NeurIPS 2026 Papers

Online video understanding requires models to robustly encode, store, and retrieve information from a continuous stream to support accurate question answering. Existing approaches rely on key--value caching to accumulate frame-level representations over time, but allocate only a limited token budget per frame, causing loss of fine-grained visual detail. We find that naively scaling this budget degrades retrieval quality and question-answering performance, as denser token streams introduce redundancy that disrupts query--frame similarity. To address this, we first introduce an adaptive selection strategy that reduces token redundancy while preserving local spatiotemporal information. Second, we propose a training-free retrieval ensemble that leverages external models to better identify relevant frames. Our method, \textbf{MemStream}, achieves +8.0\% on CG-Bench, +8.5\% on LVBench, and +2.4\% on VideoMME (Long) over ReKV with Qwen2.5-VL-7B. Furthermore, we show that MemStream consistently improves performance across several Video-LLMs for long-video understanding.

Continuous glucose monitoring (CGM) provides a dense window into metabolic physiology, yet existing generic time-series and CGM-specific foundation models typically learn from entangled glucose sequences without explicitly capturing the temporal structure of glycemic dynamics. We present GlucoFM, a lightweight CGM foundation model that aligns irregular recordings to a 24-hour chronological grid, preserves observation masks, and decomposes glucose dynamics into slow physiological state and transient event streams, capturing low-frequency glycemic baselines and short-term deviations that may reflect acute physiological responses or sensor artifacts. GlucoFM is pretrained on 109,066 hours of unlabeled CGM recordings from 477 subjects with two complementary objectives: masked contextual latent prediction over fused daily representations and temporal dynamics prediction over state and event streams. Across four diverse cohorts and seven clinical prediction tasks, GlucoFM achieves the strongest frozen linear-probing performance among evaluated baselines, improving average PR-AUC by 4.1 points over the best CGM-specific foundation model. Its gains are most pronounced on core metabolic outcomes, leading PR-AUC on all diabetes-risk and $\beta$-cell dysfunction tasks and on 3 of 4 insulin-resistance tasks. GlucoFM also achieves the best overall cross-dataset transfer performance among evaluated methods and strong few-shot adaptation, highlighting physiology-aware decomposition as an effective inductive bias for transferable CGM representation learning.

Generating accurate radiology reports from 3D medical volumes remains prone to hallucination. Existing grounding strategies often depend on external visual foundation models or on fixed medical phrase queues. The former passes segmenter-derived evidence to generation, where segmentation errors are hard to suppress. The latter has limited coverage of long-tail and compositional abnormalities. We therefore propose HeatRAD, a glance-before-tell framework that first routes suspicious slice evidence into an organ-ordered visual prefix, then generates reports from this localized evidence without explicit grounding supervision. HeatRAD uses a slice-centric signed heat-conduction encoder that performs efficient spectral mixing while separating smooth anatomy from lesion-like residues. \method{} then scores suspicious slices, binds their evidence to coarse organ contexts, and assembles an organ-ordered visual prefix, guiding the decoder toward localized abnormalities instead of redundant volumetric context. Experiments on three 3D radiology benchmarks show consistent gains in clinical fidelity and hallucination-related metrics, supporting anomaly-first evidence selection as an annotation-efficient paradigm for 3D report generation.

Layer pruning removes entire Transformer decoder blocks from large language models, but introduces a mismatch between the hidden state received by the next surviving layer and the distribution it was trained to process, leading to significant performance degradation. We propose Ghosted Layers, a training-free recovery module that addresses this issue by solving a boundary activation alignment problem. Our method derives a closed-form optimal linear operator from a small calibration set to reconstruct the activation discrepancy introduced by the pruned layers. We show that this solution corresponds to the unconstrained optimum of the alignment objective, whereas existing methods are restricted to constrained solutions over limited operator subspaces. Experiments across multiple LLM backbones and pruning strategies demonstrate that our method consistently improves accuracy and perplexity over prior training-free baselines, while preserving the efficiency gains of layer pruning.

When language models lack relevant knowledge for a given query, they frequently generate plausible responses that can be hallucinations, rather than admitting being agnostic about the answer. Retraining models to reward admitting ignorance can lead to overly conservative behaviors and poor generalization due to scarce evaluation benchmarks. We propose a post-hoc framework, Conformal Abstention (CA), adapted from conformal prediction (CP) to determine whether to abstain from answering a query. CA provides finite-sample guarantees on both the probability of participation (i.e., not abstaining) and the probability that the generated response is correct. Importantly, the abstention decision relies on prediction confidence rather than the non-conformity scores used in CP, which are intractable for open-ended generation. To better align prediction confidence with the model's ignorance, we introduce a calibration strategy using representation geometry within the model to measure knowledge involvement in shaping the response. Experiments demonstrate that we improve selective answering significantly with 75% conditional correctness.

Zeroth-order optimization (ZOO) is a promising approach for memory-efficient fine-tuning of Large Language Models (LLMs). However, scaling ZOO to larger models on a single GPU necessitates aggressive quantization, which often destabilizes training. A primary cause of this instability is that existing methods treat strictly positive quantization scale parameters as Euclidean variables, potentially leading to boundary violations and optimization difficulties. To address this issue, we propose Hyper-Octant Zeroth-order Optimization (HoZO), a geometry-aware framework that formulates fine-tuning as Riemannian optimization on the positive orthant manifold. By constructing a novel geodesically complete metric and deriving the closed-form exponential map, HoZO inherently enforces positivity constraints without bias-inducing projections. On the theoretical side, we prove that HoZO achieves the optimal oracle complexity for zeroth-order methods. On the empirical side, HoZO outperforms baselines across six downstream tasks and three model sizes. Notably, it achieves a $34\times$ memory reduction compared to first-order fine-tuning, enabling the stable fine-tuning of a 4-bit quantized Llama-2-70b model on a single 48GB GPU.

Multimodal Large Language Models (MLLMs) have shown increasing potential for 3D scene understanding and spatial reasoning from videos. However, even with strong visual geometry priors, the hidden semantics of existing 3D-enhanced MLLMs can drift across views, leading to degraded grounding, captioning, and spatial reasoning performance. The underlying reason can be their geometry injection as additional tokens or additive features, which enriches visual content but fails to define how view-dependent semantic states should transform. To address this issue, we propose Lie Routing Vision-Language Model (LieVLM), a geometry-conditioned Lie routing framework that treats geometry as an operator over hidden semantics. LieVLM first elicits local Lie states from 3D geometry tokens by predicting latent rotation states and routing coefficients. It then converts these rotation states into triplet-wise Lie group actions to route projected 2D visual semantics. Finally, a geometry-aware routed fusion module combines the original semantic path, the direct 3D context path, and the Lie-routed transformation path for language generation. This design preserves pretrained visual-language semantics while allowing geometry to actively correct view-induced semantic drift. Extensive experiments on 3D scene understanding and spatial reasoning benchmarks demonstrate that LieVLM improves view robustness, especially in severe-pose-change regimes, and consistently outperforms additive geometry fusion methods. Our results suggest a shift in 3D MLLM design, i.e., geometry should not merely be injected as context, but should operate on semantic representations.

Large language models solve complex problems by generating lengthy chains of explicit reasoning tokens. While effective, this makes reasoning expensive, length-sensitive, and constrained to (discrete) natural language. While latent reasoning offers a continuous alternative, determining useful structures for intermediate latent states is an open challenge. In this paper, we formulate latent reasoning as a geometric path-approximation problem within the model’s pretrained token-embedding space. We introduce Geometric Latent Reasoning (GLR), which uses a lightweight transition head to predict iterative direction updates in embedding space. Using textual chain-of-thought traces as anchors, GLR learns to approximate discrete reasoning trajectories while permitting continuous deviations from exact token embeddings. Evaluations on mathematical reasoning benchmarks using Qwen3 models reveal an emergent phenomenon: geometric latent reasoning induces substantially shorter generations without an explicit length objective. By replacing early explicit reasoning with continuous latent steps, models often reach correct answers using substantially fewer total generation steps. These findings suggest that continuous trajectories act as compact intermediate reasoning states, exposing a new tradeoff between latent computation budget, output length, and accuracy.

Large Multimodal Models (LMMs) often struggle with geometric reasoning due to visual hallucinations and a lack of mathematically precise Chain-of-Thought (CoT) data. To address this, we propose the GeoSym Engine, an automated and scalable neuro-symbolic framework. By leveraging a type-conditional grammar and an analytic SymGT Solver, it derives exact symbolic ground truths and seamlessly integrates with a robust rendering pipeline to produce high-precision geometric diagrams. Using this engine, we construct GeoSym127K, a difficulty-stratified dataset featuring 51K high-resolution images, 127K questions with symbolic ground truths, and 55K answer-verified CoT QA pairs. We also introduce GeoSym-Bench, an expert-curated suite of 511 complex samples for rigorous evaluation. Through extensive supervised fine-tuning (SFT), we demonstrate that GeoSym drives concentrated improvements specifically on diagram-dependent and multi-step geometry tasks. Our Qwen3-VL-8B model gains an absolute +22.21% on the MathVerse Vision-Only subset and reaches 61.52% (+6.19% improvement) on WeMath, mitigating long-horizon logic fragmentation and outperforming advanced closed-source models like Doubao-1.8. Furthermore, applying Reinforcement Learning with Verifiable Rewards (RLVR) via GRPO reveals that initializing from structural SFT checkpoints substantially elevates the performance ceiling over zero-shot RL. Driven by deterministic exact-match signals, this showcases the robust scaling potential of our verifiable reasoning synthesis. Datasets and code are available at https://huggingface.co/datasets/Tomie0506/GeoSym127K and https://github.com/Tomie56/GeoSym127K

Worldwide image geolocalization aims to predict precise GPS coordinates for images captured anywhere on Earth, which is challenging due to the large visual and geographic diversity. Recent methods mainly follow two paradigms: retrieval-based approaches that match queries against a reference database, and generation-based approaches that directly predict coordinates using Large Vision-Language Models (LVLMs). However, we observe distinct error profiles between them: retrieval excels at fine-grained instance matching, while generation offers robust semantic reasoning. This complementary heterogeneity suggests that no single paradigm is universally superior. To harness this potential, we propose GeoRouter, a dynamic routing framework that adaptively assigns each query to the optimal paradigm. GeoRouter leverages an LVLM backbone to analyze visual content and provide routing decisions. To optimize GeoRouter, we introduce a distance-aware preference objective that converts the distance gap between paradigms into a continuous supervision signal, explicitly reflecting relative performance differences. Furthermore, we construct GeoRouting, the first large-scale dataset tailored for training routing policies with independent paradigm predictions. Extensive experiments on IM2GPS3K and YFCC4K demonstrate that GeoRouter significantly outperforms state-of-the-art baselines. Our dataset, code, and checkpoints are publicly available to facilitate future research.

State space models (SSMs) provide an efficient route to global image restoration because they model long sequences with linear complexity. Recent restoration architectures further enlarge the effective context by reordering image tokens so that semantically similar pixels become neighbors in the 1D scan. This strategy improves texture aggregation, but it also exposes a mismatch between the geometry of images and the discretization used by SSMs: adjacent tokens in the reordered sequence can be spatially distant, while the recurrent update still treats them as consecutive samples of a continuous process. In addition, existing prompt-based modulation relies on a finite set of discrete prompts, which is poorly matched to the continuous variation of natural textures and offers no explicit compensation for the low-pass tendency of recurrent integration. We propose GeoMamba, a geometry-aware and continuously modulated SSM for image restoration. GeoMamba introduces geometry-adaptive discretization (GAD), which conditions the SSM time-step on the Euclidean distance between consecutive reordered tokens and attenuates history propagation across spatial jumps. It also introduces a continuous dynamic matrix (CDM), which regresses pixel-wise modulation parameters from feature and gradient cues to reduce prompt quantization and preserve high-frequency details. Experiments on lightweight and classical image super-resolution demonstrate consistent improvements over strong Transformer- and Mamba-based baselines, while achieving better or competitive performance in additional denoising and JPEG artifact reduction settings.

Several educational domains rely heavily on diagrams and visual cues, yet most existing tutoring datasets are limited to text-only interactions. This limits the development of AI tutors that can teach in visually grounded ways used by human instructors. Thus, we introduce GeoDial, a multimodal tutoring dataset of over 1.3K teacher-student dialogs in the domain of geometry collected from experienced math teachers, where instructional turns are explicitly grounded in diagram highlights. We propose a scalable annotation protocol that integrates dialog acts, visual highlighting, and feedback, enabling fine-grained supervision of both language and visual tutoring behavior. To illustrate the challenges posed by this setting, we fine-tune several vision–language models on GeoDial and evaluate their ability to generate tutoring utterances and diagram highlights. While supervised fine-tuning substantially improves the quality of generated dialog, it struggles to produce accurate diagram highlights, revealing a key limitation of current methods and highlighting the need for approaches that more effectively integrate visual reasoning with pedagogical interaction.

Vision-Language-Action (VLA) models are emerging as highly effective planning models for end-to-end autonomous driving systems. However, current works mostly rely on imitation learning from sparse trajectory annotations and under-utilize their potential as generative models. We propose Generative Scenario Rollouts (GeRo), a plug-and-play framework for VLA models that jointly performs planning and generation of language-grounded future traffic scenes through an autoregressive rollout strategy. First, a VLA model is trained to encode ego vehicle and agent dynamics into latent tokens under supervision from planning, motion, and language tasks, facilitating text-aligned generation. Next, GeRo performs language-conditioned autoregressive generation. Given multi-view images, a scenario description, and ego-action questions, it generates future latent tokens and textual responses to guide long-horizon rollouts. A rollout-consistency loss stabilizes predictions using ground truth or pseudo-labels, mitigating drift and preserving text-action alignment. This design enables GeRo to perform temporally consistent, language-grounded rollouts that support long-horizon reasoning and multi-agent planning. On Bench2Drive, GeRo improves driving score and success rate by +15.7 and +26.2, respectively. By integrating reinforcement learning with generative rollouts, GeRo achieves state-of-the-art closed-loop and open-loop performance, demonstrating strong zero-shot robustness. These results highlight the promise of generative, language-conditioned reasoning as a foundation for safer and more interpretable end-to-end autonomous driving.

Diffusion and flow-based generators are increasingly used in robot policies to produce action chunks from visual observations, proprioception, and language instructions. However, closed-loop robot control requires solving action-generation problems of varying difficulty, whereas standard diffusion and flow-matching inference typically follows a fixed integration schedule. We introduce **Generative Control as Optimization (GeCO)**, a time-unconditional framework that turns action synthesis from fixed-time integration into iterative optimization. GeCO learns a stationary velocity field over action sequences, and refines actions until the field norm becomes small. This enables adaptive computation at each planning call: simple states can terminate early, while difficult states can use additional refinement. The same field norm also provides a lightweight signal for non-convergent action generation under distribution shift. GeCO can be instantiated in both diffusion-transformer policies and flow-matching Vision-Language-Action (VLA) systems without changing the surrounding policy interface. Across simulation benchmarks and real-world tasks, GeCO matches or improves task performance while enabling convergence-based inference as a plug-and-play replacement for standard time-conditioned action generators.

Vision-language models (VLMs) are increasingly used as perception and reasoning modules for embodied intelligence, yet they remain brittle when understanding fine-grained physical processes in videos. While existing models often recognize the overall task, they frequently fail to recover the intermediate actions and result states that determine how an embodied interaction actually unfolds. We argue that this limitation stems in part from a supervision bottleneck: conventional video-language data is typically constructed through post-hoc annotation, where captions are written after video collection and often provide only coarse, loosely aligned descriptions of dynamic physical events. We propose **Gen4Understand**, a generation-for-understanding framework that reverses this data flow. Instead of captioning videos after they are observed, Gen4Understand first represents embodied interactions as structured action scripts composed of atomic events and state deltas, then synthesizes videos conditioned on these scripts and real start-end frame anchors. A verifier-calibrator scores each generated candidate using its conditioning script, endpoint anchors, and video content, filtering for semantic consistency, endpoint fidelity, physical plausibility, temporal coherence, and visual quality. The retained atomic clips are further merged into action-, subtask-, and task-level supervision and converted into captioning and question-answering data for VLM adaptation. Experiments across public embodied video splits, distractor-rich hallucination tests, simulated VLA benchmarks, and real-robot tasks show that Gen4Understand consistently improves fine-grained action understanding, result-state recognition, temporal ordering, hallucination resistance, and downstream embodied control. These results suggest that controllable video generation can serve not merely as data augmentation, but as a mechanism for creating calibrated process-level supervision for embodied intelligence.

Individual-level wildlife identification often suffers from data scarcity, as varying observations of the same animal under diverse poses, viewpoints, and motions are rarely available. Image-to-video (I2V) generation offers a promising way to mitigate this limitation by synthesizing additional observations from a single reference image. However, existing I2V models mainly emphasize global layout, semantics, and motion, and therefore often fail to preserve fine-grained local appearance cues that distinguish one wildlife individual from another, such as fur texture, stripe boundaries, spot configurations, and contour transitions. We observe that these identity-critical cues are closely related to high-frequency information, which is essential for distinguishing individuals. To address this challenge, we propose WildIcon, a high-frequency-guided I2V framework for wildlife individual consistency. Specifically, WildIcon introduces a frequency-aware identity encoding branch that extracts individual-specific high-frequency cues from the reference image. Combined with isolated foreground information, the resulting identity tokens are then injected into cross-attention blocks as identity conditioning. Building on a frozen backbone with lightweight identity adaptation, WildIcon strengthens fine-grained identity preservation while retaining the motion controllability and semantic fidelity of the base I2V model. In addition, to support the training and evaluation of wildlife individual-consistent I2V, we construct WildlifeVid, a wildlife-centric video dataset with high-quality, temporally coherent clips and individual-level identity labels. Experiments on both I2V and downstream animal re-identification (ReID) show that WildIcon achieves stronger individual consistency than existing baselines and provides useful, reliable synthetic data to improve downstream ReID performance.

Influence functions (IFs) approximate the effect of removing or reweighting training data on a trained model, but their linear approximation often lacks the accuracy required for reliable post hoc model editing. We identify two culprits and propose generalized influence functions (GIFs) and pseudo-LiSSA (p-LiSSA) to address each in turn. First, classical IFs update all parameters indiscriminately, leading to unnecessary updates to data-irrelevant parameters. GIFs instead restrict edits to a \textit{data-relevant subspace} and leave the remaining parameters fixed. We theoretically show that GIFs achieve a tighter error bound than classical IFs, and that proper subspace selection requires avoiding low-curvature directions while maintaining strong alignment with the target gradient. Second, LiSSA iterations are destabilized by negative Hessian eigenvalues, which existing methods handle via $\ell_2$ damping---at the cost of a biased inverse-Hessian-vector product (iHVP) estimate. p-LiSSA eliminates this bias by solving the \textit{restricted inverse-Hessian problem} without such damping. We further show that the presence of low-curvature directions slows p-LiSSA convergence, and that the data-relevant subspace selected by GIFs mitigates this. Across five image and text unlearning benchmarks, GIFs update only 10% of the parameters yet preserve accuracy up to 2.12 percentage points better than IF-based baselines, while improving unlearning performance by up to 2.73 percentage points. Moreover, p-LiSSA improves restricted iHVP estimation by reducing the residual by 9.6$\times$ over existing iHVP baselines, while using 19.8% less memory with comparable computation time.

Modern image generation and editing systems can produce photorealistic, prompt-aligned images, but still often render familiar objects at implausible relative sizes. To measure this failure mode, we introduce GenScale, a benchmark and evaluation protocol for real-world relative object scale in image generation and editing. GenScale contains 900 image-level entries and 1,643 pairwise anchor-target scale relations across common-object generation, human-product generation with metric dimensions, and scale correction from failed generations. We further design a human-calibrated ordinal judge for scalable pairwise scale evaluation. Last but not the least, we introduce ReScale, a model-agnostic post-processing agent for localized scale correction without modifying the source generator. Experiments reveal that state-of-the-art image generators and editors cannot reliably observe relative scale yet, while ReScale consistently improves scale plausibility across generated and edited images. Together, GenScale establishes relative object scale as a distinct, measurable, and actionable capability for image generation systems.

We introduce a new approach to high-fidelity 3D scene reconstruction from multi-view RGB images that tightly couples reconstruction with a strong generative 3D prior. We cast scene reconstruction as conditional 3D generation over a set of spatially-localized, overlapping chunks that together tile the scene, scaling generation to large scene extents. Crucially, we inherit the fidelity and completeness of state-of-the-art generative shape models -- we use Trellis.2 as an example -- which we generalize to the scene level. To this end, we propose a projection-based conditioning mechanism that lifts posed multi-view image features into a coherent 3D representation aligned with the generative model, independent of view ordering and spatially anchored to the scene, yielding high-fidelity, multi-view consistent generated geometry. This enables lifting the strong object-level prior of Trellis.2 to multi-view, scene-scale generation, producing faithful, editable PBR mesh reconstructions of indoor environments. As a result, we obtain high-fidelity results that outperform cutting-edge reconstruction methods by 16\%.