answering by up to 3.9 F1 points on MUSIQUE, with additional gains on HOTPOTQA and 2WIKIMULTIHOPQA despite comparable language-modeling perplexity. Controlled experiments show that the warm-up substantially accelerates the acquisition of deeper reasoning chains. We also explore what drives this transfer. First, the structure of the data matters: replacing the stack task with a queue fails to produce the same gains. Second, the gains are specific: performance improves on sequential reasoning chains, with no consistent benefit on tasks that combine or compare independent facts. Third, timing matters: mixing abstract data with natural language is far less effective than an initial dedicated phase, and exposure after pretraining completely removes the benefits. The early advantage persists through billions of subsequent language tokens. These results show that early abstract training can reliably shape the capabilities language models later acquire. --- ### Adaptive-GEPA: Make Your Harness Fit Heterogeneous Requests Authors: Tianyu Chen, Yasi Zhang, Ruiyi Wang, Xinran Zhao, Taoran Li, Mingyuan Zhou Published: 2026-09-30 Categories: cs.SE, cs.AI arXiv: https://arxiv.org/abs/2609.38762v1 Reflective optimizers such as GEPA improve language model prompts from execution traces and evaluator feedback; full-program extensions can also rewrite tools and control flow. In practice, a user hands the same endpoint heterogeneous requests whose effective solutions require different tools, reasoning modes, and control flow. Optimizing one shared program leaves this division of work implicit in source-code search, while optimizing a separate program per request family fixes it beforehand. We introduce Adaptive-GEPA, which learns both how to divide requests and how to solve them. It evolves a router and a library of specialist programs under one search budget. The router's instructions, each specialist's description, and its program code are plain, human-readable text, edited from feedback. To combine branches, it aligns specialists by the requests they handle and inherits descriptions together with programs. On a fixed mixture of four task families, the reported Qwen3-8B run evolves four experts without supplying family labels to the router or reflection model; its routing matches the task partition on all 651 test requests. Its family-mean test score (x100) rises from 52.6 to 70.6, compared with 62.5 for GEPA's full-program adapter and 54.0 for GRPO at a nominal budget of 18,000 scored calls. These counts do not equate total compute. Figure 1 summarizes the learning curves, final test scores, and routing agreement. --- ### Event-Driven Refresh and Recurrence Memory to Reduce Stale Grounding in Referring Video Object Segmentation Authors: Abu Hanif Muhammad Syarubany, Jaehyun Jang, Siwoo Lim, Seungyeon Ryu, Chang D. Yoo Published: 2026-09-30 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.38758v1 Referring Video Object Segmentation (RVOS) aims to produce a pixel-accurate mask sequence for an object specified by natural language. Sa2VA combines a multimodal large language model with SAM2 for grounded segmentation; however, its inference typically grounds the query from a small fixed set of initial keyframes and then relies on propagation. In long or dynamic videos, this can cause stale grounding and persistent false positives when the object composition changes (e.g., distractors enter or the target disappears/re-appears). We propose Event-Driven Refresh + Recurrence Memory (EDRRM), an enhancement that selectively re-invokes Sa2VA only at stable change points. EDRRM triggers refresh boundaries using an EMA-smoothed event score computed from tracking-derived cues (births/deaths and coarse composition/layout changes) with temporal constraints. A recurrence memory further retrieves anchor frames via CLIP similarity to re-condition the model on re-appearance events. Experiments on Ref-DAVIS17, MeViS, and ReVOS show that EDRRM achieves a competitive accuracy-efficiency trade-off relative to fixed-window and FrameDiff-SSIM baselines, maintaining comparable or superior J&F scores at substantially lower average refresh-call budgets and reducing false-positive failures. End-to-end runtime analysis further confirms that the overhead introduced by tracking, CLIP-based recurrence matching, and the identifiability gate remains modest relative to the dominant Sa2VA inference cost, thereby validating the efficiency of the proposed pipeline. --- ### Self-Evolving Algorithm-Design Agents: Escaping In-Context Evolutionary Stagnation via Population-Curated Policy Optimization Authors: Chen Lu, Ke Xue, Siyuan Xu, Mingxuan Yuan, Chao Qian Published: 2026-09-30 Categories: cs.AI, cs.NE arXiv: https://arxiv.org/abs/2609.38757v1 Large language models are increasingly participating in complex real-world tasks in the form of algorithm-design agents, designing and refining algorithms. Many successful algorithm-design agents adopt pure in-context evolutionary frameworks, but they may quickly plateau in domains that require specialized knowledge. Parametric adaptation offers a way to internalize specialized knowledge, but conventional training requires abundant domain-specific corpora while high-quality algorithms are scarce in complex algorithm-design scenarios. In this paper, we propose sample-efficient parametric self-evolution where agents can explore and learn from self-generated algorithms. First, we characterize in-context evolutionary stagnation and analytically propose the Improvement Chain proposition, showing how learning successive self-generated algorithms can locally increase the likelihood of neighboring algorithms. Motivated by this local-transfer perspective, we further propose Population-Curated Policy Optimization (PCPO) to utilize a global population and a hybrid policy update scheme for retaining and reusing high-quality, diverse self-generated algorithms, shifting the policy towards stronger algorithms. In the task of learning rate schedule design for global placement in electronic design automation, trained only on 4 chip cases, PCPO outperforms the state-of-the-art in-context evolutionary methods (e.g., OpenEvolve and ShinkaEvolve) on average across 16 chip cases. With an 8B-size base model, PCPO achieves competitive performance compared to frontier closed-source models such as GPT-5.5. PCPO also reduces inference-time token cost by internalizing grounded domain knowledge and prompt distillation. Moreover, PCPO achieves significant speedups on four GPU kernel designs, with an average of 8.27$\times$ speedup against the PyTorch Eager baseline. --- ### Agentic Relative Camera Pose Estimation via Learned Ranking and Verification Authors: Zhining Gu, Shangjie Du, Weimin Qiu, Carl Olsson, Ping Liu, Meng Tang Published: 2026-09-30 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.38755v1 A wide range of approaches have been developed for camera pose estimation, including correspondence-based methods, end-to-end pose regression, and recent 3D geometric foundation models. Our key observation is that no single estimator is optimal for diverse challenges, such as wide baselines, lack of texture, appearance changes, and occlusions. Further analysis reveals substantial performance variation across both benchmarks and individual image pairs, with different estimators exhibiting complementary strengths. We introduce PoseAgent, an agentic framework for relative camera pose estimation that dynamically orchestrates pose estimators through learnable ranking and verification. Given an image pair, a profiling agent first extracts appearance, semantic, and geometric features relevant to pose estimation, e.g., scene type. A learned ranking agent then predicts the relative competence of multiple pose estimators given the image-pair profile. The top-ranked estimator is executed, and its predicted pose is assessed by a learned verification agent that estimates the corresponding pose error. When verification fails, PoseAgent adaptively invokes lower-ranked estimators until a candidate is accepted or the execution budget is reached. For pose verification, our verification network predicts pose errors more accurately than prior models. For pose estimation, PoseAgent improves AUC@5 degree up to 4.2% over the strongest standalone estimator on each of ARKitScenes, MegaDepth, ScanNet++, and RealEstate10K. On ARKitScenes, PoseAgent also outperforms VLM-based agents, which include a VLM ranker with the same verifier and fallback policy. These results demonstrate the effectiveness of our learned ranking and verification. --- ### Where the Evidence Lives: Auditing AI Companions' Self-Descriptions Authors: Seiya Ikeda, Shin-nosuke Ishikawa Published: 2026-09-30 Categories: cs.HC, cs.AI arXiv: https://arxiv.org/abs/2609.38753v1 Companion agents describe themselves: they remember, they understand their users, the relationship has changed them. We argue that such accounts, and the experience ratings that seem to confirm them, are checkable by users only where the evidence is theirs: in the agent's behavior, or in themselves. Where the evidence lives in the machinery, fluent self-description and moderately positive ratings do not establish that the mechanisms behind them ran. We demonstrate an audit procedure that sets an agent's self-description against its users' judgements and its implementation records, reporting each claim as supported, contradicted, or unresolved, and apply it to Lita, a proactive companion we built and deployed for a month with nine colleagues. Participants endorsed stylistic claims, withheld endorsement from relational ones, and rated memory at or above midpoint, while two of three memory layers had never executed their accumulation step. Memory-bearing agents should report what their self-descriptions cannot establish. --- ### Here the World in Stereo: Learning Dynamic Spatial Correspondence for Immersive Joint Video-Audio Generation Authors: Hanmo Chen, Chengcheng Liu, Tianxiao Chen, Zheyu Zhang, Siming Zheng, Jinwei Chen, Xu Yang, Cheng Deng, Bo Li, Peng-tao Jiang Published: 2026-09-30 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.38748v1 Recent joint video-audio generation models have achieved strong semantic correspondence and temporal synchronization. However, applications such as AR/VR and interactive gaming further require stereo audio to provide an immersive sense, which remains largely overlooked. Effective stereo audio requires the perceived sound location to evolve consistently with the motion of its corresponding visual source. We refer to this property as Dynamic Spatial Correspondence and propose StereoBind, a framework that binds visual source motion to stereo sound generation. StereoBind uses motion tracks to coordinate visual motion and stereo audio through three complementary mechanisms. Visual Motion Binding establishes source-aware audiovisual correspondence, the Spatial Track Encoder captures absolute source positions, and Residual Track RoPE models relative motion. For supervision and evaluation, we construct StereoWorld-29K, a large-scale stereo audio-video dataset with paired motion tracks, and StereoWorldBench for measuring audiovisual spatial consistency. Experiments show that StereoBind substantially improves spatial alignment in stereo audio generation over existing models while preserving overall audiovisual quality. --- ### Consensus-Aware Multi-Source Fusion for Reference-Guided Camouflaged Object Detection Authors: Junyang Xia, Luocheng Zhang, Wenwen Pan, Chifeng Zhu, Yang Yang, Xinchun Liu, Jiajun Ding Published: 2026-09-30 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.38747v1 Reference-guided camouflaged object detection aims to segment a target whose visual appearance closely resembles its surroundings by exploiting auxiliary reference samples. The task remains difficult because reference samples contain inconsistent target cues, while generic visual representations are not inherently aligned with the target specified by the references. To handle these problems, we present a consensus-aware multi-source fusion framework. Reference-Conditioned Dual-Backbone Fusion (RCDF) couples trainable PVTv2 query features with frozen DINOv3 representations and uses reference-conditioned correlation to select foundation-model evidence before multi-scale fusion. The framework also aggregates multiple references through cross-reference consensus aggregation and injects reference information at semantic depths matched to the query features. Extensive experiments demonstrate the effectiveness of the proposed method. The results further show that reference consensus, target-conditioned foundation features, and hierarchical decoding provide complementary improvements under the evaluation protocol. The source code will be made publicly available upon acceptance. --- ### Matisse: Evidence-Space Reasoning for Active 3D Reconstruction Authors: Xihang Yu, Kaichen Zhou, Lorenzo Shaikewitz, Clément Jambon, Xiao Zhan, Rajat Talak, Luca Carlone Published: 2026-09-30 Categories: cs.CV, cs.RO arXiv: https://arxiv.org/abs/2609.38746v1 How can a 3D reconstruction system acquire and retain useful information to understand the geometry of a scene from partial views under a limited computation budget? Existing active view acquisition methods typically estimate uncertainty over observed or instantiated geometry, limiting their ability to reason about unseen structure, while long-horizon reconstruction methods often retain redundant observations. We introduce Matisse, a training-free framework that unifies active reconstruction and keyframe selection by leveraging evidence provided by a pretrained generative 3D model. Matisse estimates Evidential Uncertainty from cross-attention evidence associated with 3D latent tokens and derives an Evidential Information Gain to guide both view acquisition and keyframe selection based on the expected reduction in posterior entropy. Matisse supports multi-object scenes through occlusion-aware, object-balanced aggregation and propagates uncertainty through intermediate latents to avoid full reconstruction during planning. Matisse reduces Chamfer distance by 12.7%, 3.8%, and 9.2% on GSO30, YCB-V, and Replica, respectively, relative to the best baseline on each dataset, and achieves a $1.50\times$ end-to-end speedup over the best active reconstruction baseline on GSO30 with the same reconstruction backend. In the GSO30 keyframe selection experiment for long-horizon reconstruction, Matisse achieves comparable Chamfer distance using 14% of the input views compared with Stream3D. --- ### Molecular Property Prediction under Structural Shift with Tabular Foundation Models Authors: Jinmo Lee, Dooho Lee, Minho Jeong, Jaemin Yoo Published: 2026-09-30 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.38744v1 Predicting molecular properties for compounds that differ structurally from labeled training molecules is important for drug discovery and materials design. Tabular foundation models (TFMs) offer a promising approach through in-context learning, but their performance under structural shifts and the value of molecular comparisons in this setting remain underexplored. We study structural generalization in molecular property prediction and introduce MolPAIR (Molecular Pair-Augmented In-context Refinement), a framework that combines molecule-level and molecular-pair contexts without task-specific parameter updates. A global tabular foundation model (TFM) first predicts a query's property from labeled molecular examples. A second frozen TFM predicts differences in prediction errors between the query and labeled reference molecules, using these comparisons to refine the initial prediction. Across 58 MoleculeACE and Polaris tasks, CheMeleon representations combined with TabPFN-3 already outperform each evaluated baseline on a majority of tasks. MOLPAIR further improves this predictor on 46 of 58 tasks, with gains across four molecular representations and three TFM backbones. These results show that explicit molecular comparisons can strengthen tabular in-context learning for structural generalization while keeping the molecular encoder and pretrained model weights fixed. The code and datasets are available at https://github.com/nums-ai/MolPAIR. --- ### Learning to Route in Visual Space via Multi-Step Embedding Retrieval Authors: Tianyu Chen, Mingyuan Zhou, Jiaxing Wu Published: 2026-09-30 Categories: cs.AI, cs.IR arXiv: https://arxiv.org/abs/2609.38743v1 LLM agents rely on retrieval tools to access external knowledge, yet visual agentic search remains severely bottlenecked by standard single-step retrievers. In current pipelines, the agent must issue text queries for every intermediate step, struggling when visual clues are difficult to describe or when the retriever fails to surface necessary intermediate evidence within its top results. We hypothesize that offloading multi-step navigation across the entire embedding space directly to the retrieval tool resolves this performance bottleneck. To study this systematically, we introduce VHOP, a flexible data generation framework and benchmark with five core difficulty levels testing both visual matching and search planning. Using this framework, we develop VHOP-Router, an end-to-end training pipeline---combining supervised fine-tuning, online imitation learning, and reinforcement learning---that transforms a standard embedding model into an autoregressive multi-step retriever. Operating directly in the visual latent space, VHOP-Router retrieves linked image chains in a single tool call without requiring the agent to formulate intermediate text queries. Experiments show VHOP-Router boosts retrieval performance from under 5\% to 76.3\%. In agentic search, it improves task success rates by 52.7\% and reduces the average token length by 61\% from 1886 to 728, whereas upgrading the agent yields only a 3.7\% gain. Compared to a strong baseline where the agent retrieves the top 50 results per step, VHOP-Router maintains superior performance while reducing in-context images by $23\times$ and cutting the cumulative API payload by $35\times$. The models also generalize robustly to unseen difficulty levels and realistic test sets. Ultimately, VHOP and VHOP-Router provide an efficient and effective solution for visual agentic search that leaves native LLM capabilities entirely intact. --- ### Code to Control: Synthesizing Parameterized Reactive Controllers Authors: Zergham Ahmed, Joshua B. Tenenbaum, Chris Bates, Samuel J. Gershman Published: 2026-09-30 Categories: cs.AI, cs.LG arXiv: https://arxiv.org/abs/2609.38733v1 Recent LLM-based approaches to control either invoke a language model to select actions or synthesize world models that require planning at every decision, introducing latency that can limit real-time use. We introduce Code to Control, an approach that synthesizes Python controllers which execute directly as policies. Code to Control separates program structure from parameters. An LLM synthesizes the controller structure, while derivative-free search fits its parameters for continuous control using feedback from the environment. Once learned, the resulting controllers require neither LLM inference nor planning at decision time, enabling real-time gameplay and, under our timing protocol, faster action selection than a PPO policy. Across a suite of Atari games, Flappy Bird, and MuJoCo tasks, Code to Control outperforms planning-based program synthesis methods, remains competitive with deep reinforcement learning while using fewer environment interactions, transfers across substantial changes in environment dynamics, and scales to complex locomotion tasks. --- ### Anchor-ECC: Local Integrity Checking for Watermarked LLM Outputs via Error-Correcting Codes Authors: Zewei Deng, Muhammad Siddeek, Liyan Xie, Mohamed Seif, Mengdi Wang, H. Vincent Poor, Andrea Goldsmith Published: 2026-09-30 Categories: cs.CR, cs.CL arXiv: https://arxiv.org/abs/2609.38722v1 LLM watermarking has become an effective approach to distinguishing AI-generated text from human-written text by embedding detectable patterns during generation. However, a small post-generation edit may change the meaning of the text without removing its overall watermark signal, creating a risk that the modified content is still attributed to the original model. We propose Anchor-ECC, which incorporates the error-correcting code (ECC) constraints and explicit boundary anchors into the watermark structure and pairs them with a dynamic-programming decoder to detect and localize post-generation edits. Across Qwen3-8B, Mistral-7B-Instruct-v0.3, and OPT-125M, the approximate-hard setting achieves about 99.7% block-level true positive rate (TPR) with at most 7.6% false alarm rate (FAR) for edit detection under mixed insertions, deletions, and substitutions, while preserving the distinction between watermarked outputs and unwatermarked text. Additional quality experiments identify lower-perplexity configurations that retain strong edit-detection performance. Together, these results extend LLM watermarking from source identification to local integrity verification while supporting configurable trade-offs between detection reliability and generation quality. --- ### UniEvo-VL: An On-policy Self-Distillation Training Recipe for Multimodal Model Self-improvement Authors: Fang Wu, Da Xing, Yanjie Huang, Junxi Wang, Ji Wang, Hejia Geng, Guancheng Wan, Bowen Zuo, Xiaomin Li, Shixiang Tang, Xinyu Xiang, Zehong Wang et al. Published: 2026-09-30 Categories: cs.AI, cs.CV arXiv: https://arxiv.org/abs/2609.38721v1 Modern multimodal models bring generation and understanding into a single unified system, which enables them to provide and learn from their own feedback. Motivated by this unified capacity, we introduce UniEvo-VL, a self-evolving framework for multimodal models to learn from this constructive self-correction feedback during test-time compute. Instead of relying on a separate, often larger, teacher, we leverage their self-critiques as privileged information and ask a single multimodal model to act as both teacher and student with different contexts. The student only sees the vanilla question, while the teacher conditions on the privileged critique. Then training minimizes the per-state divergence between their denoising diffusion distributions over the student's own sampling trajectories. Experiments demonstrate that UniEvo-VL improves the image generation capabilities of multimodal models, while maintaining their sensitivity to additional reflection information. Specifically, we build on top of the open-source Qwen-image-2512 and observe a significant performance gain from 0.747 to 0.808 on GenEval and from 32.97 to 35.53 on GenEval2 Soft-TIFA. Moreover, attempts with more powerful external critics (e.g., GPT5.6-Luna) show that multimodal models with strong judge capabilities can anticipate a higher self-evolving ceiling. Last but not least, mixed text-rendering outcomes show that our self-improvements may not be uniform across different tasks. Our study aims to shed light on the current hot recursive self-improvement research line to enhance the user experience when using multimodal models without external supervision or guidance. --- ### MetaSteer: Context-Conditioned, nonlinear Steering via Attention-Projection Adaptation Authors: Mehdi Jafari, Hao Xue, Flora Salim Published: 2026-09-30 Categories: cs.CL arXiv: https://arxiv.org/abs/2609.38718v1 Steering large language models typically relies on linear, context-independent interventions in activation space, an assumption that recent work has challenged and that can induce an information bottleneck when a fixed representation must encode many behavioral distinctions. We introduce MetaSteer, a method that learns nonlinear interventions with context-dependent effects and applies them to attention projection matrices, producing activation effects that vary with the input context by construction and requiring no linear concept-geometry assumption. Framed as preference-based optimization, MetaSteer is trained once on a pooled preference corpus and transferred zero-shot to unseen concepts and out-of-distribution contexts. We find that, despite using low-rank adapters, MetaSteer induces structured, context-dependent changes in hidden-state trajectories while partially preserving aspects of their local trajectory dynamics, including velocity and curvature. We evaluate MetaSteer on three controlled text-generation benchmarks and three agentic settings across multiple model families and scales. MetaSteer matches or outperforms strong task-specific steering baselines on most aggregate comparisons in the zero-shot regime. Across the evaluated settings, stronger text-generation steering is associated with stronger agentic steering performance. We further discuss geometric trajectory effects, capability retention, and safety considerations raised by transferable steering. --- ### Soft Spatial Reasoning Authors: Rafi Ibn Sultan, Md. Sajid Alam Chowdhury, Saleh Zare Zade, Chengyin Li, Prashant Khanduri, Marco Brocanelli, Dongxiao Zhu Published: 2026-09-30 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2609.38717v1 Large Vision-Language Models (LVLMs) commonly perform spatial reasoning through chain-of-thought (CoT), encoding intermediate reasoning as autoregressive sequences of discrete language tokens. Such hard thinking requires committing to a single token at each step, even when the correct spatial interpretation remains uncertain. This early commitment constitutes premature discretization: an incorrect token selection can propagate errors through subsequent reasoning. We propose Soft Spatial Reasoning, a post-training framework that introduces soft thinking for spatial tasks in LVLMs. At each intermediate reasoning step, the LVLM forms a continuous soft state by mixing token embeddings rather than selecting a single token, allowing multiple candidate continuations to influence the next step. The appropriate degree of softness, however, can vary across reasoning steps: retaining multiple candidates may preserve a useful spatial interpretation, but if those candidates imply conflicting spatial relations, mixing them may interfere with subsequent reasoning. At the core of Soft Spatial Reasoning is AdaptSoft, a controller that uses the current hidden state and predictive uncertainty to adapt the degree of softness at each reasoning step. To train AdaptSoft, we introduce a gradient-alignment learning objective that provides a step-specific learning signal for softness control without intermediate reasoning supervision. Across diverse spatial benchmarks, Soft Spatial Reasoning outperforms hard and fixed-soft CoT baselines using the same backbone, as well as a range of existing LVLMs. The source code is available at https://github.com/rafiibnsultan/Soft_Spatial_Reasoning --- ### SpatialCORE: Confidence-Aware Grounded Spatial Reasoning in Large Vision--Language Models Authors: Rafi Ibn Sultan, Xiangyu Zhou, Md. Sajid Alam Chowdhury, Chengyin Li, Prashant Khanduri, Marco Brocanelli, Dongxiao Zhu Published: 2026-09-30 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2609.38716v1 Large Vision-Language Models (LVLMs) have made remarkable progress across visual perception tasks, yet spatial reasoning remains a persistent weakness, especially for questions that require reasoning over visual space. Recent spatial-reasoning methods incorporate generated grounding, where models predict bounding boxes, masks, or other localization outputs for task-relevant objects as part of their reasoning trace. However, these approaches typically optimize final-answer correctness alone, allowing correct answers to be rewarded even when the model does not reason from confidently localized task-relevant objects. We introduce SpatialCORE (Spatially COnfident REasoning), a post-training framework that turns the model's own confidence in generated grounding into a learning signal for spatial reasoning. Its central idea is to reinforce grounding that is both accurate and confident, encouraging the model to reason from confidently localized task-relevant objects. SpatialCORE realizes this through a self-regulating spatial reward that weights each predicted bounding box's matching quality by its coordinate-token confidence. An answer gate further ties grounding optimization to final-answer correctness. SpatialCORE achieves state-of-the-art results among open-source and specialized spatial reasoning models across diverse benchmarks, and transfers effectively in zero-shot settings to unseen data distributions. The source code is available at https://github.com/rafiibnsultan/SpatialCORE. --- ### Hard-Region Supervision: #1 on the Waymo Open Dataset 2D Video Panoptic Segmentation Leaderboard Authors: Jinghan Yang Published: 2026-09-30 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.38714v1 We describe our winning entry to the Waymo Open Dataset 2D Video Panoptic Segmentation Challenge. The task asks for a semantic class at every pixel of every frame and, for countable objects, an identity that holds across 100 frames and across five overlapping cameras. We build on DVIS++, a cascade of a segmenter, a tracker, and a refiner, as our baseline. We propose hard region supervision (HRS) to improve the baseline. In particular, we use the baseline to define the hard region as where it makes mistakes, and design a loss and an auxiliary prediction head for this region. The auxiliary head is used only in training and removed at test time, so at inference the model trained with HRS has the same architecture as the baseline. In addition, we propose three test-time steps that further improve the results: a two-model ensemble, a merge of the segmenter's output into the final panoptic map, and cross-camera identity linking. On the challenge test set, our entry reaches 0.3547 wSTQ, 0.2071 wAQ, and 0.6075 mIoU, ranking first on all three metrics. It is 3.6 wSTQ points ahead of the second entry and 2.4 points ahead of our DVIS++ baseline. --- ### Staying on Task: Testing the Foundations of Long-Horizon Agent Reliability Authors: Jeffrey Willette, Krishna C. Puvvada, Boris Ginsburg Published: 2026-09-30 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.38712v1 Long-horizon agentic workflows require models to sustain repeated state-dependent actions all while the context grows, sub-task complexity changes, and new data arrives. Each situation represents an independent axis along which an agent may fail. An agent reconciling a long ledger, for example, must repeatedly read its state, update the correct record, and preserve alignment across thousands of outputs. A model may accept the entire ledger yet lose its place or stop applying the operation consistently as generation proceeds. We introduce Long-Transduction, a controlled diagnostic that tests a model's ability to stay on task during long generation while continuously reading, mutating, and outputting input-context dependent operations such as arithmetic, sorting, variable lookups, and table transformations. Long-Transduction evaluation independently varies local task complexity, input data formatting, and context length isolate failures along each axis. We evaluate seven open-weight models, finding a 62.8\% decrease when scaling context length from 4-128K, a 36.5\% decrease when varying input format, and a 39.9\% decrease by increasing local task complexity. Together, these failures represent critical liabilities in long-horizon agentic workflows. --- ### SCALE: Synthetic Calibration via Agreement Labeling in Embedding Space Authors: Wenjun Liu, Saeed Hassanpour Published: 2026-09-30 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.38705v1 Foundation models for computational pathology are usually evaluated using AUC and accuracy, while calibration is often left untested. This matters because a model can be accurate on average but still assign overly confident probabilities to cases that are difficult even for pathologists. We study calibration across eight pathology foundation models. Using pathologist agreement as a measure of diagnostic difficulty, we find that calibration error is consistently higher on low-agreement cases than on high-agreement cases. This pattern is not apparent from aggregate expected calibration error (ECE) alone. We then propose synthetic agreement calibration, a method for improving calibration without collecting multi-annotator labels. Given a trained linear probe, we select high-confidence embeddings as class anchors and interpolate between anchors from opposite classes. The interpolation weights encode a continuous notion of diagnostic ambiguity, which we use as a synthetic agreement signal to retrain the probe with agreement-aware label smoothing. On MHIST, which includes annotations from seven pathologists, synthetic agreement calibration recovers most of the calibration improvement obtained by label smoothing based on real pathologist agreement, while substantially reducing low-agreement ECE relative to the uncalibrated baseline. Discrimination metrics are preserved. On PatchCamelyon and BreakHis, public histopathology datasets without multi-annotator labels, the method improves calibration across the evaluated foundation models, whereas annotator-dependent approaches cannot be used without additional expert annotation. --- ### Budget Boundary Effects in Test-Time Mathematical Reasoning Authors: Guilin Zhang, Ziqi Tan, Wulan Guo, Kai Zhao, Hongyun Yang, Mei Luo, Qi Ning, Feng Yang Published: 2026-09-30 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.38699v1 A cumulative token cap can fall inside a mathematical derivation, forcing a test-time controller to choose between stopping at the cap (strict) and allowing the current attempt to finish (advisory). We measure this boundary choice with paired offline replays of 19,200 public traces: 120 AIME, BrUMO and HMMT problems and two archive configurations of one model. Candidate order and a 16-attempt cap are fixed, and answer selection is blind to reference answers and correctness labels. Three findings emerge. First, at the 4k cap, most advisory accuracy gains replace abstention with a correct answer; strict stopping pays for an unfinished prefix that the completed-only selector cannot use. Second, comparisons along realized cost differ from same-cap comparisons: advisory 4k in low has higher accuracy than strict 8k at comparable mean completion cost, while in high its observed accuracy is 0.42 points below strict 32k using 59% of its mean tokens. These aggregate comparisons do not establish equal-compute superiority or accuracy equivalence. Third, increased candidate coverage does not guarantee higher answer accuracy: a log-probability selector loses accuracy while coverage rises, including after a source-grade consistency repair. Same-cap majority-accuracy differences shrink below 1.3 percentage points at 32k. Budget curves should jointly state the cap, realized cost, eligible candidates, stopping rule and selector information. --- ### Cascadia: A Control-Plane-Free Alternative to Hyperconverged AI Infrastructure Authors: Matias Parij, Pawan Paudel, Tate Berenbaum, Muthaiah Venkatachalam Published: 2026-09-30 Categories: cs.DC, cs.AI arXiv: https://arxiv.org/abs/2609.38697v1 We present Cascadia, a system for serving large language models on fleets of commodity Intel AIPCs using their CPU, integrated-GPU, and NPU resources. Every node embeds ingress, scheduling, and execution; inference requests require no dedicated routing control plane. Nodes join a libp2p QUIC mesh using CA-issued ed25519 admission certificates, gossip signed capabilities, exchange live load over direct peer streams, and route OpenAI-compatible requests to eligible peers. An operator-run certificate authority handles admission and fleet management outside the inference path. Three serving modes share one interface: whole-model execution on one node, load-balanced replicas, and pipeline-sharded chains using the compilation and speculative decoding mechanism of our companion paper. Optional KV-cache mobility reuses compatible conversation prefixes after a routing move, with cold recomputation on a miss. Signed response receipts and hash-chained logs support provenance and audit. A three-node Phi-3.5-mini NPU testbed delivered 3.10x the response throughput of its one-node configuration under ten concurrent requests; a separate four-node deployment recorded 4.06x the throughput of direct single-node serving. Paired latency observations, runtime measurements, and internal functional checks characterize the tested configurations. We compare Cascadia with IBM, Nutanix, VMware, and HPE platforms on deployment footprint, hardware requirements, scheduling, scaling, licensing, and trust, using vendor documentation. The paper repository provides benchmark scripts, curated measurements, and a claim-to-evidence map. --- ### Always-On Experimentation Authors: Ricardo J. Sandoval, David Arbour, Avi Feller, Michael I. Jordan Published: 2026-09-30 Categories: stat.ME, cs.AI arXiv: https://arxiv.org/abs/2609.38695v1 Generative AI has dramatically accelerated the rate at which new treatments---from novel pharmaceuticals to online marketing campaigns---can be conceived and deployed. As a result, modern experimentation platforms often run continuously, with treatments added as they are ready and removed when they underperform. We formalize this "Always-On" experimental setting, in which treatments can be dynamically generated, added to, and removed from a running experiment, and study the statistical problem of deciding whether to accept or reject each treatment while controlling for the false discovery rate. We develop sequential tests that achieve time-uniform Type-I error control under arbitrary stopping times and "predictable" treatment schedules. Our approach builds on the testing-by-betting framework: we construct test supermartingales for testing the average treatment effect of each treatment, and show that the construction of these test supermartingales is growth-rate optimal in an almost-sure sense. --- ### No Corners Cut: State-Grounded Transitions for Mid-Stream Prompt Switches in Video Generation Authors: Zejing Rao, Ketong Ren, Xiaoqiang Liu, Yiping Meng, Guoxin Zhang, Fan Tang Published: 2026-09-30 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.38691v1 Streaming video generators allow users to dynamically modulate video synthesis via mid-stream prompt switching. Existing streaming methods can respond to the updated instruction while still cutting corners, prematurely realizing goals or taking heuristic shortcuts that bypass necessary intermediate state changes needed for a plausible transition. In this study, we present SEGUE, a novel framework that makes this process explicit and trains the generator to execute these transitions faithfully. At each switch, a training-free planner parses the latest frame and prompts, writes a few segue prompts with roles and durations, and then hands control back to the user's prompt. Furthermore, to address the inherent difficulty of training causal models on short-lived temporal schedules without corrupting preparatory supervision, we introduce SPANDMD, which evaluates each active prompt using the full rollout as temporal context while retaining its DMD residual only within the prompt's assigned span. On OpenTrans-360, a benchmark of 1,800 switches that scores how the old state exits and the new one begins, SEGUE ranks first on all eight transition metrics and raises the overall score over the strongest baseline from 0.866 to 0.887. It also ranks first on four of six instruction-response metrics of StreamAV-Bench, while the planner transfers to frozen autoregressive generators without retraining. --- ### GATE-ST: Gene-Aware Text-image Encoder for Spatial Transcriptomics Authors: Lucas Ni, Jian Luo, Wentao Huang, Chao Chen Published: 2026-09-30 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.38690v1 Spatial transcriptomics enables spatially resolved gene expression analysis from slide-level images while preserving morphological features, providing valuable information for studying disease mechanisms and developing treatments. However, spatial gene expression profiling typically requires expensive and time-consuming tests. While existing image-based prediction optimizations mostly revolve around including positional embeddings and further image-based changes, text-based optimizations remain relatively unexplored. We present GATE-ST, which incorporates text-based inputs into image-based spatial gene expression predictions. With this approach, generated text descriptions of genes are utilized to better spatial transcriptomics prediction results. Gene summaries are put through a text encoder, generating embeddings that integrate with image embeddings through cross-attention layers to align with morphological features. We demonstrate the effectiveness of such text inputs by benchmarking performance against random gene embeddings and multiple other image-text fusion architectures, and show that GATE-ST outperforms these alternatives. Our results demonstrate the effectiveness of GATE-ST in pathology imaging, which may greatly reduce the time and cost of accurate spatial transcriptomic predictions, proving the potential of text-guided spatial gene expression prediction. --- ### EPIC: Epipolar-Consistent 360° Immersive Stereo Video Generation Authors: Debabrata Mandal, Dongdong Fu, Jonathon Miller, William Villareal, Xi Peng, Praneeth Chakravarthula Published: 2026-09-30 Categories: cs.CV, cs.HC arXiv: https://arxiv.org/abs/2609.38689v1 Immersive displays can enable rich and diverse virtual experiences. Manually authoring every possible experience to realize this potential, however, is prohibitively expensive, difficult to scale, and impractical. Generative AI models could remove this bottleneck, but today's models are built for conventional displays and cannot generate the high-resolution, stereoscopic $360^\circ$ content required for immersive viewing. Further, temporal and stereo inconsistencies that may be tolerable on conventional displays can become highly disruptive when viewed through an immersive headset. Here, we address this gap with a zero-shot generative pipeline that extends existing video diffusion models into 4K stereoscopic $360^\circ$ videos. Inspired from binocular vision and depth perception, we develop an epipolar-aware $360^\circ$ image matching metric that captures the temporal and stereo geometric inconsistencies across views. We then use this metric as a preference signal for direct preference optimization with limited training data. Our work enables $360^\circ$ stereo video generation and provides a scalable path for bringing generative content to immersive displays, allowing diverse mixed reality experiences on demand. --- ### Concept-Grounded Attention: A Controlled Evaluation of Graph-Injected Attention, Temporal Versioning, and Epistemic Status Authors: Sachin Dev Duggal, Pradyumna Swarnalatha Ramanna, Alexandros Vassiliades Published: 2026-09-30 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.38684v1 Knowledge-intensive language-model systems typically represent external knowledge as text chunks or static graphs, with limited support for concept evolution, point-in-time reasoning, and distinctions between validated and inferred knowledge. We introduce the Concept Lifecycle Model (CLM), which represents concepts as persistent, graph-grounded, temporally versioned entities with explicit provenance and epistemic status, and Concept-Grounded Attention (CGA), which injects concept-graph structure into transformer computation through graph-biased self-attention (Form A) and gated cross-attention over concept nodes (Form B). We evaluate the framework in controlled settings using disabled-mechanism baselines. On 200 MuSiQue and HotpotQA questions with retrieval fixed, concept-graph retrieval recovers explicit multi-hop paths but does not improve evidence recall. Form A appears to steer attention, with 2.76 times more attention on gold than distractor concepts, but the same ratio occurs when Form A is disabled; the learned bias is negligible and no answers change. An identity-preserving Form B improves F1 from 0.188 to 0.221, but control concepts yield 0.213, indicating that most of the gain reflects added capacity. On LongMemEval, explicit temporal representation improves answer accuracy by 13 to 25 points across all tested generators, up to 122B parameters, while simplified CLM version resolution performs similarly to dated serialization because concept identity is not established reliably. On a synthetic source-independence task, protocol-derived epistemic status reduces unsupported assertions from 28% to 0.1% in a fine-tuned small model and from 19-68% to 0-5% in 72-122B models. Overall, the results support making temporal validity and epistemic status explicit, while showing that graph-attention diagnostics are not informative without disabled-mechanism controls. --- ### Unveiling the Value of Motion for Cinematic Camera Trajectories Authors: Ziqi Zhou, Yujian Yuan, Laura Sevilla-Lara Published: 2026-09-30 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2609.38683v1 Cinematic camera motion is a fundamental storytelling tool, defined not only by where the camera is positioned in the scene, but also by how it moves in terms of direction and speed. Recent work on camera trajectory generation and alignment to text relies on pose-centric representations. While in principle a network could derive direction of movement and speed, we find that in practice this might not happen. In fact, in this paper we discover that decomposing the camera trajectory representation from the traditional per-frame poses to direction and speed has surprising benefits across multiple tasks, including trajectory-to-text alignment as well as text-to-trajectory generation. To accurately evaluate the former, we introduce a simple and reliable protocol that overcomes the limitations of prior evaluation baselines. For the latter, building on this representational insight, we propose a novel generative model for camera trajectories, CineGEN, that achieves superior performance across a variety of metrics. We also propose a novel dataset, CineScript, containing movie clips that are enriched with scene descriptions as well as higher-level metadata. This novel data allows us to test models' ability to capture high-level cinematographic information. We show that, despite its simplicity, representing camera trajectories through direction and speed not only helps numerically to achieve better alignment and generation, but also inherently encodes complex directorial intent. --- ### ReGain: Restoring Subject Fidelity in Personalization on Synthetic Images Authors: Shubhang Bhatnagar, Ishan Bhatnagar, Viraj Shah, Narendra Ahuja Published: 2026-09-30 Categories: cs.CV, cs.AI, cs.LG, eess.IV arXiv: https://arxiv.org/abs/2609.38680v1 Text-to-image diffusion models are personalized to a subject by DreamBooth fine-tuning on a handful of its images. Increasingly, these images come from a diffusion model rather than a camera. We show that fine-tuning on such synthetic images degrades subject fidelity, producing oversaturated color and excess high-frequency detail. To isolate the cause, we fine-tune two models from the same base model with the same DreamBooth recipe, one on real photos of a subject and one on synthetic images of that subject generated by the first. We trace the degradation to classifier-free guidance (CFG). For the model personalized on synthetic images, the angle between the conditional and unconditional noise predictions, and with it the norm of their difference, is much larger than for the model personalized on real photos. This inflation grows toward high frequencies and also appears at other prompts semantically close to the subject, such as its class noun, but not at unrelated ones. We propose ReGain, a training-free correction applied at sampling time that measures how much each frequency band of the guidance is inflated relative to the base model and scales that band down accordingly. ReGain needs no real photos. On Stable Diffusion v1.5, ReGain closes 51-64% of the subject-fidelity gap to the model personalized on real photos, as measured by DINO, DINOv2 and CLIP-I. It also improves subject fidelity on SDXL and SD 3.5 and preserves text alignment on all three backbones. --- ### Point2Part: Unified 3D Partitioning from Point Prompts Authors: Hao-Tang Tsui, Yu-Rou Tuan, Xiaoxuan Ma, Nicolas Ugrinovic, Takaaki Shiratori, Kris Kitani Published: 2026-09-29 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.38180v1 Existing 3D part decomposition methods do not necessarily partition the original shape into non-overlapping parts that collectively cover the entire shape, allowing overlaps or gaps that hinder downstream part-level applications. We instead formulate part decomposition as a joint partitioning of the entire shape, where the predicted parts are non-overlapping and jointly recover the entire shape. Our key insight is that part decomposition should consider all desired parts jointly, rather than modeling each part independently. To this end, we develop a promptable model for 3D part decomposition from images or meshes. Users can specify desired parts through 3D point prompts for controllable decomposition. Given one point prompt per desired part, our model produces the corresponding parts as a complete partition of the entire shape. We build on a pretrained 3D generation model and first obtain a shape latent from either an input image or mesh. We then introduce a prompt encoder that maps each 3D point prompt to a part token while attending to the shape latent. To decode the desired parts, we propose a novel part decoder jointly scoring the entire shape against all part tokens in a coarse-to-fine manner, assigning every position within the shape volume to exactly one part. We perform part decomposition in this shared shape latent space, enabling a unified model for image-to-part generation, mesh-to-part generation, and part segmentation. Our method outperforms existing works on all part-quality metrics across all three tasks, and improves compatibility among parts by an order of magnitude over previous SOTA methods. Code and models will be released. --- ### Skill-Space Shooting for Autonomous Robot Policy Improvement Authors: Zihang Rui, Renhao Wang, Haoxu Huang, Yang Gao Published: 2026-09-29 Categories: cs.RO, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2609.38178v1 Robots deployed in the physical world must be able to improve beyond their initial training as they encounter new situations and failures. For this improvement to scale across tasks, it must make effective use of experience without requiring human demonstration of each correction. Recent agentic systems offer a way to reduce this reliance on human effort by using foundation models to autonomously compose learned behaviors to complete tasks. Yet completing tasks this way does not itself teach a task policy to overcome its own failures; that requires turning these behaviors into learnable corrections for the policy. Our insight is that many such corrections are familiar short behaviors, or skills: they recur across tasks and describe actions that foundation models can reason about from a scene. We introduce skill-space shooting, which uses foundation model guidance to explore corrections through these reusable skills and turn successful trials into policy improvement. Real-world experiments show repeated improvement in policies acting autonomously, while skills can also be shared to reduce the teaching needed to improve on new tasks. By making reusable skills a source of corrective supervision, skill-space shooting enables scalable and generalizable policy improvement within and across tasks. Additional results and videos at https://skill-space-shooting.github.io. --- ### Imagine3D-LLM: Teaching MLLMs to Imagine 3D Scenes Before Answering Authors: Jaewoo Jung, Hyeonseo Yu, Honggyu An, Jisang Han, Mungyeom Kim, Minkyeong Jeon, Heeseong Shin, Wonjun Moon, Federico Tombari, Daniel Barath, Marc Pollefeys, Seungryong Kim et al. Published: 2026-09-29 Categories: cs.CV, cs.CL arXiv: https://arxiv.org/abs/2609.38177v1 Reasoning about the 3D world from multi-view images remains a fundamental challenge for Multimodal Large Language Models (MLLMs). While modern MLLMs handle single-image inputs effectively, they struggle to integrate evidence across viewpoints into a coherent 3D understanding. A growing body of work attempts to close this gap by injecting 3D awareness into MLLMs, either by boosting fine-grained pixel-level cross-view correspondence or by fusing features from 3D geometry foundation models, yet a substantial gap to human reasoning persists. In this work, we revisit human spatial reasoning, which suggests that rather than relying on fine-grained geometry cues, humans roughly identify common objects across views, infer the relative geometry between viewpoints, and assemble a coarse 3D layout of the scene. Inspired by this process, we introduce Imagine3D-LLM, an MLLM that learns to assemble a similar compact 3D representation of the scene and conditions its answer on this representation. Concretely, we append a small set of learnable summary tokens after the image tokens, decode them into a compact 3D Gaussian Splatting representation supervised by a photometric reconstruction loss, and train jointly with the standard next-token prediction objective. Notably, although only the summary tokens receive direct reconstruction supervision, this objective also induces stronger cross-frame correspondence within the LLM's underlying image features, suggesting that learning to reconstruct propagates 3D-aware signals throughout the model. As a result, Imagine3D-LLM consistently outperforms prior approaches across multiple spatial reasoning and 3D understanding benchmarks, suggesting that imagining the scene can be more effective than being told its pixel-wise geometry. --- ### Breakdown of Local Denoising as Semantic Speciation Authors: Guangkuo Liu, Mert Okyay, Yifan F. Zhang, Fangjun Hu, Rahul Nandkishore, Xun Gao Published: 2026-09-29 Categories: cs.LG, cond-mat.dis-nn, cond-mat.stat-mech arXiv: https://arxiv.org/abs/2609.38176v1 The dynamics of generative models exhibit two apparently distinct temporal windows: a speciation window, in which a sample commits to a semantic class, and a nonlocality window, in which local context windows become insufficient for generation. Motivated by evidence of their near-concurrence in a variety of frontier models, we investigate their relationship through the spatial distribution of semantic information. Under a "common cause" hypothesis, we prove that the nonlocality window must lie in the speciation window. This hypothesis postulates that semantic labels explain a fraction of the correlations between distant tokens, a condition that is natural for many real datasets. We further give conditions under which both windows shrink to a single limiting time as system size grows, defining a "phase transition", and verify this behavior analytically in Gaussian mixtures. Together, these results identify conditions under which semantic information explains the concurrence of speciation and nonlocality, connecting two complementary perspectives on the emergence of semantic structure in generative modeling. --- ### Counterfactual Video Generation Enables Scalable Humanoid Loco-Manipulation Authors: Zihan Wang, Zhen Wu, Pieter Abbeel, Rocky Duan, Jitendra Malik, Carmelo Sferrazza, C. Karen Liu, Guanya Shi, Angjoo Kanazawa Published: 2026-09-29 Categories: cs.RO, cs.CV, cs.GR arXiv: https://arxiv.org/abs/2609.38172v1 Teaching humanoids loco-manipulation skills, such as carrying diverse objects, via visual imitation is a promising path toward generalist robots. However, collecting diverse, high-quality interaction videos, such as clips that clearly show a person's full body and unoccluded interactions with objects, poses a practical barrier to scaling this approach. We propose PRISM, a real-to-sim-to-real framework that overcomes this limitation by amplifying a handful of real videos into a large, diverse training set. PRISM first generates hundreds of diverse "counterfactual" human-object interaction videos via video-to-video (V2V) generation from a few exemplar real videos. Our contact-anchored real-to-sim pipeline then reconstructs both human and object motions, retargeting this imperfect video data into physically plausible trajectories. The intra-class variability across these counterfactual videos lets us train a single policy that generalizes to unseen objects within each category. We demonstrate the full pipeline by deploying this policy on a real robot without any real-world fine-tuning. Using only onboard depth observations, our humanoid picks up, carries, and drops objects, including boxes, barrels, bins, and balls, across novel instances, sizes, and initial configurations. --- ### Adversarial Training for Pixel Diffusion Authors: Xin Lin, Zhifei Zhang, Yuqian Zhou, Haitian Zheng, Zhe Lin, Ming-Hsuan Yang, Truong Nguyen Published: 2026-09-29 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.38170v1 Pixel diffusion models generate RGB images directly, avoiding the bottleneck of an autoencoder, yet their outputs still systematically underrepresent fine-scale natural-image statistics. We show that adversarial learning provides an effective post-training correction for this deficiency. Starting from a pretrained model, we retain its original diffusion or flow-matching objective and add an adversarial loss to the predicted output at non-high-noise timesteps, leaving the model architecture and sampling procedure unchanged. To our knowledge, this is the first systematic study of adversarial post-training for pixel diffusion. Across two pixel backbones, the method jointly improves distribution fidelity, coverage, prompt alignment, and perceptual quality. We further investigate why it works. Frequency-band and power-law analyses show that the original models systematically underproduce natural-image high-frequency content, while adversarial post-training restores this missing spectral power. In contrast, perceptual loss also increases high-frequency content but sacrifices distribution fidelity and prompt alignment. Nearest-neighbor, recall, and matched no-GAN SFT controls further rule out memorization, mode dropping, and additional optimization as simple explanations. Finally, we examine the boundary of this effect. Under the tested latent diffusion configurations, the same procedure does not produce comparable joint gains and adds almost no decoded high-frequency power. These results identify direct output access to the image statistics being corrected as a key factor governing when adversarial post-training succeeds. --- ### STEPQuant: When and Where Errors Matter in Delta-Rule Recurrent State Quantization Authors: Bingchen Yao, Haobo Xu, Haokun Lin, Yichen Wu, Ziyu Guo, Renrui Zhang, Zhichao Lu, Zhenan Sun, Ying Wei Published: 2026-09-29 Categories: cs.CL, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2609.38169v1 Linear attention replaces growing KV caches with fixed-size recurrent states, yet these persistent states can become a substantial memory bottleneck under concurrent serving. Directly quantizing recurrent states to low precision often leads to severe accuracy degradation, as quantization errors propagate through successive state updates. We discover that the impact of these errors depends on two complementary dimensions: temporally, errors in long-lived memory can persist across many decoding steps; spatially, errors in different key rows affect model outputs differently, while state magnitudes vary substantially along both rows and columns. Motivated by these observations, we propose STEPQuant, a spatial-temporal post-training quantization framework for Delta-rule recurrent states. STEPQuant allocates precision according to error magnitude and memory lifetime, and jointly fits key-row and value-column scales based on state distributions and key-row impact on output error. Experiments on Qwen3.8-27B and Kimi-Linear-48B-A3B-Instruct across both long- and short-generation benchmarks show that STEPQuant closely matches FP32-state accuracy under a nominal 6-bit budget and outperforms uniform INT8 in its 4-bit configuration. Integrated into SGLang with optimized GPU kernels, 6-bit STEPQuant achieves over 5x recurrent-state compression and reduces total serving memory by up to 68.7%. Our code is available at https://github.com/Dreamer-Toby/STEPQuant. --- ### LeapQuant: Efficient Linear Attention with Accurate Recurrent State Quantization Authors: Yi Pan, Haocheng Xi, Kan Zhu, Xingyang Li, Yibo Wu, Mayank Mishra, Hongtao Zhang, William X. Zheng, Baris Kasikci, Song Han, Kurt Keutzer, Rishabh Iyer et al. Published: 2026-09-29 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2609.38166v1 Recent LLMs increasingly adopt hybrid designs that replace standard attention with linear attention, such as Gated DeltaNet (GDN) and Kimi Delta Attention (KDA). Although they compress the context into a fixed-size recurrent state and substantially reduce the cost of long-context processing, repeatedly reading and updating that state remains a major inference bottleneck. Quantization offers a natural way to reduce this cost, but can significantly degrade model quality, due to the accumulation of rounding errors and the presence of outlier rows and columns in the state. To address these challenges, we propose LeapQuant, a training-free method that achieves near-lossless performance under 8-bit recurrent-state quantization. First, to mitigate error accumulation, we propose per-window quantization, which leaps over a window of tokens and quantizes the state only once at its end. Within a window, outputs are computed from the fixed low-bit state together with high-precision buffered updates. Second, to reduce the error introduced by each quantization, LeapQuant retains the state's largest outliers as a few high-precision Compensator Tokens, which share the update path of real tokens. We then smooth the remaining residual before quantization to further reduce the error. Comprehensive experiments across the Qwen, Kimi, and GLM model families show that LeapQuant substantially reduces memory and compute costs during inference. With accuracy comparable to the FP32 baseline, it achieves average speedups of 2.05--3.70$\times$ at the kernel level and 1.47$\times$ for end-to-end inference on NVIDIA B200, RTX PRO 6000, and RTX 5090 GPUs. --- ### Cropland PAtteRNS: Parallel Dimensional Attention Networks and Attention to Dataset Disparity for Crop Segmentation in Satellite Imagery Time Series Data Authors: Joseph Metcalfe, Sara Sharifzadeh, Fabio Caraffini Published: 2026-09-29 Categories: cs.CV, cs.LG arXiv: https://arxiv.org/abs/2609.38165v1 The landscape of satellite imagery time series datasets and boundary-pushing architectures for cropland segmentation has never been richer. However, in this gold rush, important truths are being missed on both fronts, as a drive for the most novel concepts or the largest datasets pushes finer details to the side. In this paper, we present our hybrid transformer-convolutional model, Cropland Parallel Attention and Refinement Network for Segmentation (PAtteRNS), the first model to use self-attention mechanisms separately for each of the temporal, spectral, and spatial aspects of Sentinel-2 multispectral SITS data. To achieve fully-factorised attention in our proposed model, we introduce a novel parallel transformer architecture which significantly reduces the computational complexity of triple-factorised self-attention. We validate our architecture with an in-depth ablation study, and analyse the performance of our model against state-of-the-art crop segmentation models on multiple tile-size variants of the popular PASTIS and MTLCC datasets. Our findings show our model to outperform all others in the task of crop class segmentation, verified across multiple important segmentation metrics, with especially strong performance against compared models seen in the often under-reported parcel delineation quality, for which we use the Boundary IoU metric. We also find that flawed class groupings within datasets can have a significant negative impact on model performance, and report that alternate tile-size variants of crop segmentation datasets produce results incomparable to one-another, invalidating fair comparison between model performance when trained on different tile-sizes. Based on these findings, we suggest further work is required to standardise best practices when constructing SITS crop segmentation datasets, and to enable future dynamic-tile-sizing for ideal model performance. --- ### Rethinking Representations for World-Action Modeling Authors: Haoyi Jiang, Liu Liu, Xinjiang Wang, Zhihao Sun, Zequn Chen, Sen Wang, Xinjie Wang, Xia Chen, Jingfeng Yao, Weiheng Zhao, Shanglin Yuan, Zhizhong Su et al. Published: 2026-09-29 Categories: cs.CV, cs.RO arXiv: https://arxiv.org/abs/2609.38163v1 World-action models jointly learn robot policies and predict future observations, making the representation space an interface between control and prediction. We study the design of this space through controlled comparisons, finding that neither reconstruction fidelity nor pre-trained perceptual features alone ensure effective policy learning. These findings motivate ReWAM, a representation-centric world-action model built on pre-trained DINO features. Feature Calibration and a Temporal Representation Bottleneck organize these features into compact world states suited to dynamics modeling. Action-Grounded Representation Shaping routes only action-loss gradients to the bottleneck, thereby letting the policy shape what the representation encodes while the world model learns how it evolves. Without generative video pre-training, ReWAM achieves 93.6% success on RoboTwin 2.0. On RoboDojo, it achieves an average score of 12.29 and a success rate of 8.28% using approximately 600 hours of embodied pre-training data. --- ### A Spectral Theory of Distortion in LLM Graph Reconstruction: Sharp Bounds and Empirical Characterization Authors: Jianru Shen Published: 2026-09-29 Categories: cs.LG, cs.DM arXiv: https://arxiv.org/abs/2609.38161v1 Evaluations of graph reconstruction by language models typically report a single aggregate distance between the original and the reconstructed graph. We prove that for the Wasserstein distance between Laplacian spectra such a summary is bracketed by two edge counts, the net change in edge number from below and the symmetric difference from above, each scaled by $2/n$ where $n$ is the number of vertices. The bracket is sharp: its two ends coincide exactly when the reconstruction only adds edges or only deletes them, and on that class the distance is a rescaled edge count that says nothing about which edges changed. When the ends differ, the residual between the distance and the lower end is positive only if the reconstruction both invented and lost edges, which turns it into a certificate of mixed editing computable from the reported summaries alone. We characterize these regimes in 135 reconstructions produced by three open-weight models over 45 synthetic graphs. Seventy-seven outputs are one-sided and 29 mixed outputs have $X > 0$, including cases where edge count is exactly preserved while nineteen edges were simultaneously invented and lost. The three models differ in editing policy, ranging from copying the input to attempting completion at the cost of large hallucination volume, a distinction that aggregate distortion does not reveal. --- ### EmoRES-TTS: Residual-Enhanced Vector Steering for Emotional Speech Generation Authors: Kuan-Po Huang, Haohe Liu, Puyuan Peng, Haibin Wu, Zhaoheng Ni, Hung-yi Lee, Jinwon Lee, Neha Chachra Published: 2026-09-29 Categories: cs.SD, cs.CL, eess.AS arXiv: https://arxiv.org/abs/2609.38157v1 Emotion-conditioned text-to-speech (TTS) models may fail to express the requested emotion reliably, and improving controllability by additional training is costly in both computation and emotion-labeled speech training data. We therefore study vector steering, a training-free approach that modifies the internal representations of a frozen model. CoCoEmo, a conventional vector steering method for emotion TTS, treats each emotion vector as an indivisible direction controlled by a single global strength, limiting adherence to the requested emotion. In this work, we first discover that an emotion vector can be decomposed into a shared component that moves speech away from neutral expression and a residual component that directs generation toward the requested emotion. Building on this finding, we propose Emotion Residual-Enhanced Steering for TTS (EmoRES), a novel method that controls the two components without retraining the backbone. On IEMOCAP, EmoRES outperforms CoCoEmo across all four objective emotion metrics on the IndexTTS-2 and CosyVoice2 backbones. Rank correlation improves by 26.13 and 12.97 percentage points, corresponding to relative gains of 118.8% and 33.1%, while emotion hit rate improves by 12.95 and 6.92 points, corresponding to relative gains of 20.1% and 9.8%. Human evaluation further shows a relative improvement up to 35.0% in the rate at which listeners correctly identified the dominant requested emotion and up to a 17.3% improvement in fidelity, while listeners prefer EmoRES for naturalness in up to 63.8% of pairwise comparisons. Component ablations further demonstrate that effective control benefits from preserving the shared component while strengthening the residual of the emotion steering vectors. --- ### DMA$^2$: Pixel-space Distribution Matching with Adversarial and Anchor Losses Authors: Xin Lin, Zhifei Zhang, Yuqian Zhou, Haitian Zheng, Shaoteng Liu, Lehan Yang, Zhe Lin, Ming-Hsuan Yang, Truong Nguyen Published: 2026-09-29 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.38156v1 Distribution matching distillation (DMD) provides a general framework for few-step diffusion generation, but its modern text-to-image instantiations have been developed primarily around latent diffusion. It therefore overlooks key properties and design opportunities of native RGB. We revisit two DMD interfaces for pixel-space teachers. On the teacher-matching side, diagnostics show low-noise RGB matching is dominated by a local-texture cue, motivating a fixed high-noise matching band. On the real-data side, native clean-RGB outputs allow guidance from an external visual representation without traversing a decoder or sharing the heavy fake-score critic. DINO-Adv removes this critic from the adversarial gradient path and supplies local parametric patch guidance. For distribution-level guidance, we introduce AF-Loss, a parameter-free auxiliary semantic distribution-field objective designed for text-to-image DMD. It operates on detached rolling real and generated supports in the shared DINOv2 space while preserving prompt-conditioned teacher supervision. AF-Loss adds no learnable parameters or inference-time computation. Together these designs form DMA$^2$. Across DPG-Bench, GenEval, VQAScore, and COCO30K, the four-step DMA$^2$ student performs better than the 25-step teacher and evaluated few-step distillers. --- ### Beyond the Timeline: Augmenting Long-Video Memory with Grounded Entity Biographies Authors: Hui Ren, Lei Fan, Henry Pao, Han Guo, Zeeshan Zia, Ying Chen, Alexander Schwing, Gang Hua Published: 2026-09-29 Categories: cs.CV, cs.AI, cs.CL, cs.IR, cs.LG arXiv: https://arxiv.org/abs/2609.38155v1 Answering questions about long videos often requires connecting events involving the same objects across hours or days. Chronological descriptions and text-derived entities can leave physical identity unresolved: different objects may share a description, while observations of the same object remain disconnected across events. Retrieving relevant events therefore does not necessarily recover the "biography" of the particular entity a question concerns. To address this, we introduce Grounded Entity Biographies (GEB), a long-video memory framework that groups visually grounded observations of the same physical instance across clips into retrievable biographies while preserving the context of each moment. During question answering, the biography is retrieved alongside episodic evidence, allowing the model to follow an entity through events using identity links established during memory construction. Evaluations across four benchmarks, including day-long and week-long recordings, demonstrate improvements over prior memory frameworks in both multiple-choice and open-ended question answering. On EgoLifeQA, GEB achieves 72.0% accuracy, 4.4 percentage points above the best published result. Ablations show that grounded identity association and biography reading both contribute to the gains, which additional descriptions alone do not fully recover. --- ### LongLive-Plug: Once-for-All Distillation for Video Generation Authors: Shuai Yang, Luozhou Wang, Wei Huang, ZhiFei Chen, Bohan Zhang, Xiao Fu, Qianli Ma, Chen-Hsuan Lin, Weian Mao, Bryan Chu, Song Han, Yukang Chen Published: 2026-09-29 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.38154v1 Video diffusion models are increasingly developed into specialized models for diverse downstream tasks, and this development often includes a distillation stage, for example to accelerate sampling or to improve long-video generation. This stage is typically repeated for every specialized model. We introduce LongLive-Plug, a once-for-all distillation framework that learns reusable capabilities as LoRAs on a base model for training-free, plug-and-play deployment to compatible downstream models. These capabilities include single-pass classifier-free guidance, few-step sampling, and long-context error correction for autoregressive generation. The adapters remain reusable even when downstream models add conditioning branches, expand output channels. Despite training at a fixed guidance scale, our dedicated CFG LoRA provides text guidance control through its inference weight. Combining it with a few-step LoRA simultaneously preserves few-step generation and CFG controllability on downstream tasks. We verify training-free deployment on 54 downstream models across three backbone families and eight task categories, including world modeling, robotics, editing, and multimodal generation. The approach may support additional compatible models. Each capability can thus be distilled once per backbone family and reused without per-target retraining. --- ### PowerSim: Differentiable Physics Simulation and Rendering with Power Diagrams Authors: Trong-Tung Nguyen, Anand Bhattad Published: 2026-09-29 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.38153v1 We introduce PowerSim, a method to bring physically grounded, differentiable dynamics to PowerFoam's power diagram based 3D representation. PowerSim directly couples a pre-trained PowerFoam scene to the Material Point Method (MPM) by exploiting a natural alignment between the two: the geometric and appearance properties of each primitive correspond closely to the quantities MPM already tracks as an object deforms. Consequently, simulated motion can drive the scene's geometry and appearance directly, without an auxiliary representation in between. Built on this framework, we enable a range of applications on real and synthetic scenes: (1) simulating a static scene under user interaction, (2) recovering spatially varying material fields, (3) compositing primitives from independently captured scenes into a single simulation-ready scene and (4) ray-tracing reflections that update consistently as the object deforms. Our results suggest that PowerSim excels over previous frameworks for physically grounded dynamics, while unlocking unique advantages-such as secondary ray lighting effects on dynamic scenes. Results are best viewed on our project website: https://power-sim.github.io/. --- ### FracGen: Learning How Objects Stretch and Tear with Physics-Informed Video Generation Authors: Trong-Tung Nguyen, Jiahan Zhang, Anand Bhattad Published: 2026-09-29 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.38152v1 We introduce FracGen, a fracture-aware video generation model that produces plausible, controllable fracture dynamics from a single image of an intact object, conditioned on physics signals. To train FracGen, we build FracSim, a fracture-aware simulation framework that augments material point method (MPM) simulation with a continuum damage model, producing paired fracture videos and dense, pixel-aligned physical fields at no additional cost beyond standard rendering. FracGen leverages these maps in two ways: it is trained to jointly predict them alongside RGB video, encouraging the model to capture physical state rather than surface appearance; and it is supervised with physics-informed losses that encourage consistency among the predicted maps. As a result, FracGen captures distinct material-specific fracture behavior without expensive test-time simulation or per-scene tuning, while offering fine-grained control over where an object tears, how fast the crack propagates, and how much deformation precedes failure. We further introduce a benchmark for evaluating the physical plausibility of generated fracture video, and show through extensive experiments that FracGen outperforms existing video generation baselines in both physical and visual fidelity. Results are best viewed in our project website: https://fracgen.github.io/. --- ### Pretraining Latent Information Feedback Transformers with Teacher Supervision Authors: Dor Tirosh, Ido Amos, Mor Geva Published: 2026-09-29 Categories: cs.CL arXiv: https://arxiv.org/abs/2609.38149v1 Transformer language models (LMs) are feed-forward: deep-layer representations are never fed back to shallower layers, and the only pathway for information to flow downward across generation steps is the decoded token. This narrow channel forces models to recompute intermediate results and to discard alternative continuations. In this work, we remove this bottleneck during pretraining, introducing the LIFT (Latent Information Feedback Transformer) architecture and training method which enable LMs to propagate state across generation. We achieve this by turning recurrent-state learning into a teacher-forced prediction problem: each input token is paired with an information-dense state, derived from the next-token distribution of an off-the-shelf pretrained LM. The model, extended with a small number of additional parameters, is then trained to predict both the next token and the next state. As the input states are precomputed, pretraining remains fully parallel across positions. At inference, the model's own predicted states are fed back, with a minor computational overhead that decreases with model size. Experiments with pretrained models ranging from 135M to 1B parameters show that LIFT consistently outperforms standard Transformers and baselines on language modeling, downstream reasoning tasks, and procedural tasks under token-matched budget, while being on par with or ahead of compute-matched Transformers. Moreover, a controlled study on a state-tracking task shows that a tiny LIFT outperforms same-size Transformers trained on 8x more data, even when trained with the states of a Transformer that fails the task. Overall, we show that LMs can learn to exploit deep-to-shallow feedback during pretraining via scalable teacher supervision. --- ### Thinking Before Thinking: Scaling Agentic Inference Through Meta-Reasoning Authors: Paras Dahal, Anton Bakhtin, Taco Cohen, Zhengxing Chen, Carole-Jean Wu, Rob Fergus, Scott Yih, Gabriel Synnaeve, Ruslan Salakhutdinov, Sanjeev Arora, Jason Weston, Anirudh Goyal Published: 2026-09-29 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.38147v1 As agents take on longer and more complex problems, controlling the execution becomes a task in its own right. Each step in the run brings new control choices, like which partial work to build on, whether to start fresh, or when to stop. We introduce agentic meta-reasoning, an inference-time harness that makes these choices an explicit and structured reasoning process. Workers carry out the task-level computation, while a controller consolidates what the run has established, explores next options, assesses what each option is worth under the remaining budget, and dispatches the chosen work with context drawn from persistent memory. Between decisions the controller carries only a compact account of the run rather than replaying its full history. Our baselines span production coding agents and research harnesses, together with a Direct Control Agent using the same workers and compute budget allowance. On ProgramBench, which tests long-horizon agentic capability through program reconstruction, meta-reasoning achieves 71.5% with GPT-5.5 against 58.0% for Codex; with Opus 4.8 it achieves 67.2% against 65.5% for Claude Code. On the other benchmarks, spanning abstract reasoning, multi-domain long-horizon reasoning, and proof generation, it gains between 3.6 and 4.2 points over direct control, averaged across three frontier models. It keeps improving over the tested budget ranges where direct control plateaus, though its overhead can hurt at small budgets. Artifact-graph analysis reveals more reuse of earlier work, higher coverage of correct solutions in most settings, and nonuniform gains in final selection. These results indicate that spending computation on structured control becomes more important as agents scale to longer runs. --- ### LIFT: Layout-In-Future Video Generation under Large Viewpoint Change via On-Policy Self-Distillation Authors: Shengxiang Ji, Boyang Wang, Haiyang Xu, Bingnan Li, Yucheng Mao, Zeyuan Chen, Xiaojun Shan, Xiang Zhang, Gang Hua, Jianwen Xie, Zezhou Cheng, Zhuowen Tu Published: 2026-09-29 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.38146v1 We introduce LIFT, a unified image-to-video generation framework that complements camera control with Layout-In-FuTure control, enabling users to specify what should appear in a future view and where it should appear. This addresses a practical need in controllable video generation: given an initial image, users often care not only about how the camera moves, but also about what the scene should look like at key future moments, especially the final frame. Existing camera controls specify viewpoint trajectories, while text prompts provide only coarse semantic guidance; neither precisely determines the content and spatial layout of future views. This limitation becomes particularly pronounced under large viewpoint changes, where the camera reveals regions that are not visible in the first frame. LIFT therefore uses the last-frame layout as an explicit control signal for the desired future scene. Since learning from such sparse layout guidance is substantially more challenging than conditioning on dense per-frame layouts, we introduce on-policy self-distillation (OPSD) to transfer the control capability of a dense-layout teacher to a last-frame-layout student. We further curate LIFT-Vista, a dataset featuring large viewpoint changes with camera and temporally consistent layout annotations. Experiments show that LIFT improves video quality, future-layout controllability, and camera controllability over other methods. --- ### Learning Meta-Skills for Agent Harness Design in Test-Time AI4AI Authors: Cheng Qian, Kunlun Zhu, Beibin Li, Zhenhailong Wang, Heng Ji Published: 2026-09-29 Categories: cs.AI, cs.CL, cs.LG arXiv: https://arxiv.org/abs/2609.38143v1 Agent performance depends on both reasoning ability and the environment in which it acts. We study test-time AI-for-AI, asking how a Builder can learn to construct better execution environments for a Target while both models' weights remain fixed. To make the Builder's experience reusable, we introduce Meta-Skill: principles specifying when support is needed and what resources to provide. The Builder learns these principles from Target's execution feedback on the development set, then uses the frozen skill bank to construct harnesses for unseen tasks. Across Harness-Bench and NewtonBench, full-bank meta-skills improve macro-average performance by 8.95 percentage points over no-skill construction, and 12.02 points over direct delivery of the same bank to the Target. These results highlight the value of translating experience into executable support. Gains when the same model serves both roles further suggest a path to system level self-improvement through learning to build better environments. ---