NeurIPS 2026 Papers

Large Vision-Language Models (LVLMs) incur substantial inference costs due to the processing of a vast number of visual tokens. Existing methods typically struggle to model progressive visual token reduction as a multi-step decision process with sequential dependencies and often rely on hand-engineered scoring rules that lack adaptive optimization for complex reasoning trajectories. To overcome these limitations, we propose TPRL, a reinforcement learning framework that learns adaptive pruning trajectories through language-guided sequential optimization tied directly to end-task performance. We formulate visual token pruning as a sequential decision process with explicit state transitions and employ a self-supervised autoencoder to compress visual tokens into a compact state representation for efficient policy learning. The pruning policy is initialized through learning from demonstrations and subsequently fine-tuned using Proximal Policy Optimization (PPO) to jointly optimize task accuracy and computational efficiency. Our experimental results demonstrate that TPRL removes up to 66.7\% of visual tokens and achieves up to a 54.2\% reduction in FLOPs during inference while maintaining a near-lossless average accuracy drop of only 0.7\%.

Direct Preference Optimization (DPO) minimizes empirical risk over observed preference comparisons, implicitly assuming that the training preference distribution matches the deployment preference distribution. In practice, this i.i.d. assumption often fails, as preference distribution shift exhibits a natural hierarchical structure across three levels: group-level shift, prompt-level shift, and conditional preference-level shift. Such multi-level distributional mismatch poses a critical challenge for safe LLM deployment, yet existing robust DPO methods typically model either overall distributional uncertainty or a single robustness level, leaving the hierarchical structure of OOD preference shifts underexplored. We propose TPO, a tri-level distributionally robust framework for OOD preference optimization, which jointly models all three levels of distributional uncertainty through a hierarchical rectangular ambiguity set. By instantiating it with Wasserstein distance, we derive a tractable and interpretable objective that combines a closed-form preference-level margin penalty, prompt-level adversarial embedding perturbation, and worst-case group reweighting, each governed by an independent radius. We establish finite-sample convergence guarantees and empirically validate the effectiveness of \method{} over existing robust DPO methods.

Probabilistic neural surrogates for PDEs enable ensemble forecasting, but often exhibit miscalibration when used for uncertainty quantification, especially in compute-constrained, few-step inference regimes. Existing generative approaches typically rely on fixed global stochasticity, which cannot adapt predictive uncertainty to the local dynamical state. We propose TORNADO, a latent probabilistic forecasting framework that formulates next-step prediction as stochastic transport between VAE encoder posteriors. Unlike existing approaches with globally prescribed stochasticity, TORNADO learns a SDE with state-dependent diffusion, enabling the predictive uncertainty to adapt to the local latent dynamics. The model is trained using analytic conditional endpoint targets derived from a Gaussian interpolant, enabling simulation-free learning. An additional energy distance regularization term improves distributional alignment. Empirically, TORNADO improves uncertainty calibration over representative generative baselines, including stochastic interpolants, diffusion models, and flow matching, while maintaining comparable predictive accuracy. Our results show that adapting stochasticity to the latent state, rather than prescribing it globally, is an effective mechanism for improving calibration in practical neural PDE forecasting settings.

Soft Actor-Critic (SAC) and its variants dominate Multi-Task Reinforcement Learning (MTRL) due to their off-policy sample efficiency, while on-policy methods such as Proximal Policy Optimization (PPO) remain underexplored. We diagnose that PPO in MTRL suffers from a previously overlooked issue: critic-side gradient ill-conditioning, which may cause tail tasks to stall while easy tasks dominate the value function's updates. To address this, we propose TOPPO (Tail-Optimized PPO), a reformulation of PPO via Critic Balancing---a set of modules that improve gradient conditioning and balance learning dynamics across tasks. Unlike prior approaches that rely on modular architectures or large models, TOPPO targets the optimization bottleneck within PPO itself. Empirically, TOPPO achieves stronger mean and tail-task performance than published SAC-family and ARS-family baselines while using substantially fewer parameters and environment steps on Meta-World+ benchmark. Notably, TOPPO matches or surpasses strong SAC baselines early in training and maintains superior performance at full budget. Ablations confirm the effectiveness of each module in TOPPO and provide insights into their interactions. Our results demonstrate that, with proper optimization, on-policy methods can rival or exceed off-policy approaches in MTRL, challenging the prevailing reliance on SAC and highlighting critic-side gradient conditioning as the central bottleneck.

Generating high-quality and controllable camera motion is essential for AI-assisted cinematography, video synthesis, and 3D scene understanding. We introduce TKCAM, a Text- and Keyframe-conditioned CAMera-motion synthesis framework based on generative masked modeling. To overcome the instability of direct 3D regression, we formulate camera dynamics as continuous 12-DoF kinematic sequences and discretize them into hierarchical motion tokens via a Residual Vector Quantizer (RVQ). A two-stage masked transformer architecture then learns to reconstruct and refine these tokens, utilizing explicit self- and cross-attention modules for robust multi-modal conditioning. A central feature of our framework is animator-style keyframe control: users can provide free-form text prompts alongside sparse key poses, and TKCAM seamlessly inpaints temporally coherent in-between trajectories. Furthermore, to advance evaluation standards, we curate RealEstate10K-Cap, a large-scale text-camera dataset, and establish a comprehensive cross-domain benchmark with a Universal CLaTr Evaluator. Extensive experiments demonstrate that TKCAM significantly surpasses recent state-of-the-art baselines on Fréchet Inception Distance (FID), text-motion matching scores, and retrieval metrics (R@K), producing cinematographically plausible motions while enabling precise spatial guidance. Our code, models, and dataset will be open-sourced upon acceptance.

We introduce and analyze Target-Induced Loss Tilting (TILT) for unsupervised domain adaptation under covariate shift. It is based on a novel objective function that decomposes the source predictor as $f+b$, fits $f+b$ on labeled source data while simultaneously penalizing the auxiliary component $b$ on unlabeled target inputs. The resulting fit $f$ is deployed as the final target predictor. At the population level, we show that this target-side penalty implicitly induces relative importance weighting at the population level, but in terms of an estimand $b^*_f$ that is *self-localized* to the current error, and remains *uniformly bounded* for any source-target pair (even those with disjoint supports). We prove a general finite-sample oracle inequality on the excess risk, and use it to give an end-to-end guarantee for training with sparse ReLU networks. Experiments on controlled regression problems and shifted CIFAR-100 distillation show that \tilt improves target-domain performance over source-only training, exact importance weighting, and relative density-ratio baselines, with a stable dependence on the regularization parameter.

Multimodal Large Language Models (MLLMs) can exhibit strong text-based reasoning yet fail to apply the same capabilities to semantically equivalent visual inputs. We study this failure mode in a controlled setting by rendering text-only reasoning problems as images, preserving semantic content while changing only the input modality. Across multiple MLLM families, models solve the same problems substantially better from text than from images. We characterize this discrepancy as a reasoning-access gap, where models may correctly perceive visual content but still fail to route that content into the latent reasoning machinery used for text-based tasks. To bridge this gap, we propose TIGER (Text-to-Image Gap-targeted Training for Enhanced Reasoning), a framework that repurposes text-only reasoning corpora into effective multimodal training data. By mining "modality counterfactuals", instances where a model succeeds on a text problem but fails on its semantically-equivalent rendered image, TIGER provides targeted supervision for reasoning-access failures without requiring manually curated multimodal datasets. We implement TIGER using image-conditioned Group Relative Policy Optimization (GRPO), along with SFT and DPO variants, and show how it consistently narrows the modality gap and yields generalized visual reasoning performance gains in multimodal reasoning benchmarks such as MathVerse and EMMA. We show that Reinforcement Learning with Verifiable Reward-based models still exhibit modality-dependent reasoning gaps, and that our training recipe can further reduce these gaps. Further analysis using reasoning-subspace activation and activation patching shows that TIGER enables visual representations that better activate reasoning-relevant latent subspaces of the language backbone. Our results suggest that advancing multimodal intelligence requires moving beyond perceptual alignment toward explicit visual access to existing reasoning machinery.

E-commerce image search often takes a cropped image as the query, while each candidate is represented by full item images and structured text. This image-to-multimodal retrieval setting presents two asymmetries: a modality disparity -- a visual query must match image--text items, and a granularity disparity -- a cropped query must be compared with full images containing background context and possible distractors. Detection-based pipelines handle the granularity disparity through explicit localization but incur extra cost and error propagation, whereas CLIP-style encoders avoid detection, but are vulnerable to backgrounds or irrelevant items. To address these limitations, we propose TIGER-FG, a text-guided implicit fine-grained grounding framework for image-to-multimodal e-commerce retrieval. TIGER-FG uses item text as semantic guidance to produce target-focused item representations without object detection for retrieval. We further introduce dual distillation objectives that preserve target-region spatial consistency and query--item similarity structure, yielding more stable and discriminative multimodal representations. In addition, we construct ECom-RF-IMMR, a realistic benchmark suite with a 10M-pair training set and two evaluation benchmarks covering standard and cluttered item layouts. TIGER-FG improves Recall@1 over the strongest baseline by 6.1 and 34.4 percentage points on the two evaluation benchmarks, respectively, with only 85.7M query-side parameters and 256-dim embeddings. Results on public e-commerce benchmarks further demonstrate its generalization to noisy and one-to-many retrieval scenarios. Code and data will be released.

Selective state space models (SSMs), such as Mamba, achieve strong per-token expressivity by making the time discretization step $\Delta$ a learned function of the input. However, in doing so, $\Delta$ ceases to represent a physical sampling interval, limiting its irregular time series modeling capability. Continuous-time SSMs, such as S5, preserve the physical meaning of $\Delta$ and handle irregular timestamps natively, but their dynamics remain linear time-invariant (LTI), limiting per-token expressivity. We propose \textbf{TIDES}, a selective SSM variant that reconciles selective and continuous architectures by moving input-dependence off the step size and onto the diagonal state matrix. As a result, $\Delta$ retains its physical meaning, tied to the state discretization, allowing the model to handle irregular timestamps natively without sacrificing the per-token expressivity that makes selective SSMs effective. We show this on a novel \emph{Fading Flash} experimental benchmark, a compact controlled diagnostic for sequence models that jointly tests input-dependence and extrapolation to out-of-distribution $\Delta$ values, and isolates the distinct failure modes of current state-of-the-art architectures that TIDES avoids by construction. On large-scale benchmarks, TIDES sets the new state-of-the-art average rank on UEA time-series classification and the Physiome-ODE regression benchmark.

Watermarking text-to-image diffusion models provides a practical mechanism for provenance tracking and responsible deployment. Post-processing methods are easy to deploy but vulnerable to image-space transformations. In-processing methods improve robustness by embedding the watermark within the denoising trajectory. Yet most existing methods inject the signal into the initial noise or a fixed intermediate latent, without controlling how it evolves during the remaining denoising steps. This can entangle watermark preservation with text-guided generation, risking weaker verification or degraded content fidelity. We propose TIDE, a trajectory-aware watermarking method for text-to-image diffusion models. TIDE injects a learnable watermark into an intermediate latent and jointly optimizes it with an auxiliary watermark condition, using preservation and detection objectives to limit deviation from the unwatermarked reference while maintaining separable watermark evidence. During sampling, TIDE estimates recent text-guidance directions and redirects watermark guidance toward the residual component less aligned with this local text subspace, reducing avoidable interference with semantic generation. Verification uses null-prompt partial inversion to recover the injection latent and match its masked coefficients to the target watermark. Experiments show that TIDE achieves reliable verification on unattacked images, improves robustness under common attacks, and better preserves semantic and visual fidelity.

Recent Continuous Thought Machine architecture decouples internal computation from external inputs via neural dynamics, but relies on multi-layer perceptrons without stability guarantees. We propose to model neural dynamics using asymmetric Excitatory-Inhibitory (E-I) networks, which can be stabilized via principles from network theory and can be expressed as energy-based systems optimized through a game-theoretic loss. Building on this perspective, we introduce Temporal Inhibitory-Excitatory Dynamic Engine (TIDE), a neuro-inspired architecture that computes internal representations through neural dynamics stabilized by incorporating the Wilson-Cowan dynamics and lateral inhibition. TIDE balances biological realism by, for instance, using Hierarchical Receptive Fields and enforcing Dale's principle to ensure a realistic $80:20$ E-I balance ratio with an end-to-end trainable architecture. The aim of this paper is to introduce a new architecture that brings neuro-inspired learning to the forefront. We present proofs of convergence, stability, and complexity bounds, along with empirical ablation studies. Overall, TIDE surpasses CTM with under $50$\% of the training time and improves $\texttt{top-1}$ accuracy by an average of $+1.65$\% on ImageNet under various perturbations.

The integration of artificial intelligence into computer-aided design frameworks has sparked a shift in the design of analog integrated circuits (ICs), transitioning the field from using manual and algorithmic-based solutions to adopting automated and intelligent paradigms. In this scenario, the GDSII file represents the industry-standard database containing the ultimate and most accurate source of information of the analog circuit, encapsulating the complex physical geometries and parasitic realities that define tape out performance. This paper proposes THEIA, a novel dataset containing thousands of layout images paired with question-answer conversations, along with a benchmark that employs a fine-tuned vision-language model (VLM) to analyze GDSII files of analog circuits, enabling designers to interact with and query physical layouts as intuitive, meaningful entities. Experimental results using thousands of analog designs across five realistic tasks demonstrate that the proposed fine-tuned VLM outperforms state-of-the-art general-purpose VLMs by a significant margin (up to 73%), highlighting a fundamental gap between general-purpose multimodal reasoning and domain-specific layout understanding.

Efficient exploration often remains a central bottleneck in reinforcement learning with verifiable rewards (RLVR). Although temperature control and test-time scaling strategies can increase rollout diversity of large language models (LLMs), they either expand the sample budget at rollout time or leave the benefit of exploration unquantified. To this end, we propose Temperature-Grouped Reinforcement Learning (TGRL), which turns temperature-induced diversity into an explicit training signal. For each prompt, TGRL partitions its rollout group into low- and high-temperature subsets, estimates exploration gain through their reward contrast, and allocates this group-level signal as token-level credit using Jensen--Shannon (JS) divergence between the corresponding temperature-scaled next-token distributions induced by the same logits. Notably, TGRL reaches equivalent accuracy up to 36% faster than strong RLVR baselines without expanding the rollout budget. Across 11 benchmarks from diverse domains, TGRL broadly improves over strong RLVR baselines: it improves the six-benchmark math average by 1.6% at 32B, raises CodeForces rating by 196.7 points and LiveCodeBench Pass@16 by 4.4%, and improves ALFWorld/WebShop success rates by 6.3%/4.9%. Comprehensive ablations and wall-clock analysis confirm the efficacy of all proposed components. Code is available at https://anonymous.4open.science/r/TGRL-4877/.

Robot task planning in real-world environments requires mapping abstract natural language instructions to executable action sequences under long horizons and complex constraints. While large language models (LLMs) provide strong commonsense reasoning, they often fail to generate reliable and feasible plans. In contrast, symbolic planners ensure feasibility and optimality but require well-specified goals and cannot directly interpret high-level human intent. We formulate robot task planning as learning to generate verifiable subgoals in the Planning Domain Definition Language (PDDL), bridging language understanding and symbolic planning. To address the challenges of sparse and noisy supervision, we propose Trace-Guided Policy Optimization (TGPO), a reinforcement learning framework that improves structured subgoal generation through (i) verifier-grounded rewards, (ii) external correction of intermediate reasoning traces, and (iii) constrained policy updates that incorporate corrected traces into training. We evaluate TGPO on large-scale household planning tasks with long horizons, abstract instructions, and complex constraints. TGPO significantly outperforms prompting-based and reinforcement learning baselines, with the largest gains on abstract tasks. Furthermore, TGPO integrates naturally with symbolic planners and language-conditioned executors, enabling robust long-horizon planning and combinatorial generalization in realistic environments.

Learning control policies for complex, long-horizon tasks is a central challenge in autonomous systems. Signal Temporal Logic (STL) offers an expressive language for specifying such tasks, but its non-Markovian nature and inherent sparse reward make it hard for standard Reinforcement Learning (RL) to solve. Prior RL approaches focus only on limited STL fragments or use robustness scores as sparse rewards. Our new method TGPO (Temporal Grounded Policy Optimization) decomposes STL into timed subgoals and invariant constraints and tackles the problem in a hierarchical fashion. Its high-level module proposes time allocations for subgoals, and the low-level time-conditioned policy learns to achieve the sequenced subgoals using a dense, stage-wise reward. In inference, we use the critic to efficiently sample time allocations and select the most promising assignment for the policy to rollout. We evaluate in five environments, from low-dimensional navigation to manipulation, drone, and quadrupedal locomotion, and TGPO significantly outperforms baselines (especially for high-dimensional and long-horizon cases), with 31.6% higher success rate compared to the best method.

Partially Relevant Video Retrieval (PRVR) aims to retrieve untrimmed videos containing moments relevant to a given text query. Despite recent progress, existing PRVR methods suffer from two key limitations: a fixed video decomposition scheme that causes semantic dilution, and weak cross-domain robustness due to task-specific training. In this paper, we propose TF-PRVR, the first training-free framework for PRVR designed to improve real-world generalization. TF-PRVR leverages frozen vision-language features to construct video-specific hierarchical representations. It derives temporal semantic signals from frame-level features and applies frequency-based multi-scale analysis to identify adaptive temporal boundaries, producing hierarchical segments with coherent event-level semantics. Built on these segments, TF-PRVR constructs a unified multi-scale graph and propagates query relevance across temporally and semantically related nodes. A moment-aware scoring strategy then aggregates temporally aligned relevance across scales, emphasizing consistently supported moments while suppressing isolated false responses. Without task-specific training, TF-PRVR preserves the general-purpose alignment capability of pre-trained vision-language models and avoids dataset-specific overfitting. Extensive experiments on standard PRVR benchmarks demonstrate competitive retrieval performance and strong robustness under cross-domain evaluation, suggesting a practical direction for real-world PRVR.

Diffusion models have achieved remarkable success in image, video, and 3D content creation. However, extending these successes to native 4D model generation remains challenging due to the intricate coupling between temporal geometry and appearance dynamics. Existing approaches typically generate 4D content by synthesizing videos from novel viewpoints, yet they often suffer from poor spatio-temporal consistency across views and time. To address these limitations, we introduce TESLA, a novel feedforward approach for generating native 4D Gaussian Splatting (4DGS) representations. Our key insight is to decouple temporal geometry and appearance generation through a temporally structured latent representation. This decomposition allows us to first establish coarse temporal geometry that captures fundamental structural dynamics, and then synthesize fine-grained appearance details conditioned on this geometric scaffold. To bridge this decoupled representation to the final output, we design a native 4DGS decoder that directly transforms our temporal features into native 4DGS representations, enabling flexible temporal rendering across arbitrary viewpoints and timesteps. Trained on a large collection of curated animated sequences, we extensively evaluate TESLA on the 4DGS generation task, showing superior performance over existing baselines.

Large Reasoning Models (LRMs) achieve impressive performance on complex reasoning tasks via Chain-of-Thought (CoT) reasoning, which enables them to generate intermediate thinking tokens before arriving at the final answer. However, LRMs often suffer from significant overthinking, spending excessive compute time even after the answer is generated early on. Prior work has identified the existence of an optimal reasoning length such that truncating reasoning at this point significantly shortens CoT outputs with virtually no change in performance. However, determining optimal CoT lengths for practical datasets is highly non-trivial as they are fully task and model-dependent. In this paper, we precisely address this and design Terminator, an early-exit strategy for LRMs at inference to mitigate overthinking. The central idea underpinning Terminator is that the first arrival of an LRM's final answer is often predictable, and we leverage these first answer positions to create a novel dataset of optimal reasoning lengths to train Terminator. Powered by this approach, Terminator achieves significant reductions in CoT lengths of 14%–55% on average across four challenging practical datasets: MATH-500, AIME 2025, HumanEval, and GPQA, while outperforming current state-of-the-art methods and reducing inference latency by more than 2$\times$ compared to the original LRM.

In unsupervised multivariate time series anomaly detection (MTSAD), local variation can inflate reconstruction and forecasting errors, making observation-space discrepancy an unreliable indicator of abnormal system behavior. Since latent-space scoring can mitigate this limitation, we reframe one-class MTSAD as current-point system-state compatibility estimation and propose **TENET**, the **T**ime-point **E**ncoding **Net**work for Multivariate Time Series Anomaly Detection. We define the current-point system state as a latent representation of the target time point, constructed from variable-level temporal context in the input window. TENET constructs this state through a temporal layer that retrieves window context relative to the current point, encodes time with current-point-centered embeddings, and adaptively mixes the retrieved context with the current representation. With this design, TENET treats the input window not as the object to reconstruct, forecast, summarize, or score, but as context for constructing a latent representation evaluated with a fixed standard-normal negative log-likelihood score. Across benchmark datasets, TENET achieves state-of-the-art performance. Additional experiments further show that TENET remains effective under alternative one-class objectives and outperforms substitutions based on generic sequence encoders, suggesting that its advantage comes from decision-aligned representation construction rather than from a specialized scoring rule. These results indicate that one-class MTSAD can be strong when the latent representation is designed around the current time point.