NeurIPS 2026 论文列表

While spatiotemporal deepfake detectors have shown high AUC, our experiments reveal that they remain susceptible to evasion attacks. These models tend to overfit on fragile temporal spectrum cues, rather than learning robust semantic causality. To mitigate this vulnerability, we propose SpInShield, a temporal spectral-invariant defense framework explicitly designed to decouple semantic motion from manipulatable spectral artifacts. We propose a learnable spectral adversary that dynamically synthesizes severe spectral deformations, simulating extreme attack scenarios. By employing a shortcut suppression optimization strategy, SpInShield compels the encoder to extract reliable forensic cues while autonomously purging unstable spectral statistics from the latent space. Experiments show that SpInShield obtains competitive performance on widely used datasets and outperforms the strongest baseline by 21.30\% AUC under simulated amplitude spectral attacks.

Recent multimodal large language models (MLLMs) achieve strong results on video question answering, yet often fail to recover the temporal structure that makes a video coherent. They may merge neighboring events, miss single-event progression, or infer implausible relations across events. We study structured temporal understanding as the ability to recover event structure from visual evidence, including event boundaries, single-event progression, and multi-event relations. We argue that this gap arises from a mismatch between current post-training objectives and temporal reasoning, where caption-style supervision allows models to rely on language shortcuts rather than visual temporal grounding. To address this, we propose $\textbf{EventLens}$, a vision-centric reinforcement learning framework that learns video understanding through event-structure recovery}. Instead of reconstructing captions, EventLens trains models to recover temporal structure under controlled perturbations, such as altered temporal granularity, reversed progression, and shuffled event order. We instantiate this framework with a three-level temporal ontology and derive three verifiable task families: event segmentation, progression discrimination, and multi-event relation reasoning. These tasks admit deterministic rewards, enabling task-specialized GRPO training without human preference labels or learned reward models. We further introduce multi-teacher on-policy distillation (MT-OPD) to consolidate specialized policies into a unified model. Experiments on Qwen3-VL backbones show consistent improvements on temporal grounding and event-sensitive reasoning benchmarks, while preserving general video-QA performance. Learning curves further demonstrate that EventLens achieves comparable downstream transfer with significantly reduced training cost compared to direct mixed RL optimization.

Multimodal language models increasingly show promise for enabling the large-scale computational analysis of film, opening up new avenues for learning about film history and the evolution of narrative techniques. But the creation of stable benchmarks built around Hollywood films is complicated by copyright protections. In this work, we address these concerns directly, by building a new collection of Hollywood films defined by two criteria: box office popularity (where we publish the first large-scale, open collection of weekly box office earnings reported by Variety magazine from 1922-1979); and likely public domain status (by researching copyright registrations and renewals in the US Catalog of Copyright Entries). We build a new multimodal MCQ benchmark on top of this collection that focuses on narrative elements that directly evaluate the abilities of models to inform meaningful research on film narrative; we find that many vision-language models struggle on this task (with many performing at near-chance levels of accuracy), while audio-visual models (including those that use audio in captioning scenes) reach a maximum accuracy of 61.1%, well below human-level performance.

We introduce epistemic pairwise maximin share (EPMMS), a new fairness notion for fair division of indivisible goods. Two fundamental notions in this setting are envy-freeness up to any item (EFX) and pairwise maximin share (PMMS), with PMMS being stronger than EFX. While EFX has been extensively studied, far less is known about PMMS. Recent work shows that relaxing EFX via an epistemic perspective leads to substantial progress on the EFX problem, raising the question of whether a similar approach can advance our understanding of PMMS. Motivated by this, we initiate the study of EPMMS, the epistemic relaxation of PMMS. EPMMS is more challenging than EEFX: the key approaches underlying recent progress on epistemic EFX inherently fail to extend to EPMMS. We establish the following results. 1. For additive valuations, $4/5$-EPMMS allocations exist and can be efficiently computed. 2. For bivalued valuations, EPMMS allocations exist and can be efficiently computed; in fact, we obtain the stronger guarantee of epistemic groupwise maximin share (EGMMS), which also strengthens the existence of MMS allocations for this setting. 3. EPMMS allocations exist in two settings where MMS allocations need not exist: instances with three additive agents or two types of additive agents.

Real-world endoscopic diagnosis unfolds across native video, depending on continuous traversal of gastrointestinal organs where a single positive finding triggers a multi-minute reasoning chain over location, morphology, severity, and intervention. Existing endoscopy multimodal benchmarks evaluate only static images or single-disease videos that fail to capture the temporal and clinical reasoning demands of real-world endoscopy reporting. We introduce EndoSCOP-V, a video-grounded multi-turn evaluation framework for long-horizon clinical reasoning in endoscopy. The benchmark contains 1,307 real-world endoscopy and colonoscopy videos (63.2 h; 97% 1080p; mean ~174 s) — the time-scale at which endoscopists actually reason about a finding. The corpus is paired with 20,275 question–answer turns deterministically instantiated from clinician-filled structured endoscopy-reporting EHRs, avoiding synthetic or LLM-rephrased per-case annotations. Questions are derived from 1,033 clinician-reviewed templates spanning a clinically aligned 9-task × 30-sub-task reporting taxonomy across 54 World Endoscopy Organization (WEO) MST 3.0 diseases at ~18 templates per disease, substantially deeper per-disease examination than prior endoscopy benchmarks. To mirror real-world case mix, 22% of procedures carry co-existing pathologies, up to six per case. To model the hierarchical and conditional nature of clinical reporting, EndoSCOP-V introduces a Markovian context-pruned oracle-forcing protocol that evaluates multi-turn dialogues turn-by-turn. By injecting the gold answer before each subsequent turn, the protocol isolates per-turn reasoning from cascading dialogue errors and enables fine-grained analysis of clinical reasoning capabilities. Clinician-marked key-frame anchors further separate temporal-localization from frame-level comprehension failures. A panel of 15 board-certified endoscopists anchors a 78% strict-accuracy baseline on a 1,500-question audit. Across 14 open-weight LMMs, the best 27B-class model reaches 59.8% strict accuracy, falling to 53.4% once format-trivial disease-rejection items are excluded. On a between-subjects spot-check, the best of 3 frontier proprietary models trails the clinician panel by 23 pp on strict accuracy. Sub-task analysis reveals consistent gaps in quantitative measurement, multi-variable grading, and intervention reasoning on long-horizon video, providing a clinically grounded benchmark for evaluating multimodal reasoning in endoscopy.

Understanding how linguistic structure emerges in language models is central to interpreting what these systems learn from data and how much supervision they truly require. In particular, semantic role understanding (``who did what to whom’’) is a core component of meaning representation, yet it remains unclear whether it arises from pre-training alone or depends on task-specific fine-tuning. In this paper, we study whether semantic role understanding emerges during language model pre-training or requires task-specific fine-tuning. We freeze decoder-only transformers and train linear probes to extract semantic roles, using performance to infer whether role information is already encoded in pre-training or learned during adaptation. Across model scales, we find that frozen representations contain substantial semantic role information, with performance improving but not fully matching fine-tuned models. This indicates partial but incomplete emergence from pre-training alone. We show that semantic role structure emerges from language modeling objectives, but its internal implementation shifts toward more distributed representations as model scale increases.

Recent advances in 3D object generation have enabled the creation of high-fidelity 3D assets from 2D images. This *2D-to-3D* capability offers a promising path for alleviating 3D data scarcity, which remains a key bottleneck for scalable simulation in robotics and AR/VR applications. However, existing 3D generation methods primarily emphasize visual plausibility and geometric fidelity, with limited consideration of object-level interaction and structural understanding that are critical for downstream embodied AI tasks. To bridge this gap, we introduce **EmbodiedObject**, a unified framework for embodied 3D object understanding built upon latent representations produced by modern *2D-to-3D* generative models such as SAM3D. We first systematically analyze the part-level locality of these latent representations against existing unified 3D point encoders and show that they naturally preserve rich structural semantics suitable for object understanding. Motivated by this observation, we then design a unified DiT-style decoder that operates directly on the latent representation and supports multiple object-level tasks, including open-vocabulary 3D affordance prediction, 3D part segmentation, and 3D articulation estimation. Given a real-world 2D image containing an object of interest, **EmbodiedObject** extends *2D-to-3D* generation beyond visual reconstruction by jointly reasoning about affordance, articulation, and part semantics. This transforms generated 3D assets into actionable object representations suitable for downstream embodied AI applications. Our source-code will be open-sourced.

Hour-long videos, streaming feeds, and movie-length content are pushing vision-language models (VLMs) past their breaking point—the standard recipe of encoding each frame into visual tokens yields sequences that grow linearly with video length, causing prohibitive memory consumption and inference latency. Existing approaches, such as token reduction and keyframe selection, attempt to shorten the token sequence but leave the $O(n)$ scaling intact, failing to resolve the fundamental bottleneck. We propose \model, \emph{a strikingly simple, training-free frame packing framework that compresses arbitrarily long videos into a fixed-size visual representation at constant memory cost.} \model requires no fine-tuning, no architectural modification, and no auxiliary models—it plugs directly into off-the-shelf VLMs at inference time, yet consistently surpasses more elaborate baselines. Drawing inspiration from FramePack in generation literature, we propose a new content-aware compression strategy that operates under a fixed token budget for visual understanding. Specifically, \model proceeds in two steps: (1) scoring each frame by its semantic similarity to the text prompt and ranking frames in a diversity-driven manner, and (2) packing the selected frames at varying resolutions into a fixed-length context window. This design concentrates the memory budget on frames that are both instruction-relevant and visually diverse. While in theory \model can pack arbitrarily many frames into the fixed budget, we identify an empirical instantiation that generalizes robustly across different settings. Extensive experiments on LongVideoBench, MLVU, and Video-MME show that \model consistently outperforms full-resolution baselines as well as prior token reduction and keyframe selection methods, across diverse model families, parameter scales, and frame budgets. Notably, on Qwen3.5-35B-A3B with a 128-frame budget, \model yields gains of \textbf{+12.9\%, +10.6\%, and +9.7\%} over the official model on the three benchmarks, respectively.

Children acquire language with remarkable speed and robustness from limited linguistic input in ways that surpass today's best large language models. Recent research suggests this efficiency stems from learning through multimodal sensory experiences where semantic alignment between what children see and hear grounds language in the external world. However, vision-language models (VLMs) trained on curated web data fail to generalize to the sparse, weakly-aligned egocentric streams produced by wearable devices, embodied agents, and infant head-cams—and no fixed evaluation pipeline exists for measuring progress on this regime. We train VLMs on datasets with varying degrees of semantic alignment between visual and linguistic inputs, including naturalistic infant egocentric videos, and evaluate them with a comprehensive suite spanning multimodal language grounding and unimodal vision and language tasks. At the core of this suite is _Machine-DevBench_, a corpus-grounded benchmark of lexical and grammatical competence, automatically generated from the model's training vocabulary across logarithmic frequency bins to eliminate the train/eval mismatch and low statistical power of prior developmental benchmarks. Our results show that current VLM paradigms hinge on the tight semantic alignment of curated data and fail to exploit the weakly-aligned signal that dominates naturalistic egocentric input—the very regime in which humans thrive. To motivate progress, we introduce the EgoBabyVLM Challenge to drive the development of models capable of grounded language learning from the kind of naturalistic data that human infants experience.

Unified 3D foundation models aspire to generate 3D assets and reason about them in language within a single backbone, but their text–3D interaction remains largely implicit: existing methods concatenate text and 3D tokens into a flat sequence and rely on self-attention, collapsing coarse structural cues and fine geometric details into one undifferentiated representation. We argue that the central design problem is not richer geometry alone, but **scale-coupled cross-modal collaboration**. We introduce **ELSA3D**, a unified 3D model that addresses this by structuring language and geometric reasoning jointly along matched abstraction scales. The two streams are coupled through **Anchor Tokens**: a small, dynamic set of cross-modal fusion units that bind selected text tokens to geometric evidence at a learned scale and write the fused signal back, keeping interaction sparse yet precise. A lightweight per-block router makes both computation and reasoning elastic, choosing which text tokens instantiate anchors at which geometric scale so that cross-modal capacity concentrates where alignment is hardest. ELSA3D establishes new state-of-the-art on image-to-3D, text-to-3D, and 3D captioning, outperforming the strongest unified baseline on every reported metric while roughly **halving** FLOPs and inference latency relative to a non-elastic version of the same model.

Diffusion models have achieved impressive performance in video generation, but their iterative denoising process remains computationally expensive due to the large number of tokens processed at each timestep. Recently, progressive resolution sampling has emerged as a promising acceleration approach by reducing latent resolution in early stages. However, scaling this idea to video generation remains challenging, as the additional temporal dimension introduces diverse spatio-temporal demands across different videos, and compressing only a single dimension often leads to limited acceleration or degraded quality. Therefore, we propose \textbf{DVG}, a \textbf{D}ynamic \textbf{V}ideo \textbf{G}eneration framework that jointly allocates computation across time and space, automatically selecting content-aware acceleration strategies without manual tuning or retraining. \textbf{DVG} achieves near-lossless acceleration across models and tasks, reaching up to \textbf{7$\times$} speedup on HunyuanVideo and HunyuanVideo-1.5, and \textbf{18$\times$} when combined with distillation, demonstrating its potential as a key component in today’s large-scale efficient video generation systems. \emph{Our code is in supplementary material and will be released on Github.}

Video generation must account for two sources of motion, one induced by the observer's camera path and the other caused by scene dynamics. An ideal camera-controlled video model should account for both motions: let users move the camera while evolving the scene dynamics. While current models handle camera-induced motion well in static settings, they struggle for dynamic scenes: objects are static, move incorrectly, or degrade in generation quality. We introduce DynaTokens, a lightweight set of learnable scene-specific tokens that teach dynamics to an existing camera-controlled world model. Our method is motivated by a simple asymmetry between the two sources of motion: whereas camera motion affects the generated view globally, object dynamics are spatially localized. Through cross-attention, DynaTokens trains the learnable tokens from a few example trajectories for a scene while keeping the base model frozen, and enables dynamics under new query camera paths. DynaTokens achieves a better simultaneous dynamics-camera tradeoff on VBench2 and WorldScore evaluations than LoRA, block finetuning, and specialized trainable-layer baselines. Analyses of token attention, ablations, and motion temporality suggest that matching the trainable interface to the structure of the learning target is important for effective adaptation.

Vision-language models (VLMs) have made steady progress as in-vehicle driving assistants, yet existing evaluations largely remain offline, focusing on perception and short-term reasoning over pre-recorded driving scenes. A complementary capability is still underexplored: whether a model can continuously monitor a causal driving video stream and issue a calibrated alert when a user-specified condition is met. Existing driving benchmarks mainly assess reactive understanding, while recent streaming video benchmarks are developed largely outside the driving domain. To this end, we introduce DriveStreamBench, a driving and streaming video benchmark that organizes this capability spectrum into four cognitive levels constructed from a shared pool of driving events, to support level-wise capability profiling and gated evaluation. We evaluate various vision-language models along with DriveStream-SFT, a supervised reference baseline trained on the benchmark's training split, under a unified streaming protocol. DriveStreamBench tests a basic but necessary capability, and all evaluated models fall short of it; current vision-language models are not yet ready for user-conditioned alerting in driving. Resources are available at https://anonymous.4open.science/r/DriveStreamBench/ and https://dataverse.harvard.edu/previewurl.xhtml?token=01f6bc4c-d809-4e40-b025-b6d783c86571.

Understanding human kinematics in outdoor, real-world environments is critical for assistive robots and intelligent infrastructure. Many perception systems rely on cameras and LiDAR, which infer motion indirectly from spatial observations, making estimates sensitive to discontinuities and ambiguity under low lighting or at distance. In contrast, mmWave radar is robust to lighting and weather, operates at long range, and directly measures motion through Doppler signatures. However, Doppler-based kinematic understanding remains largely unexplored in the wild because existing datasets are primarily controlled and scripted. We introduce DopplerWild, a dataset and benchmark for evaluating Doppler-based kinematic understanding in the wild. DopplerWild comprises 447k radar frames across four outdoor locations, including an unlabeled subset for self-supervised representation learning and a labeled subset of 539 subjects with nearby-person interference (56\%) and occlusion (16\%). The benchmark spans coarse and fine-grained motion classification, as well as velocity estimation. Beyond aggregate metrics on controlled datasets, DopplerWild evaluation isolates real-world factors: subject variation, location shift, multi-person interference, occlusion, sensing geometry, and low-label regimes. DopplerWild reveals three findings. First, self-supervised pretraining matches supervised performance with less than half the labeled data and improves performance under interference, occlusion, and lateral motion. Second, we identify key challenges for single-view Doppler sensing, including physical limits in lateral motion, subtle asymmetric limb motion, and location distribution shifts. Third, we observe bidirectional transfer asymmetry between DopplerWild and an external controlled Doppler dataset, suggesting that real-world coverage is important for transferable Doppler representations. DopplerWild provides an evaluation framework for characterizing the generalization, limits, and transferability of Doppler-based kinematic representations in the wild.

Post-hoc explanation methods for video action recognition typically conflate spatial appearance and temporal dynamics into a single saliency or importance map. We propose a perturbation-based framework that disentangles these contributions into a spatial appearance mask ($M_a$) and a temporal motion mask ($M_m$). The two masks are jointly optimized against a frozen action recognition model through factored perturbation architecture, with nearest frame replacement to identify important temporal dynamics. A separation penalty discourages the appearance mask from spending budget on frames that the motion mask has already discarded, and a one-sided ReLU area constraint allows the realized masks to fall below the optimization budget when the input does not require its full extent. To evaluate these explanations, we introduce a separation protocol that sweeps spatial pixels and temporal frames independently, backed by proofs characterizing the inherent geometric bias of standard Insertion metrics. Across EPIC-Kitchens-55, EGTEA Gaze+, and Something-Something V2, our method attains the lowest deletion AUC against leading baselines while producing masks roughly five times sparser.

Autoregressive video diffusion models (AR-VDMs) enable real-time, long-horizon video generation but degrade rapidly once the inference horizon outruns the training horizon, due to compounding exposure bias. To stabilize long rollouts, recent methods heuristically adopt frame sink, which keeping initial frames as a persistent KV cache context, inspired by attention sinks in streaming LLMs. We revisit this design and find that the mechanism by which initial frames stabilize long-range consistency in AR-VDMs is not an LLM-style attention sink; instead, consistency is anchored through a sparse subset of attention heads, while the majority of heads predominantly capture local temporal dynamics. This observation reveals an over-conditioning pathology in existing frame-sink strategies: by universally exposing every head to the pinned initial-frame context, the model becomes prone to collapsing motion toward the low-dynamic modes of the teacher during model distillation. We propose Disen-Forcing, a distillation-time plugin for AR-VDMs that disentangles and reinforces this head specialization through disentangled context routing and motion-disentangling regularization. On long-horizon video generation benchmarks, Disen-Forcing simultaneously improves semantic consistency and motion dynamics of base AR-VDMs: a trade-off that sink-based methods fail to balance.

Video diffusion alignment has been heavily reliant on scalar rewards. These rewards are typically derived from learned reward models in human preference datasets, requiring additional training and extensive collection. Moreover, scalar rewards provide coarse, global supervision, offering limited prompt-generation mismatch credit assignment and making models prone to reward exploitation and unstable optimization. We propose \textbf{Diffusion-DRF}, a free, rich, and differentiable reward framework for video diffusion fine-tuning. Diffusion-DRF employs a frozen, off-the-shelf Vision-Language Model (VLM) as the critic, eliminating the need for reward model training. Instead of relying on a single scalar reward, it decomposes each user prompt into multi-dimensional questions with freeform dense VQA explanation queries, yielding information-rich feedback. By direct differentiable optimization over this rich feedback, Diffusion-DRF achieves stable reward-based tuning without preference datasets collection. Diffusion-DRF achieves significant gains both quantitatively and qualitatively, outperforming state-of-the-art Flow-GRPO by 4.74 points in overall performance on unseen VBench-2.0.

Long-video question answering is challenging since the answer often hinges on a few decisive moments scattered throughout the video, while memory constraints force Multimodal Large Language Models (MLLMs) to sample frames at extremely low rates, resulting in extreme sparsity that risks omitting key evidence. Though aggressively enlarging the context size is intuitive, we argue that performance remains fundamentally constrained by the bias, $\textit{i.e.}$, the misleading cues caused by spurious linguistic correlations and salient yet irrelevant visual observations. In this paper, we first introduce S-MME, a benchmark designed to systematically diagnose shortcut-biased behavior in MLLMs. To mitigate this failure, we further propose a training-free framework, Counterfactual Long-video Evidence-Aware Reasoning (CLEAR). By treating the full video-question pair as a complete view, CLEAR estimates bias effects via other views by only keeping the question or locally salient visual clips. Then, we apply hidden-state intervention to mitigate the discrepancy at inference such that the correct answer can be predicted confidently. With experimental studies across multiple benchmarks, we show that our method is generalizable and can be integrated in different MLLMs, yielding consistent gains in prediction and robustness. We hope CLEAR can offer a complementary direction in test-time scaling for long-video understanding and will publish code.

Muon has recently demonstrated strong empirical performance in large language model training, but the theoretical role of momentum in Muon remains unclear. Existing analyses of Muon either remove momentum to study spectral updates in isolation, or retain momentum without explaining why it improves empirical performance. Our work bridges this gap by showing momentum in Muon acts as a spectral filter. Under a structured signal-plus-perturbation gradient model, we prove that momentum suppresses perturbations while preserving the dominant signal, thereby enlarging the spectral gap between them. This enlarged gap stabilizes the singular subspaces of the matrix passed to Muon's orthogonalization step, making the resulting update more reliable. We further show that applying momentum before orthogonalization achieves provably stronger alignment with the signal component of the gradient than either reversing this order or simply removing momentum. Experiments across diverse tasks, including LLM training, support our theoretical analysis. More broadly, our theory offers a starting point for understanding the benefits of momentum in other matrix-based optimizers.