Neural network checkpoints have quietly become a large-scale data resource: millions of trained weight vectors now exist, each encoding task-, domain-, and architecture-specific knowledge. This position paper argues that model checkpoints should be treated as a first-class data modality, and that generative modeling in weight space should be standardized as a core machine learning primitive. Recent advances demonstrate that neural weights can be synthesized on demand, often matching fine-tuning performance while reducing adaptation cost by orders of magnitude. We contend that these results reflect an underlying structural fact: high-performing models occupy low-dimensional, highly structured regions of weight space shaped by symmetry, flatness, modularity, and shared subspaces. Building on this view, we organize existing methods into a standardized five-stage pipeline for weight-space generation and survey applications where the approach is already practical, such as parameter-efficient adaptation, mid-scale model synthesis, and on-device learning. We then confront alternative views, clarify current limits, and issue a concrete call to action. Our goal is to shift the community’s default mindset from optimizing models per task to sampling models from learned weight distributions, accelerating toward an era in which AI systems routinely generate other AI systems.
论文检索
输入标题、作者或关键词，从 13,033 篇学术成果中精准定位
Deep Learning · Algorithms
Prompt learning has recently emerged as a novel, parameter-efficient paradigm to tackle the missing modalities challenge. However, existing prompting methods often overlook the internal structural information within prompt vectors, limiting their effectiveness in guiding frozen backbone models under diverse missing modality scenarios. To address this limitation, we propose a Structured Prompt Refining (SPR) network that refines the internal structure of prompt vectors across multiple dimensions: (1) a Global Interaction Fusion Module captures bidirectional interactions across prompt layers, thereby mitigating sub-optimal adaptation from inconsistent guidance under missing modalities, (2) a Local Feature Refinement Module structures adjacent prompt vectors into coherent semantic units, leveraging local contextual relationships to maintain semantic integrity during modality absence, and (3) a Channel Feature Selection Module uses point-wise gating to adaptively suppress noise and enhance critical channels based on the specific missing modality. Using only 0.8% trainable parameters, SPR achieves significant improvements on three mainstream multimodal classification datasets. Notably, it surpasses state-of-the-art by 3.8% in F1-Macro on the MM-IMDB dataset, even at a 90% modality missing rate. Extensive experiments and in-depth ablations validate SPR's effectiveness and robustness under various missing conditions.
Applications · Robotics
Trajectory world models play a crucial role in robotic dynamics learning, planning, and control. While recent works have explored trajectory world models for diverse robotic systems, they struggle to scale to a large number of distinct system dynamics and overlook domain knowledge of physical structures. To address these limitations, we introduce *WestWorld*, a kno**W**ledge-**E**ncoded **S**calable **T**rajectory **World** model for diverse robotic systems. To tackle the scalability challenge, we propose a novel system-aware Mixture-of-Experts (Sys-MoE) that dynamically combines and routes specialized experts for different robotic systems via a learnable system embedding. To further enhance zero-shot generalization, we incorporate domain knowledge of robot physical structures by introducing a structural embedding that aligns trajectory representations with morphological information. After pretraining on 89 complex environments spanning diverse morphologies across both simulation and real-world settings, *WestWorld* achieves significant improvements over competitive baselines in zero- and few-shot trajectory prediction. Additionally, it shows strong scalability across a wide range of robotic environments and significantly improves performance on downstream model-based control for different robots. Finally, we deploy our model on a real-world Unitree Go1, where it demonstrates stable locomotion performance ([see our demo on the website](https://westworldrobot.github.io/)).
Deep Learning · Attention Mechanisms
We study two recurring phenomena in Transformer language models. First, \emph{massive activations}, where a small number of hidden channels attain extremely large values for a few tokens. Second, \emph{attention sinks}, where certain tokens attract a disproportionate share of attention across many heads and layers. We present a unified inference-time mechanism explaining how massive activations emerge and propagate through layers, and how normalization transforms these tokens into sparse, nearly fixed vectors that reshape the attention space and induce sink or no-sink behavior. We further conduct ablations on models trained from scratch to disentangle architectural and training factors governing both phenomena. We find that attention sinks persist across architectures and can arise even without massive activations. The normalization strategy primarily determines the emergence of massive activations, while head dimension and context length modulate the frequency of attention sink formation.
Applications · Robotics
We introduce Distribution Contractive Reinforcement Learning (DICE-RL), a framework that uses reinforcement learning (RL) as a “distribution contractor” to refine pretrained generative robot policies. DICE-RL turns a pretrained behavior prior into a high-performing “pro” policy by amplifying high-success behaviors from online feedback. We pretrain a diffusion-based policy for broad behavioral coverage, then finetune it with a stable, sample-efficient residual off-policy RL framework that combines selective behavior regularization with value-guided action selection. Extensive experiments and analyses show that DICE-RL reliably improves performance with strong stability and sample efficiency, enabling mastery of complex long-horizon manipulation skills both in simulation and on a real robot. Project website: [dice-rl-anonymous.github.io](https://dice-rl-anonymous.github.io/).
Deep Learning · Algorithms
Low-Rank Adaptation (LoRA) is the most widely adopted method for fine-tuning large language models. Notably, LoRA is inherently overparameterized: multiple pairs of low-rank factors can yield the same adapted weight matrix. We show—both theoretically and empirically—that these pairs exhibit significantly different condition numbers. As a result, converging to different loss minimizers directly impacts the convergence rate of LoRA. Building on this observation, we introduce Balanced Low-Rank Adaptation (BaLoRA), a variant of LoRA that projects iterates onto a balanced manifold. This manifold improves the conditioning of the loss landscape while preserving the adapted matrix. The projection step is computationally lightweight and integrates seamlessly into existing fine-tuning pipelines. Empirically, BaLoRA converges faster than standard LoRA and achieves superior performance across a range of fine-tuning tasks.
General Machine Learning · Evaluation
Large language models are beginning to introduce a new paradigm for compilation: instead of only assisting at the source level, they can operate directly on **intermediate representations (IRs)**, the compiler’s internal code representation, Early studies suggest that LLM-guided optimization can sometimes rival traditional compiler optimizations on selected programs, but evidence remains fragmented. Yet the community still lacks a rigorous IR-level benchmark that tests whether a model not only understands IR but can rewrite it under compiler-grade semantic constraints with meaningful performance impact. We present **CIRBench**, a benchmark of 800 curated IR instances spanning four compiler-oriented tracks: Analysis infers IR properties, Repair fixes invalid IR, Refactor applies a single semantics-preserving compiler optimization, and Transform performs performance-oriented rewrites, together mirroring core optimization responsibilities in modern compilers. CIRBench combines verifier, equivalence checking, and end-to-end performance measurement into a unified, layered correctness-aware evaluation of LLMs on IR. On six mainstream LLMs, CIRBench shows that current models fail on many IR analysis and rewriting instances and on median underperform the compiler baseline, but we also observe a maximum speedup of $4.96\times$ over -O3. These findings highlight both the opportunities and the remaining challenges of using LLMs inside optimizing compilers.
Deep Learning · Large Language Models
Reliable reward models (RMs) are critical for ensuring the safe alignment of large language models (LLMs). However, current RM evaluation methods focus solely on preference perception accuracies in given specific scenarios, obscuring the critical vulnerabilities of RMs in real-world scenarios. We identify the true challenge lies in assessing a novel dimension: Suitability, defined as conditional reliability under specific real-world perturbations. To this end, we introduce Reward Auditor, a hypothesis-testing framework specifically designed for RM suitability inference. Rather than answering “How accurate is the RM's preference perception for given samples?”, it employs scientific auditing to answer: “Can we infer RMs exhibit systematic vulnerabilities in specific real-world scenarios?". Under real-world perturbed scenarios, Reward Auditor quantifies statistical significance and effect size by auditing distribution degradation of RM preference perception confidence. This enables inference of both the certainty and severity of RM vulnerabilities across diverse real-world scenarios, thereby laying a solid foundation for building next-generation LLM alignment systems that are verifiably safe, more robust, and trustworthy.
Applications · Computer Vision
Concept erasure aims to prevent image generative models from producing unsafe content while preserving their general generative capability. Meanwhile, next-scale autoregressive (AR) image generation has recently emerged as a new generative paradigm characterized by next-scale prediction, for which concept erasure remains largely unexplored. In this paradigm, semantic information is highly compressed at early scales, leading to severe entanglement between unsafe and unrelated semantics. In this paper, we propose ScaleErasure, an inference-time concept erasure method that performs minimal intervention. ScaleErasure precisely selects and guides predicted logits that are most relevant to the unsafe concept, thereby enabling effective erasure under severe semantic entanglement. Specifically, ScaleErasure performs two additional forward passes conditioned on the unsafe concept and the corresponding safe concept, and leverages their outputs to guide the target logits away from unsafe concepts toward safe concepts. To enable precise and minimal intervention, logits selection and guidance are conducted across three dimensions: scales, tokens, and bit channels. Experiments demonstrate that ScaleErasure outperforms adapted baselines in the next-scale AR paradigm, achieving more precise concept erasure while largely preserving general generative capability.
Audio-Visual Segmentation (AVS) targets pixel level localization of sounding emitting objects in videos. However, existing models rely on dense cross-modal attention with quadratic computational cost, limiting their suitability for resource efficient deployment. Most efficiency oriented methods focus on backbone reduction and overlook the interaction module as the primary bottleneck. This paper proposes LightAVSeg, a lightweight framework that replaces heavy attention with a decoupled design for semantic filtering and spatial grounding, resulting in interaction costs that scale linearly with spatial resolution. Furthermore, we introduce an auxiliary alignment loss to enforce semantic consistency during training with zero inference overhead. Extensive experiments demonstrate that LightAVSeg achieves a new state-of-the-art among lightweight methods: with 20.5M parameters (~1/7 of AVSegFormer), it reaches 50.4 mIoU on the MS3 benchmark and enables efficient inference on a mobile processor.
Deep Learning · Large Language Models
Recent progress in spatial reasoning with Multimodal Large Language Models (MLLMs) increasingly leverages geometric priors from 3D encoders. However, most existing integration strategies remain passive: geometry is exposed as a global stream and fused in an indiscriminate manner, which often induces semantic-geometry misalignment and redundant signals. We propose GeoThinker, a framework that shifts the paradigm from passive fusion to active perception. Instead of feature mixing, GeoThinker enables the model to selectively retrieve geometric evidence conditioned on its internal reasoning demands. GeoThinker achieves this through Spatial-Grounded Fusion applied at carefully selected VLM layers, where semantic visual priors selectively query and integrate task-relevant geometry via frame-strict cross-attention, further calibrated by Importance Gating that biases per-frame attention toward task-relevant structures. Comprehensive evaluation results show that GeoThinker sets a new state-of-the-art in spatial intelligence, achieving a peak score of 72.6 on the VSI-Bench. Furthermore, GeoThinker demonstrates robust generalization and significantly improved spatial perception across complex downstream scenarios, including embodied referring and autonomous driving. Our results indicate that the ability to actively integrate spatial structures is essential for next-generation spatial intelligence.
Optimization · Zero-order and Black-box Optimization
Meta Black-Box Optimization (MetaBBO) has emerged as a promising paradigm by employing meta learning to automatically optimize the configurations of low-level black-box optimizers. Despite its potential, the generalization of MetaBBO remains significantly constrained when facing unseen, complex objective landscapes. We identify that this bottleneck stems from a restricted performance upper bound inherent in current training mechanisms: by learning from scratch in a self-supervised or unsupervised manner, meta optimizers are never exposed to advanced, high-quality optimization behaviors, forcing them to converge on suboptimal strategies. In this paper, we propose $\texttt{MetaDistill}$, a general MetaBBO training framework designed to lift the strategy ceiling through pretraining and test-time fine-tuning. In the pretraining stage, we represent high-quality strategies from classical algorithms as expert optimization trajectories and utilize them for diversity-preserving distillation, enabling the learnable optimizer to internalize advanced optimization behaviors. In the optional fine-tuning stage, we perform self-supervised fine-tuning as a warm-start procedure to further refine the distilled knowledge on unseen tasks. We evaluate our $\texttt{MetaDistill}$ framework on the BBOB test suite and three control tasks. The results demonstrate that $\texttt{MetaDistill}$ significantly improves the generalization ability of various learnable optimizers compared to their original training paradigms.
General Machine Learning · Everything Else
Dynamic Sparse Training (DST) methods train neural networks by maintaining sparsity while dynamically adapting the network topology. Despite the promise of reduced computation, DST methods converge significantly slower than dense training, often requiring comparable training time to achieve similar accuracy. We demonstrate both analytically and empirically that Batch Normalization (BN) adversely affects sparse training, and propose SparseOpt — a sparsity-aware optimizer — to address this. Experiments on ResNet models across CIFAR-100 and ImageNet demonstrate consistently faster convergence and improved generalization with our proposed method. Our work highlights the limitations of current normalization layers in sparse training and provides the first systematic study of the interaction between Batch Normalization, sparse layers, and DST, taking a significant step toward making DST practically competitive with dense training.
Finding human-understandable circuits in language models is a central goal of the field of mechanistic interpretability. We train models to have more understandable circuits by constraining most of their weights to be zeros, so that each neuron only has a few connections. To recover fine-grained circuits underlying each of several hand-crafted tasks, we prune the models to isolate the part responsible for the task. These circuits often contain neurons and residual channels that correspond to natural concepts, with a small number of straightforwardly interpretable connections between them. We study how these models scale and find that making weights sparser trades off capability for interpretability, and scaling model size improves the capability-interpretability frontier. However, scaling sparse models beyond tens of millions of nonzero parameters while preserving interpretability remains a challenge. In addition to training weight-sparse models de novo, we show preliminary results suggesting our method can also be adapted to explain existing dense models. Our work produces circuits that achieve an unprecedented level of human understandability and validates them with considerable rigor.
Deep Learning · Large Language Models
We introduce Context Tuning, a simple and effective method to significantly enhance few-shot adaptation of language models (LLMs) without fine-tuning model parameters. While prompt-based adaptation techniques have demonstrated the effectiveness of lightweight adaptation methods for LLMs, they typically initialize a trainable prompt or prefix with irrelevant tokens for the task at hand. In contrast, Context Tuning initializes the trainable prompt or prefix with task-specific demonstration examples, leveraging the model’s inherent In-Context Learning (ICL) ability to extract relevant information for improved few-shot learning performance. Extensive evaluations on benchmarks such as CrossFit, UnifiedQA, MMLU, BIG-Bench Hard, and ARC demonstrate that Context Tuning outperforms traditional prompt-based adaptation methods and achieves competitive accuracy to Test-Time Training with significantly higher training efficiency.
Deep Learning · Large Language Models
Scaling test-time compute enhances long chain-of-thought (CoT) reasoning, yet existing approaches face a fundamental trade-off between computational cost and coverage quality: either incurring high training expense or yielding redundant trajectories. We introduce The Geometric Reasoner (TGR), a training-free framework that performs manifold-informed latent foresight search under strict memory bounds. At each chunk boundary, TGR scores candidate latent anchors via a lightweight look-ahead estimate combined with soft geometric regularizers that encourage smooth trajectories and diverse exploration. Chunk-wise KV cache resets keep memory linear in chunk length. On challenging math and code benchmarks, TGR improves robust trajectory coverage, measured by the area under the Pass@$k$ curve (AUC), by up to 13 points on Qwen3-8B, with negligible overhead of about 1.1--1.3$\times$.
Optimization · Non-Convex
Recent developments in deep learning optimization have brought about radically new algorithms based on the Linear Minimization Oracle (LMO) framework, such as Muon and Scion. After over a decade of Adam's dominance, these LMO-based methods are emerging as viable replacements, offering several practical advantages such as improved memory efficiency, better hyperparameter transferability, and most importantly, superior empirical performance on large-scale tasks, including LLM training. However, a significant gap remains between their practical use and our current theoretical understanding: prior analyses (1) overlook the layer-wise LMO application of these optimizers in practice, and (2) rely on an unrealistic smoothness assumption, leading to impractically small stepsizes. To address both, we propose a new LMO-based framework called Gluon, capturing prior theoretically analyzed methods as special cases, and introduce a new refined generalized smoothness model that captures the layer-wise geometry of neural networks, matches the layer-wise practical implementation of Muon and Scion, and leads to state-of-the-art convergence guarantees. Our experiments with NanoGPT and CNN confirm that our assumption holds along the optimization trajectory, ultimately closing the gap between theory and practice.
Probabilistic Methods · Everything Else
A key challenge in probabilistic regression is ensuring that predictive distributions accurately reflect true empirical uncertainty. Minimizing overall prediction error often encourages models to prioritize informativeness over calibration, producing narrow but overconfident predictions. However, in safety-critical settings, trustworthy uncertainty estimates are often more valuable than narrow intervals. Realizing the problem, several recent works have focused on post-hoc corrections; however, existing methods either rely on weak notions of calibration (such as PIT uniformity) or impose restrictive parametric assumptions on the nature of the error. To address these limitations, we propose a novel nonparametric re-calibration algorithm based on conditional kernel mean embeddings, capable of correcting calibration error without restrictive modeling assumptions. For efficient inference with real-valued targets, we introduce a novel characteristic kernel over distributions that can be evaluated in $\mathcal{O}(n \log n)$ time for empirical distributions of size $n$. We demonstrate that our method consistently outperforms prior re-calibration approaches across a diverse set of regression benchmarks and model classes.
Optimization · Non-Convex
MARS has recently emerged as a state-of-the-art optimizer, consistently outperforming AdamW in large language model (LLM) training. It modifies the momentum-based variance reduction (MVR) update by introducing a multiplicative coefficient $\gamma$, which scales the momentum correction term. However, the existing theory of Yuan et al. (2025) does not explain why this modification improves the convergence of MARS over MVR. In this paper, we provide a rigorous theoretical explanation for the superiority of MARS over MVR. We introduce the novel similarity condition, called **$\gamma$-similarity**, which generalizes standard similarity and smoothness assumptions for analyzing stochastic algorithms. Under this condition, we derive gradient complexity guarantees for MARS, which explicitly depend on $\gamma$ and a $\gamma$-similarity constant $\delta_\gamma$. We prove that by appropriately tuning $\gamma \in [0,1]$, MARS achieves strictly lower complexity than MVR. Finally, experiments on GPT pretraining corroborate our theoretical findings, demonstrating that MARS with an optimal choice of $\gamma$ improves token efficiency over MVR, and yields substantial gains compared to AdamW.
Accuracy and Normalized Accuracy under Length Bias: Analysis, Guidelines, and a Bayesian Alternative
Multiple-choice benchmarks that rank candidate completions by conditional log-probability suffer from a length bias: because log-probabilities sum over tokens, longer answers tend to be penalized relative to shorter ones in practice. A common mitigation is to normalize scores by completion length, but we show empirically that this heuristic frequently over-corrects, introducing a bias toward longer answers instead. We first analyze these scoring rules, characterizing when standard and length-normalized accuracy are appropriate and how their length biases depend on the distribution of completion lengths. Motivated by this analysis, we introduce *Bayesian accuracy*, a scoring rule that computes the posterior probability of each candidate under an explicit prior over answer length, thereby removing linear length effects. Bayesian accuracy is a drop-in replacement for likelihood-based multiple-choice evaluation, requires no additional forward passes, and consistently exhibits lower empirical length bias than both standard and length-normalized accuracy across benchmarks and few-shot settings.