An abstract framework for neural flows with composition and separation structures is proven to universally approximate any operator, recovering ResNet and plain architectures via discretization.
super hub Mixed citations
Proceedings of the IEEE conference on computer vision and pattern recognition , pages=
Mixed citation behavior. Most common role is background (44%).
103
Pith papers citing it
Background
44% of classified citations
hub tools
citation-role summary
background 7
method 2
citation-polarity summary
claims ledger
- method This appendix supplies the predictive-variance machinery deferred from Section 2.5. Kriging conditional variance.Given an observation set Oj⊆{1,...,N}at sample j, the fixed-rank kriging conditional variance (Cressie and Johannesson, 2008) at a query locations∗is ˆv(s∗,j) :=ˆσ2 + ˆϕ(s∗)⊤ˆΛ cond(Oj)ˆϕ(s∗),(16) with plug-in conditional score covariance ˆΛ cond(Oj) := ( ˆΛ−1+ˆσ−2Φ⊤ OjΦOj )−1 .(17) WhenOj =∅(new-sample regime),ˆΛ cond =ˆΛ and ˆvreduces to the marginal variance; when Oj is the trainin
- background Multi-Generation Regression Outputs.For a minibatch of inputs {x1, x2, . . . , xB}, GRPO samples K independent generation trajectories for each input. This yields a set of numeric predictions: q(xi) = q1(xi), q2(xi), . . . , qK (xi) ⊤ ,(1) which naturally encode prediction variability. We summa- rize these outputs using their empirical mean: µ(xi) = 1 K KX k=1 qk(xi),(2) which provides a stable, low-variance estimate for each compared sample during reward computation. Batch-Level Relational C
- background (28) In particular,Γ j,R is not a fold trace (e.g., it excludesz(y) =|y 1|along{y 1 = 0}), and the trace locally separatesU j,R into two nonempty strict-sign sides. (A3) Non-redundancy of intersecting traces (essentiality).For everyj∈ Kwith Zj ∩int(P)̸=∅, there exists a point y∈Z j ∩int(P)such thatz k(y)̸= 0∀k∈ K \ {j}.(29) Equivalently, Zj ∩int(P)̸⊆ [ k∈K\{j} Zk.(30) Lemma 3 (Counting cells intersectingPvs.int(P))LetP⊂R d be a convex closed polytope withint(P)̸=∅, and letPbe any collection of f
- background In Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 770-778, 2016. Patrick Helber, Benjamin Bischke, Andreas Dengel, and Damian Borth. Eurosat: A novel dataset and deep learning benchmark for land use and land cover classification.IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 12(7):2217-2226, 2019. JulianIbarz, JieTan, ChelseaFinn, MrinalKalakrishnan, PeterPastor, andSergeyLevine. Howtotrainyour robot with deep reinforcement l
- background For FEDADAGRAD , we setβ1 =β2 = 0 (as typical versions of ADAGRAD do not use momentum). For FEDADAM and FEDYOGI we setβ1 = 0.9,β 2 = 0.99. While these parameters are generally 22 Published as a conference paper at ICLR 2021 Algorithm 5 FEDADAGRADFEDADAGRADFEDADAGRAD , FEDYOGIFEDYOGIFEDYOGI , and FEDADAMFEDADAMFEDADAM - Batched data Input:x0,v−1≥τ 2, optionalβ1,β 2∈ [0, 1) for FEDYOGI and FEDADAM fort = 0,··· ,T − 1 do Sample a subsetS of clients xt i =xt for each clienti∈S in parallel do fore =
- background Jingkang Yang, Pengyun Wang, Dejian Zou, Zitang Zhou, Kunyuan Ding, Wenxuan Peng, Haoqi Wang, Guangyao Chen, Bo Li, Yiyou Sun, et al. Openood: Benchmarking generalized out-of- distribution detection.Advances in Neural Information Processing Systems, 35:32598-32611, 2022. Jingkang Yang, Kaiyang Zhou, Yixuan Li, and Ziwei Liu. Generalized out-of-distribution detection: A survey.International Journal of Computer Vision, 132(12):5635-5662, 2024. 11 A TRAININGCONFIGURATION All experiments share a com
co-cited works
representative citing papers
Dataset distillation creates a tiny synthetic training set that, when used with a fixed network initialization, produces models whose performance approximates that of models trained on the full original dataset.
GeoMix achieves new state-of-the-art results in descriptor-free 2D-3D matching by adding directional embeddings, learnable global context nodes, and multi-detector training, cutting rotation and translation errors by up to 90% on standard benchmarks.
FlexiSLM is the first spoken language model supporting dynamic and controllable frame rates on speech input and output, outperforming fixed-rate 7B models at high quality and enabling faster inference at lower rates like 6.25 Hz.
Vision models show a phase transition where accuracy on a zero syntactic-distance global-semantics task collapses to chance beyond a critical scale and does not recover with larger models or data.
PatternGSL introduces a learnable specification language for sewing patterns that lets vision-language models reconstruct explicit, simulation-ready 3D garments from single images, backed by a new 300K paired dataset.
Derives ODE limits of Adam-DA showing that first- and second-order momentum parameters reverse their convergence roles in zero-sum games compared to minimization, validated on GAN experiments.
Proposes pointwise Riemannian Dimension from feature eigenvalues to derive tighter, representation-aware generalization bounds for deep networks in the nonlinear regime.
BESplit mitigates non-IID bias in split federated learning via evidential aggregation, bias-compensated client pairing, and dual-teacher distillation, outperforming prior methods on five benchmarks.
Presents first online L2D algorithm for multiclass classification with bandit feedback and varying experts, achieving O((n+n_e)T^{2/3}) regret generally and O((n+n_e)√T) under low noise.
High-probability generalization bounds for D-SGD are derived at the optimal rate O(1/sqrt(mn) log(1/δ)) via pointwise uniform stability across convex and non-convex settings.
Fully Looped Transformer stabilizes looped training up to 12 iterations via distributed inter-loop signals and attention injection, improving downstream performance by up to 13.2%.
The Sinkhorn treatment effect is a new entropic optimal transport measure of divergence between counterfactual distributions that admits first- and second-order pathwise differentiability, debiased estimators, and asymptotically valid tests for distributional treatment effects.
A Bayesian predictive model adaptively selects martingale factors to construct asymptotically log-optimal confidence sequences for bounded means while preserving anytime validity under misspecification.
Risk-controlled post-processing yields a threshold-structured policy that follows the baseline except where an oracle fallback sharply reduces conditional violation risk, achieving O(log n/n) expected excess risk in i.i.d. settings and exact risk control under exchangeability.
Empirical tests with quad-mesh filling indicate that decision regions in modern image classifiers are simply connected.
A complex-weight extension to the Equation Learner enables stable recovery of symbolic expressions containing real-domain poles and unconstrained use of singular operators such as division and logarithm.
NodePFN pre-trains on synthetic graphs with controllable homophily and causal feature-label models to achieve 71.27 average accuracy on 23 node classification benchmarks without graph-specific training.
LLaMA-Adapter turns frozen LLaMA 7B into a capable instruction follower using only 1.2M new parameters and zero-init attention, matching Alpaca while extending to image-conditioned reasoning on ScienceQA and COCO.
LLM.int8() performs 8-bit inference for transformers up to 175B parameters with no accuracy loss by combining vector-wise quantization for most features with 16-bit mixed-precision handling of systematic outlier dimensions.
A learned local rule (Weight Transformer) iteratively self-organizes task-network weights from local graph neighborhoods and generalizes across unseen MLP, CNN, and ResNet architectures up to ~2M parameters.
Activation-space divergence detects and corrects compromised LLM agents in multi-agent systems without interaction graphs or synchronized rounds, outperforming graph baselines especially under async stealthy attacks.
Entropy-adaptive per-class budgets let clients generate far fewer synthetic samples yet still close most of the accuracy gap caused by label skew in federated learning.
MOSAIC learns overlap-aware shared-specific representations, fits a first-stage predictor on overlapping data, and calibrates the gap using target-pattern samples, with non-asymptotic error bounds decomposing overlap size, calibration gap, and representation error.
citing papers explorer
-
Neural Flow Operators can Approximate any Operator: Abstract Frameworks and Universal Approximations
An abstract framework for neural flows with composition and separation structures is proven to universally approximate any operator, recovering ResNet and plain architectures via discretization.
-
Dataset Distillation
Dataset distillation creates a tiny synthetic training set that, when used with a fixed network initialization, produces models whose performance approximates that of models trained on the full original dataset.
-
GeoMix: Descriptor-Free Visual Localization via Global Context and Multi-Detector Training
GeoMix achieves new state-of-the-art results in descriptor-free 2D-3D matching by adding directional embeddings, learnable global context nodes, and multi-detector training, cutting rotation and translation errors by up to 90% on standard benchmarks.
-
FlexiSLM: A Dynamic and Controllable Frame Rate Spoken Language Model
FlexiSLM is the first spoken language model supporting dynamic and controllable frame rates on speech input and output, outperforming fixed-rate 7B models at high quality and enabling faster inference at lower rates like 6.25 Hz.
-
Can Machines Really See Objects in Images? A Study Based on Syntactic Distance and Visual Self-Referential Instances
Vision models show a phase transition where accuracy on a zero syntactic-distance global-semantics task collapses to chance beyond a critical scale and does not recover with larger models or data.
-
PatternGSL: A Structured Specification Language for Template-Free and Simulation-Ready 3D Garments
PatternGSL introduces a learnable specification language for sewing patterns that lets vision-language models reconstruct explicit, simulation-ready 3D garments from single images, backed by a new 300K paired dataset.
-
Understanding Dynamics of Adam in Zero-Sum Games: An ODE Approach
Derives ODE limits of Adam-DA showing that first- and second-order momentum parameters reverse their convergence roles in zero-sum games compared to minimization, validated on GAN experiments.
-
Pointwise Generalization in Deep Neural Networks
Proposes pointwise Riemannian Dimension from feature eigenvalues to derive tighter, representation-aware generalization bounds for deep networks in the nonlinear regime.
-
BESplit: Bias-Compensated Split Federated Learning with Evidential Aggregation
BESplit mitigates non-IID bias in split federated learning via evidential aggregation, bias-compensated client pairing, and dual-teacher distillation, outperforming prior methods on five benchmarks.
-
Online Learning-to-Defer with Varying Experts
Presents first online L2D algorithm for multiclass classification with bandit feedback and varying experts, achieving O((n+n_e)T^{2/3}) regret generally and O((n+n_e)√T) under low noise.
-
Unveiling High-Probability Generalization in Decentralized SGD
High-probability generalization bounds for D-SGD are derived at the optimal rate O(1/sqrt(mn) log(1/δ)) via pointwise uniform stability across convex and non-convex settings.
-
Simply Stabilizing the Loop via Fully Looped Transformer
Fully Looped Transformer stabilizes looped training up to 12 iterations via distributed inter-loop signals and attention injection, improving downstream performance by up to 13.2%.
-
Sinkhorn Treatment Effects: A Causal Optimal Transport Measure
The Sinkhorn treatment effect is a new entropic optimal transport measure of divergence between counterfactual distributions that admits first- and second-order pathwise differentiability, debiased estimators, and asymptotically valid tests for distributional treatment effects.
-
Asymptotically Log-Optimal Bayes-Assisted Confidence Sequences for Bounded Means
A Bayesian predictive model adaptively selects martingale factors to construct asymptotically log-optimal confidence sequences for bounded means while preserving anytime validity under misspecification.
-
Risk-Controlled Post-Processing of Decision Policies
Risk-controlled post-processing yields a threshold-structured policy that follows the baseline except where an oracle fallback sharply reduces conditional violation risk, achieving O(log n/n) expected excess risk in i.i.d. settings and exact risk control under exchangeability.
-
Empirical Evidence for Simply Connected Decision Regions in Image Classifiers
Empirical tests with quad-mesh filling indicate that decision regions in modern image classifiers are simply connected.
-
Complex Equation Learner: Rational Symbolic Regression with Gradient Descent in Complex Domain
A complex-weight extension to the Equation Learner enables stable recovery of symbolic expressions containing real-domain poles and unconstrained use of singular operators such as division and logarithm.
-
Learning Posterior Predictive Distributions for Node Classification from Synthetic Graph Priors
NodePFN pre-trains on synthetic graphs with controllable homophily and causal feature-label models to achieve 71.27 average accuracy on 23 node classification benchmarks without graph-specific training.
-
LLaMA-Adapter: Efficient Fine-tuning of Language Models with Zero-init Attention
LLaMA-Adapter turns frozen LLaMA 7B into a capable instruction follower using only 1.2M new parameters and zero-init attention, matching Alpaca while extending to image-conditioned reasoning on ScienceQA and COCO.
-
LLM.int8(): 8-bit Matrix Multiplication for Transformers at Scale
LLM.int8() performs 8-bit inference for transformers up to 175B parameters with no accuracy loss by combining vector-wise quantization for most features with 16-bit mixed-precision handling of systematic outlier dimensions.
-
Architecture Generalization with MetaNCA
A learned local rule (Weight Transformer) iteratively self-organizes task-network weights from local graph neighborhoods and generalizes across unseen MLP, CNN, and ResNet architectures up to ~2M parameters.
-
When Agents Go Rogue: Activation-Based Detection of Malicious Behaviors in Multi-Agent Systems
Activation-space divergence detects and corrects compromised LLM agents in multi-agent systems without interaction graphs or synchronized rounds, outperforming graph baselines especially under async stealthy attacks.
-
WHERE to Generate Matters: Budget-Aware Synthetic Augmentation for Label Skewed Federated Learning
Entropy-adaptive per-class budgets let clients generate far fewer synthetic samples yet still close most of the accuracy gap caused by label skew in federated learning.
-
Pattern-Calibrated Multimodal Prediction under Blockwise Missingness
MOSAIC learns overlap-aware shared-specific representations, fits a first-stage predictor on overlapping data, and calibrates the gap using target-pattern samples, with non-asymptotic error bounds decomposing overlap size, calibration gap, and representation error.
-
MixTTA: Low-Rank Cross-Channel Mixing for Reliable Test-Time Adaptation
MixTTA equips normalization layers with low-rank cross-channel transformations plus decoupling and spectral projections to correct cross-channel structural changes during test-time adaptation.
-
Why SGD is not Brownian Motion: A New Perspective on Stochastic Dynamics
SGD is reformulated via a master equation from discrete updates, producing a discrete Fokker-Planck equation that predicts non-stationary variance growth proportional to learning rate in flat Hessian directions.
-
Partial Fusion of Neural Networks: Efficient Tradeoffs Between Ensembles and Weight Aggregation
Partial fusion of neural networks — merging only the most similar neurons via partial optimal transport or clustering — yields a monotone, tunable accuracy/parameter tradeoff between weight aggregation and ensembles.
-
Divide and Contrast: Learning Robust Temporal Features without Augmentation
Di-COT is an unsupervised contrastive method that stochastically partitions time-series windows into overlapping sub-blocks to learn representations without augmentation, reporting SOTA results on classification and transfer tasks across multiple benchmarks while cutting training time.
-
STELLAR: Scaling 3D Perception Large Models for Autonomous Driving
STELLAR trains up to 500M-parameter multi-modal models on 50M driving scenes and reports empirical scaling trends plus new state-of-the-art results on the Waymo Open Dataset.
-
Symmetrization of Loss Functions for Robust Training of Neural Networks in the Presence of Noisy Labels
Symmetrizing cross-entropy produces the unique convex multi-class unhinged loss, which locally approximates other symmetric losses, and enables new interpolating losses SGCE and alpha-MAE with competitive performance on noisy-label benchmarks.
-
StruMPL: Multi-task Dense Regression under Disjoint Partial Supervision and MNAR Labels
StruMPL is a multi-task dense regression model that jointly addresses disjoint partial supervision, MNAR labels, and inter-task physical constraints for improved forest biomass estimation from Earth observation.
-
Low Latency Gaze Tracking via Latent Optical Sensing
A hardware prototype performs gaze estimation by optically encoding task-relevant features with a microlens array and mask, captured on a 4x4 phototransistor array and decoded by a small neural network, reaching 3.4 ms latency with competitive accuracy.
-
GCE-MIL: Faithful and Recoverable Evidence for Multiple Instance Learning in Whole-Slide Imaging
GCE-MIL is a backbone-agnostic wrapper that directly optimizes MIL evidence for sufficiency, necessity, and recoverability, yielding modest gains in Macro-F1 and C-index plus more faithful patch selection across many backbones and datasets.
-
SwAIther-Precip: Lead-Time-Aware Bias Correction Enables Kilometer-Scale Downscaling of Global AI Precipitation Forecasts over Switzerland
SwAIther-Precip uses lead-time-conditioned U-Net bias correction followed by diffusion-based generative downscaling to reduce CRPS by 48% and achieve ~4 km effective resolution from 0.25° AIFS forecasts.
-
Response-Conditioned Parallel-to-Sequential Orchestration for Multi-Agent Systems
Nexa learns a response-conditioned policy that starts with parallel agent execution and adds at most one round of sequential message passing via a predicted sparse DAG, strictly subsuming pure parallel mode.
-
Self-Pruned Key-Value Attention: Learning When to Write by Predicting Future Utility
SP-KV trains a utility predictor jointly with the LLM to dynamically prune low-utility KV cache entries, achieving 3-10x memory reduction during generation with negligible performance loss.
-
Understanding Generalization through Decision Pattern Shift
DPS quantifies deviation of per-sample decision patterns from class averages and shows linear correlation with generalization gaps while unifying degradation scenarios into a continuous trajectory.
-
SynIB: Informational Bottleneck for Maximizing Synergy in Multimodal Learning
SynIB is an information-theoretic objective that adds a penalty for unimodal confidence to standard task loss, improving accuracy on synergy-dependent examples by up to 7.8% across synthetic XOR tasks and five real-world multimodal benchmarks.
-
MambaPanoptic: A Vision Mamba-based Structured State Space Framework for Panoptic Segmentation
MambaPanoptic is a fully Mamba-based panoptic segmentation model that uses MambaFPN for multi-scale features and a QuadMamba kernel generator to outperform PanopticDeepLab and PanopticFCN on Cityscapes and COCO while using fewer parameters than Mask2Former.
-
DocAtlas: Multilingual Document Understanding Across 80+ Languages
DocAtlas introduces model-free rendering pipelines to create DocTag-annotated datasets across 82 languages and shows DPO adaptation improves multilingual performance without base-language degradation.
-
Spatial Adapter: Structured Spatial Decomposition and Closed-Form Covariance for Frozen Predictors
The Spatial Adapter equips frozen predictors with a spatially regularized orthonormal basis for residuals and derives a closed-form low-rank-plus-noise covariance for spatial prediction and kriging.
-
Towards Understanding Continual Factual Knowledge Acquisition of Language Models: From Theory to Algorithm
Theoretical analysis of continual factual knowledge acquisition shows data replay stabilizes pretrained knowledge by shifting convergence dynamics while regularization only slows forgetting, leading to the STOC method for attention-based replay selection.
-
Post-hoc Selective Classification for Reliable Synthetic Image Detection
ReSIDe generalizes logit-based confidence scores to intermediate layers of synthetic image detectors and uses preference optimization to aggregate them, cutting area under the risk-coverage curve by up to 69.55% under covariate shifts.
-
Head Similarity: Modeling Structured Whole-Head Appearance Beyond Face Recognition
Head Similarity extends identity recognition to structured whole-head similarity by capturing intra-identity appearance variations via hierarchical supervision on a weakly-labeled video benchmark.
-
PragLocker: Protecting Agent Intellectual Property in Untrusted Deployments via Non-Portable Prompts
PragLocker generates function-preserving but non-portable prompts for LLM agents via code-symbol semantic anchoring followed by target-model feedback noise injection.
-
Distributionally Robust Multi-Objective Optimization
DR-MOO adds distributional robustness to multi-objective optimization and gives single-loop MGDA algorithms reaching epsilon-Pareto-stationary points in O(epsilon^{-4}) samples for nonconvex problems.
-
When Life Gives You BC, Make Q-functions: Extracting Q-values from Behavior Cloning for On-Robot Reinforcement Learning
Q2RL extracts Q-values from a BC policy and applies Q-gating to enable efficient offline-to-online RL, outperforming baselines on D4RL/robomimic tasks and achieving up to 100% success on real-robot manipulation in 1-2 hours.
-
Local Intrinsic Dimension Unveils Hallucinations in Diffusion Models
Hallucinations in diffusion models are driven by local intrinsic dimension instabilities on the manifold, which Intrinsic Quenching corrects by deflating it.
-
On the (In-)Security of the Shuffling Defense in the Transformer Secure Inference
An attack aligns differently shuffled intermediate activations from secure Transformer inference queries to recover model weights with low error using roughly one dollar of queries.
-
PHALAR: Phasors for Learned Musical Audio Representations
PHALAR achieves up to 70% relative accuracy gain in stem retrieval over prior art using under half the parameters and 7x faster training by enforcing musical equivariances via spectral pooling and complex heads.