NeurIPS 2026 Papers

Lifelong learning requires embodied agents to continuously understand the environment and learn how to use that understanding to guide action, enabling future interactions to yield more informative experiences. Prior language model (LM)-based agents have primarily addressed these requirements in isolation, emphasizing either declarative knowledge acquisition or procedural knowledge reuse, leaving the continual interplay between the two during lifelong learning underexplored. We introduce Neurosymbolic Knowledge Compilation (NeSyKC), a lifelong learning framework that couples declarative knowledge formation with incremental compilation into procedural knowledge through a shared symbolic representation. In NeSyKC, accumulated experience is abstracted into symbolic rules that capture reusable environmental structure, and these rules provide the basis for deriving executable procedures that are incrementally compiled into neural adapters. This creates a feedback loop in which experience is abstracted into symbolic rules and compiled into procedures that guide the LM's reasoning on future tasks, yielding further experience for learning. Across four open-ended environments instantiated from established embodied benchmarks, NeSyKC expands and refines reusable knowledge across tasks, improving task performance and inference efficiency while supporting generalization to unseen tasks. Real-world experiments on two robots further show that the agent continues to learn from post-deployment experience, allowing acquired knowledge to be compiled for reuse in future tasks.

Survival models are often trained at target clinical sites where outcomes are limited and censored, while patient-level data from external cohorts may be unavailable because of institutional data-sharing constraints. In many applications, the transferable knowledge is not a dataset, a calibrated survival curve, or estimated regression coefficients, but a probability-free prognostic artifact such as a clinician-defined scorecard, guideline-based staging rule, registry-derived risk index, or coarse clinical risk group. These weak teachers may encode human expertise or evidence from prior cohorts, yet they often have only ordinal meaning and need not share a numerical scale with the target model. We propose NNCoxKL, a transfer-learning framework that trains a deep Cox model on target time-to-event data while distilling such human- or model-derived prognostic signals through Cox risk sets. For each event risk set, NNCoxKL converts both the external signal and the neural Cox risk score into Plackett--Luce distributions and regularizes the Cox partial likelihood through a temperature-scaled KL alignment term. This transforms clinical knowledge into a censoring-aware ranking prior over who is most likely to fail next, rather than treating the external score as a covariate or requiring it to represent calibrated survival probabilities. The external signal enters only through this risk-set regularizer, allowing the model to learn nonlinear target-cohort effects while borrowing relative-risk information from external clinical knowledge. Experiments on independent clinical transfer tasks and controlled survival benchmarks show that NNCoxKL improves discrimination and calibration-sensitive prediction over internal-only deep Cox models and common score-transfer baselines.

Recent advances in agent-based methods have demonstrated strong promise for issue localization, a critical prerequisite for software issue resolution. However, most existing agent-based methods rely on a single agent with a growing context, where long-context accumulation compresses the effective reasoning space. Meanwhile, the flat exploration structure hinders the balance between file-level breadth and function-level depth. To address these limitations, we propose NINJA, a hierarchical Navigator-INspector Joint Architecture for context-efficient issue localization. NINJA decomposes repository exploration into global navigation and local inspection: the navigator maintains the global search state, performs file-level search, and dispatches selected entry files, while inspectors independently conduct function-level exploration around the assigned files in separate contexts. Through multi-round interactions, inspectors return suspicious locations as feedback, and the navigator updates the global state to decide whether to continue exploration or finalize localization. This hierarchical design balances file-level breadth with function-level depth, while independent inspector contexts prevent local exploration traces from accumulating in a single growing context. To further strengthen both global coordination and local exploration, we introduce a two-stage agentic fine-tuning strategy. Extensive experiments across multiple benchmarks and LLM backbones show that NINJA consistently outperforms competitive baselines. Notably, after fine-tuning, Qwen3-Coder-30B-A3B-Instruct surpasses the strong closed-source Claude-Haiku-4.5 model. Our code is available at https://anonymous.4open.science/r/NINJA-67FB/.

Interactive narrative is becoming a practical frontier for language-model agents, spanning game NPCs, collaborative story production, emotionally responsive companions, and story-grounded interface generation. Yet most current evaluations still rely on static prompts, isolated story outputs, or post-hoc ratings, leaving open whether frontier models can sustain coherent, adaptive, emotionally grounded interaction over time. We introduce NARRA-Gym, an executable evaluation environment that turns a sparse emotional seed into a complete interactive story episode and records the full agent loop: story construction, memory updates, planning, pacing control, and optional artifact generation. The benchmark jointly tests five coupled capabilities---creative story generation, long-context state tracking, character simulation, empathic personalization, and interactive artifact generation---through a fixed LLM-as-judge sweep and a human preference study with user-provided experiences. Across nine generator models and eight benchmark personas, Claude Sonnet 4.6 is the strongest and most robust performer, while Claude Opus 4.6 forms a high-variance second tier and several middle-tier models show different tradeoffs between story quality and user experience. Human rankings recover the same top tier but shift the middle of the ordering, indicating that automated judges are useful for broad screening while human preference remains important for style, tone, and emotional resonance. Finally, failure analysis shows that the key remaining bottleneck is not surface fluency but resistance-sensitive personalization: models can write polished scenes while losing track of what the user is resisting or emotionally unable to accept.

Catastrophic forgetting is a central obstacle in single-stage fine-tuning of large language models (LLMs) on a new task, especially when the original pretraining data are unavailable. Synthetic data offers a practical mitigation path, but candidate pools are often noisy and redundant, making effective selection essential. Existing methods often rely on indirect proxy signals whose alignment with actual training benefit can be limited, while utility--diversity selection over large candidate pools remains computationally demanding. To address these challenges, we propose $\textbf{NA}$vigator-guided $\textbf{D}$ata $\textbf{S}$election ($\textbf{NADS}$), a fine-tuning framework for LLMs. NADS first follows the target fine-tuning recipe without preservation regularization to obtain a navigator model, whose deviation from the pretrained model exposes recipe-induced capability drift. It then uses the predictive divergence between the navigator and the pretrained model to score forgetting-aware utility for each synthetic candidate, and constructs a constraint set through an efficient utility--diversity selection strategy. Finally, NADS fine-tunes on the new-task data while distilling pretrained behavior on the selected constraints to preserve general capabilities. Experiments show that NADS delivers a stronger balance between new-task performance and general capability preservation, while reducing utility--diversity selection overhead.

Athletic performance represents the pinnacle of human motor intelligence, demanding rapid choices, precise control, agility, and coordinated physical execution. Replicating this seamless combination of capabilities remains elusive in current artificial intelligence and robotic systems. Concurrently, understanding the biological mastery of these movements is hindered because complex muscle coordination is rarely measured in vivo due to the limitations of physical equipment. To bridge this fundamental gap in understanding, $\textit{MyoChallenge}$ at NeurIPS 2025 established a pioneering benchmark for motor control intelligence in sports, leveraging high-fidelity musculoskeletal models within physics simulation combined with machine learning-driven algorithms. The competition introduces two distinct tracks emphasizing either upper or lower limbs control: a table tennis rally task utilizing a biomechanic upper limb composed of an arm with a hand (MyoArm) and a trunk (MyoTorso); and a soccer penalty kick using a biomechanic model of legs (MyoLeg) and a trunk (MyoTorso). Marking the fourth iteration of the MyoChallenge series, this event attracted almost 70 teams and over 560 submissions globally, uniting a diverse community ranging from physicians and neuroscientists to machine learning experts. The competition facilitated the development of several state-of-the-art control algorithms for a musculoskeletal system capable of sports agility, leveraging techniques such as physics-based motion planners, on-policy behaviour cloning, hierarchical planning, and muscle synergies that significantly surpassed our provided baseline performance. By integrating standardized tasks and physiologically realistic models into the open-source framework of MyoSuite, MyoChallenge'25 serves as a reproducible and reusable testbed to accelerate interdisciplinary research across machine learning, biomechanics, sports science, and neuroscience. Project page: https://www.myosuite.org//myochallenge/myochallenge-2025

Due to the challenge of uncovering biologically meaningful spatial tissue patterns from standard H\&E-stained whole-slide images (WSIs) without relying on multiplexed staining or fixed-scale modeling, a unified framework is proposed for spatial biomarker discovery that integrates cell classification, adaptive multi-scale microenvironment construction, and colocalization pattern analysis. Cell-level predictions assign discrete, semantically meaningful types that serve as the basis for microenvironment modeling, where physical adjacency initializes local ecological features and a combination of spatial competition-based seed selection with homogeneity-constrained recursive region growing allows microenvironments to expand from locally pure regions in a data-driven manner, capturing heterogeneity across scales without predefined spatial ranges. Microenvironment-level cell composition features are then used to construct a colocalization feature space, explicitly modeling spatial relationships among cell types, and differential biomarkers are identified via a distributional comparison framework that systematically detects group-specific colocalization patterns. Experiments demonstrate that this framework reliably discovers discriminative and biologically interpretable spatial biomarkers from H\&E images alone, exhibiting robustness in tissues with complex structures and heterogeneous scales, and offering a flexible, extensible paradigm for spatial phenotypic biomarker discovery in standard histopathology.

Cross-modal person re-identification (ReID) adversarial attacks face a fundamental generalization dilemma: existing methods suffer from source-pair overfitting, where the learned perturbations become entangled with the feature covariance and alignment patterns of the source modality pair, thereby limiting their transferability under heterogeneous domain shifts. We argue that the core issue is the difficulty of reducing source-pair-specific bias while maintaining attack effectiveness. To address this, we propose Multiform Attack (MA). The framework first learns a universal attack direction via Mahalanobis-guided gradient optimization to capture the intrinsic covariance structure of the feature manifold, though this perturbation may still retain source-pair-specific bias. To overcome this bias and achieve cross-distribution generalization, the second stage leverages multiple heterogeneous source distributions to optimize for generalization, performing sparse, structure-sensitive residual correction in the discrete pixel space via attention-shift-filtered multi-objective evolutionary search. This two-stage design models the attack as a base direction plus a structured residual, which helps reduce overfitting to the source modality pair and improve transferability to unseen modalities. Extensive experiments show that MA achieves superior transferability across unseen modalities, datasets, and model architectures.

LiDAR scene flow estimation is essential for autonomous driving, as it provides 3D motion for each point. Self-supervised approaches use static-dynamic classification to mitigate the imbalance between static and dynamic points, deriving targeted supervision. However, existing methods rely on sparse geometric observations for this classification, making them vulnerable to data sparsity and occlusions. The resulting noisy labels provide incorrect motion guidance and degrade scene flow learning. To address this, we introduce TrackCue, a tracking-guided framework for improving dynamic object representation in LiDAR scene flow estimation. In particular, TrackCue repurposes point tracking to obtain dense image-space trajectories anchored to LiDAR points, providing motion cues beyond sparse geometric observations. Furthermore, we present a visually consistent motion compensation strategy that compares the tracked trajectories with ego-induced rigid trajectories in the image plane, effectively isolating true object motion from ego-induced apparent motion. To transfer these isolated motion cues back to the LiDAR domain, we perform visual motion cue lifting, which associates ego-compensated image trajectories with LiDAR points for static-dynamic label refinement. As a result, TrackCue produces more accurate static-dynamic classification and provides more reliable supervision for scene flow learning. Experimental results show that TrackCue significantly improves the precision and F1 score of dynamic labels, leading to performance gains in self-supervised scene flow estimation.

Differentiable physics solvers underpin solver-in-the-loop ML training, gradient-based optimal control, and inverse problems, yet the practical cost of obtaining correct, usable gradients from a given solver on a given problem is largely undocumented. Integration effort, computational cost, gradient accuracy, and numerical conditioning vary widely across solvers and are discoverable only by trial and error. We introduce Mosaic, an extensible benchmarking framework for differentiable PDE solvers that standardizes access to solver gradients. Each solver is packaged as a containerized component (Tesseract) exposing a uniform gradient API regardless of language or automatic differentiation (AD) strategy, enabling researchers to evaluate, compare, and build on non-trivial physical solvers. Our evaluation of 14 solvers across fluid dynamics, structural mechanics, and heat transfer demonstrates that the benchmark surfaces practically relevant differences: order-of-magnitude variation in computational cost and Jacobian conditioning, alongside structural incompatibilities that eliminate solvers from realistic tasks entirely. Despite this variation, all solvers that produce gradients converge to similar optima, indicating that the practical barriers are memory limits, numerical stability, and setup compatibility rather than gradient accuracy alone. Mosaic is open-source and available at https://anonymous.4open.science/r/mosaic-bench.

Microscopy images contain rich information about how cells respond to perturbations, making them essential to applications like drug screening. To quantify images, researchers often use representation extraction methods, and recent years have seen a proliferation of deep learning methods. While measuring the quality of these representations is essential, evaluation remains fragmented, with each proposed model evaluated on different tasks and datasets, using custom pipelines and metrics, making it difficult to fairly compare models. Here, we introduce MorphoHELM, a comprehensive open benchmark for evaluating feature extraction methods for cell painting, the most widely-used morphological profiling assay. MorphoHELM consolidates evaluation standards in the field, extends and corrects them to be more robust, and evaluates on the widest range of methods to date. A defining feature of the benchmark is that each task is evaluated at different degrees of batch effects (or technical noise), directly quantifying how the ability of methods to detect biological signal degrades as noise increases. Together, these properties enable MorphoHELM to detect trade-offs between methods, and we demonstrate that models that excel at certain kinds of biological signal are weaker at others. We show that no existing model outperforms classic computer vision analytic strategies across all settings, which remain the strongest general use-case representations. All datasets, code, and evaluation tools are publicly available.

Autoregressive generation in Large Language Models (LLMs) is constrained by the memory and computational demands of attention mechanisms. Sparse attention methods mitigate this cost by selecting only high-probability entries of the attention matrix. We observe that in many such methods, this renders the probability-value multiplication negligible, shifting the bottleneck to the query-key step. Key heads can therefore be reduced to accelerate inference, while value heads preserve capacity at no additional cost. We introduce Sparse Asymmetric Group-Query Attention (SAGA), which decouples key and value head counts to exploit this principle, and pair it with approximate top-N (Atop-N) attention, a simple GPU-agnostic sparse attention method designed to isolate sparsity's effect on decoding-optimized architectures. We formalize the benefits of this asymmetry theoretically and validate it empirically, achieving end-to-end decoding speedups exceeding $2\times$ at long contexts across multiple model scales and benchmarks. Models trained from scratch with SAGA nearly match the quality of comparable GQA variants while delivering substantial efficiency gains. To facilitate adoption, we introduce an efficient fine-tuning method that converts pretrained models to the SAGA architecture at a small to negligible cost to quality, enabling practitioners to benefit from our approach without costly retraining.

We introduce a notion of monoculture robustness in Probably Approximately Correct (PAC) learning, as a way to mitigate the problem of *algorithmic monoculture* (Kleinberg & Raghavan, 2021): the tendency of models sharing the same components (e.g., training data) to fail on the same inputs. Specifically, we ask whether a learner with access to a single dataset can output $k$ individually accurate hypotheses whose joint failure probability is comparable to training on $k$ independent datasets. We say that such a learning procedure achieves *monoculture robustness*, and we quantify this by bounding the *conjunctive error* that all $k$ hypotheses misclassify points in an individual sense. We design methods for achieving this type of guarantee by drawing a surprising connection between replicability (Impagliazzo et al., 2022) and monoculture robustness. Specifically, we show that if the base learner is $\rho$-replicable, then running it independently $k$ times on the same sample yields individual monoculture error that shrinks exponentially with $k$, controlled by the single-run error of the base learner at the point and $\rho$. We complement these positive results with a lower bound: for individual monoculture robustness, we prove a worst-case lower bound showing that, for learners, driving the individual conjunctive error down to $\gamma$ can require a sample-size multiplier of $t = \Omega(\log(1/\gamma))$.

Continual learning (CL) aims to adapt models to new tasks while preserving previously learned knowledge. Recently, Low-Rank Adaptation (LoRA), a representative Parameter-Efficient Fine-Tuning (PEFT) method, has gained increasing attention in CL due to its efficiency and scalability. Several LoRA-based merging methods maintain a single inference model by scaling the newly learned LoRA update with one coefficient and merging it into the previous weights after each task. However, they usually control the whole update with a shared scalar factor, although different update elements may affect previous and current tasks in different ways. To address this issue, we propose Merging Strength Modulation via Joint Loss Estimation (M²LE) It consists of Hessian Information-guided Adaptive Merging (HIM) and Task Simulation Perturbation for Merging (TSP). Instead of searching for a global merge coefficient, HIM directly solves for the target merged update under a joint objective over previous and current tasks. Solving this objective gives an analytic solution that naturally takes the form of element-wise modulation over the current task LoRA update. To improve the stability of the learned update under subsequent merging, we introduce TSP which simulates different tasks interference. Experiments on multiple benchmarks show that M²LE consistently outperforms existing PEFT-based continual learning methods. Our code is provided in the supplementary material.

Dynamic runtime latency and memory constraints necessitate flexible large language model (LLM) deployment, where an LLM can be inferred with various quantization precisions based on available computational resources. Recent work on such any-precision quantization either relies on hardware-inefficient vector quantization or induces additional scaling factors when switching between bit-widths. Meanwhile, existing post-training quantization (PTQ) methods calibrated for a fixed low precision show poor generalizability under runtime precision change. In this work, we attribute the source of poor generalization across bit-widths to a precision-dependent \textit{outlier migration} phenomenon where the distribution of PTQ-sensitive tokens changes across precisions. Motivated by this observation, we propose \texttt{MoBiQuant}, a novel any-precision Mixture-of-Bits quantization framework that adjusts weight precision for flexible LLM inference based on token sensitivity. Specifically, we propose a many-in-one recursive residual quantization that can iteratively reconstruct higher-precision weights at runtime and mitigates \textit{outlier migration} with a token-aware router to dynamically select the optimal inference precision of each token. Extensive experiments show that \texttt{MoBiQuant} matches or surpasses frontier single-precision PTQ while exhibiting strong elasticity, achieving significant memory savings and throughput gains of up to $1.34\times$ over state-of-the-art any-precision methods. Source code is provided in the supplementary material.

Neural Combinatorial Optimization (NCO) is a promising paradigm for solving routing problems, yet its generalization across diverse data distributions remains a critical bottleneck. Existing methods typically tackle cross-distribution generalization through complex multi-distribution training or intricate meta-learning schemes, implicitly presuming that single-distribution training yields a weak zero-shot baseline. In this paper, we demonstrate that the potential of single-distribution training has been substantially underestimated and can be unlocked by a simple architectural inductive bias: Mix Normalization. To approach this, we first construct a comprehensive benchmark spanning three representative routing problems and 179 fine-grained datasets, capturing diverse shifts in node coordinates, customer demands, and time windows. Building on this benchmark, we propose MixRoute, a neural framework that adaptively combines normalization statistics at different granularities to mitigate varying distribution shifts. Trained exclusively on uniform distributions, MixRoute generalizes effectively to a wide range of unseen distributions without any distribution-specific adaptation. Extensive experiments demonstrate that MixRoute consistently achieves state-of-the-art zero-shot generalization. Our work revisits normalization as a parsimonious alternative for zero-shot generalization against complex training-adaptation schemes.

We provide theoretical guarantees for convergence of discrete-time policy mirror descent with inexact advantage functions updated using temporal difference (TD) learning for entropy regularised MDPs in Polish state and action spaces. We rigorously derive sufficient conditions under which the single-loop actor-critic scheme is stable and convergent. To weaken these conditions, we introduce a variant that performs multiple TD steps per policy update and derive an explicit lower bound on the number of TD steps required to ensure stability. Finally, we establish sub-linear convergence when the number of TD steps grows logarithmically with the number of policy updates, and linear convergence when it grows linearly under a concentrability assumption.

We study two variants of the mirror descent-ascent (MDA) algorithm for solving min-max problems on the space of measures: simultaneous and alternating. We work under assumptions of convexity-concavity and relative smoothness of the payoff function with respect to a suitable Bregman divergence, defined on the space of measures via flat derivatives. We establish non-asymptotic convergence rates to mixed Nash equilibria, measured in the Nikaido-Isoda error, proving an $\mathcal{O}(N^{-1/2})$ rate for simultaneous MDA and an improved $\mathcal{O}(N^{-2/3})$ rate for alternating MDA. The main technical contribution is an infinite-dimensional dual space analysis that relates Bregman divergences on measures to dual Bregman divergences on spaces of bounded continuous functions, allowing us to control asymmetric commutator terms created by alternating updates. The results substantially generalize prior analyses restricted to bilinear objectives and also apply to nonlinear convex-concave problems on measure spaces, thereby providing a unified theoretical foundation for MDA in mean-field min-max optimization.

Despite rapid progress in artificial intelligence, current systems struggle with the interconnected challenges that define real-world decision making. Practical domains such as business management require open-ended optimization, actively learning environment dynamics from sparse experience, planning over long horizons in stochastic settings, and reasoning over spatial information. Yet no existing human-AI benchmarks assess how well agents integrate these challenges in a grounded decision-making context. To this end, we introduce Mini Amusement Parks (MAPs), an amusement-park simulator designed to evaluate an agent’s ability to model its environment, anticipate long-term consequences under uncertainty, and strategically operate a complex business. We provide expert human performance and a comprehensive evaluation of state-of-the-art agents, finding experts outperform these systems by 11.4× on easy mode and 15.3× on medium mode. Our analysis reveals persistent weaknesses in long-horizon planning, sample-efficient learning, spatial reasoning, and modelling uncertainty. By unifying these challenges within a single environment, MAPs offers a new foundation for benchmarking agents capable of adaptable decision making.