Scales attention for long contexts in LLMs.
problem Development of attention mechanisms for long context inference.
method Scale-invariant total attention and sparsity conditions, with a position-dependent transformation of logits.
result Scale-invariant attention scheme improves validation loss and long-context retrieval.
The paper explores how sinks and diagonal patterns prevent attention oversmoothing.
problem Preventing attention oversmoothing in neural networks.
method Analyzing geometric conditions and conditions for dense vs. sparse attention, proving equivalence between sinks and hard attention switch, and comparing the costs of sinks vs. diagonal patterns.
result Sinks and diagonal patterns effectively prevent attention oversmoothing, and diagonal patterns provide a more flexible approach.
Investigates the benefits of multi-head attention in Transformers, deriving convergence and generalization guarantees.
problem Underexplored dynamics of multi-head attention in Transformer training and generalization.
method Derives convergence and generalization guarantees for gradient-descent training of a multi-head self-attention model.
result Establishes conditions for initialization that ensure multi-head attention's realizability.
Generative models use kernel smoothing for conditioning on small example sets.
problem Improving generative models' performance with limited conditioning examples.
method Showed that cross-attention conditioning is equivalent to kernel smoothing, specifically a Nadaraya--Watson kernel smoother.
result The approach predicts and confirms three failure regimes for kernel-based conditioning.
We aim to better understand attention over nodes in graph neural networks (GNNs) and identify factors influencing its effectiveness. We particularly focus on the ability of attention GNNs to generalize to larger, more complex or noisy graphs. Motivated by insights from the work on Graph Isomorphism Networks, we design …
DAGR improves navigation by refining goal representations conditioned on the current state.
problem Goal-conditioned reinforcement learning lacks state awareness, leading to inefficient policy recovery.
method DAGR refines static goal embeddings into state-conditioned ones using gated cross-attention with a state-goal discrepancy map.
result DAGR improves navigation tasks on OGBench, matching or outperforming base methods.
OSA overcomes instability in skipless Transformers.
problem Instability in skipless Transformers using Softmax Self-Attention.
method OSA parametrizes attention matrix to be orthogonal via skew-symmetric matrix exponential.
result OSA allows for training non-causal Transformers without skip connections and normalisation layers.
In this paper, we present a novel approach for incorporating external knowledge in Recurrent Neural Networks (RNNs). We propose the integration of lexicon features into the self-attention mechanism of RNN-based architectures. This form of conditioning on the attention distribution, enforces the contribution of the most…
Graph attention is not always beneficial; conditions for perfect node classification are identified.
problem Understanding when graph attention mechanisms improve node classification performance.
method Theoretical analysis using Contextual Stochastic Block Models (CSBMs).
result Graph attention mechanisms are more effective when structure noise exceeds feature noise, and simpler graph convolution operations are better when feature noise predominates.
Linearized attention fails to converge to NTK limit even at large widths.
problem Understanding the convergence of attention mechanisms to the kernel regime.
method Analyzes linearized attention and its relationship to the NTK limit, considering practical widths and conditions.
result Linearized attention does not converge to its NTK limit at any practical width, revealing a fundamental trade-off.
Develops a mean-field theory for multi-head self-attention under cross-entropy training.
problem Mean-field analysis of multi-head self-attention under cross-entropy training.
method Mean-field theory for a simplified single-layer causal multi-head self-attention model.
result Proves a static finite-head approximation bound for the optimal risk.
Attention models can overfit without harming test performance.
problem Understanding benign overfitting in single-head attention models.
method Analyzing conditions for benign overfitting in a single-head softmax attention model.
result A single-head attention model can overfit without harming test performance under certain conditions.
Neural attention has become central to many state-of-the-art models in natural language processing and related domains. Attention networks are an easy-to-train and effective method for softly simulating alignment; however, the approach does not marginalize over latent alignments in a probabilistic sense. This property …
Model predicts Bitcoin prices influenced by market attention.
problem Predicting Bitcoin prices considering market attention.
method Model uses a mean-reverting Cox-Ingersoll-Ross process to model market attention, affecting Bitcoin volatility with a delay.
result The model provides semi-closed formulae for European call and put prices, and compares favorably to other models.
AGBoost uses attention weights to improve GBM for regression problems.
problem Improving gradient boosting machine for regression tasks.
method Attention-based modification of GBM with trainable attention weights.
result AGBoost achieves better performance on regression datasets.
We propose the Gaussian attention model for content-based neural memory access. With the proposed attention model, a neural network has the additional degree of freedom to control the focus of its attention from a laser sharp attention to a broad attention. It is applicable whenever we can assume that the distance in t…
We investigate Relational Graph Attention Networks, a class of models that extends non-relational graph attention mechanisms to incorporate relational information, opening up these methods to a wider variety of problems. A thorough evaluation of these models is performed, and comparisons are made against established be…
HANET combines LSTM and attention mechanisms for better financial forecasting.
problem Lack of distinct macroeconomic regimes in financial datasets.
method Hierarchical Cross-Attention mechanism integrating long-run macro contexts with high-frequency market dynamics.
result HANET outperforms neural forecasters, especially during turbulent periods.
A new model captures complex event data using attention and Fourier kernels.
problem Capturing complex non-linear temporal dependencies in discrete event data.
method Integrates attention mechanism into point processes' conditional intensity function and uses Fourier kernel embedding.
result Established theoretical properties and demonstrated competitive performance.
Study examines asset pricing using various attention models, finding global self-attention and sliding window sparse attention models perform well.
problem Traditional asset pricing models miss temporal dependency and short memory issues.
method Investigates RNN attention models with various attention mechanisms for large-cap US stocks.
result Global self-attention and sliding window sparse attention models outperform in deriving returns and hedging risks, especially during the pandemic.
Neural HMM with AGA captures multi-scale dynamics in financial markets.
problem Capturing multi-scale temporal dynamics in financial markets.
method Parallel multi-resolution encoders, adaptive gating, and multi-head attention.
result Outperforms fixed-resolution baselines in predicting price movements and liquidity shocks.
Stochastic attention learns to retrieve and generate from memory without training.
problem Learning to retrieve and generate from memory efficiently.
method Langevin dynamics on modern Hopfield energy for stochastic attention.
result Stochastic attention can retrieve and generate from memory without training, matching gold standards.
A new perspective on self-attention models using MLPs.
problem Improving sequence modeling with self-attention mechanisms.
method Introducing HyperMLP and HyperGLU, which use dynamic two-layer MLPs with reverse-offset layout.
result HyperMLP/HyperGLU consistently outperform softmax-attention baselines.
AGNN improves network localization accuracy by 37-53% in NLOS conditions.
problem Massive network localization under Non-Line-of-Sight conditions.
method Attentional Graph Neural Network (AGNN) with Adjacency Learning Module (ALM) and Multiple Graph Attention Layers (MGAL).
result Significant improvement in localization accuracy, approaching fundamental lower bounds.
A new LSTM architecture improves time series forecasting efficiency.
problem Efficiency and accuracy in time series forecasting using linear models.
method Attention-free LSTM architecture for time series prediction.
result Improved prediction capacity and efficiency compared to LSTM.
TabPFN model shows strong robustness to noisy data.
problem TabPFN tackles robustness to noisy and imperfect tabular data.
method Empirical robustness analysis of TabPFN's attention mechanisms under various perturbations.
result TabPFN maintains high predictive performance and coherent internal behavior under noisy and imperfect data.
Transformers' self-attention mechanism is mapped to a generalized Potts model.
problem Uncertainty in what type of data distribution self-attention can efficiently learn.
method Decouple word positions and embeddings, then show self-attention learns a generalized Potts model.
result Training self-attention is equivalent to solving the inverse Potts problem.
EFA extends self-attention to handle mixed data types and dynamic relevance.
problem Handling high-dimensional, mixed data types with dynamic relevance.
method Probabilistic generative model using self-attention and latent factor model.
result EFA consistently outperforms existing models in complex latent structure capture and reconstruction.
Extracts important peaks from XRD spectra using Attention mechanism.
problem Identifying significant peaks in XRD patterns for material properties.
method Convolutional neural network with Attention mechanism to analyze deep features.
result Selected lattice constant predicts cathodic material cell voltage.
Applying neural-networks on Question Answering has gained increasing popularity in recent years. In this paper, I implemented a model with Bi-directional attention flow layer, connected with a Multi-layer LSTM encoder, connected with one start-index decoder and one conditioning end-index decoder. I introduce a new end-…
This paper analyzes deep and wide transformer training dynamics.
problem Understanding the training dynamics of infinitely deep and wide transformers.
method Develops a mean-field framework for gradient-based training of transformers, controlling a neural PDE.
result Establishes a rigorous foundation for gradient-based transformer training, proving convergence to global minima.
NeoMLP improves neural fields by adding self-attention for better downstream tasks.
problem Complexity and lack of conditioning mechanisms in neural fields.
method Transformed MLP into NeoMLP with self-attention and weight-sharing.
result NeoMLP outperforms state-of-the-art methods in downstream tasks.
Model detects depression from transcribed interviews using affective language.
problem Detecting depression from transcribed clinical interviews.
method Hierarchical Attention Network with affective conditioning.
result Model achieves state-of-the-art F1 scores in depression detection.
Structured State-Space Duality connects SSMs to masked attention.
problem Connecting SSMs and attention mechanisms for efficient modeling.
method Formalizing and generalizing SSD from scalar-identity to diagonal state matrices.
result Diagonal SSMs match training complexity lower bounds and support richer dynamics.
HKT improves sequence processing with multi-scale attention and kernel analysis.
problem Processing sequences at multiple scales with efficient attention mechanisms.
method Trainable causal downsampling and convex weights for level-specific score matrices.
result HKT achieves consistent gains over standard attention across various tasks.
Attention-only transformers learn from context via two stages of inference.
problem Learning from corrupted token sequences in minimal transformers.
method Two-stage empirical Bayes interpretation: kernel-weighted posterior mean and particle dynamics.
result Effective denoising without explicit noise schedules, showing posterior-mean recovery under asymptotic conditions.
Study predicts stock transaction durations using LSTM and attention mechanism.
problem Estimating the probability density function of transaction durations in financial markets.
method Proposes a hybrid model combining LSTM networks and attention mechanism to extend ACD model.
result Demonstrates superior performance of the hybrid model on large-scale financial data.
Researchers analyze neural process architectures and their representational capacities.
problem Understanding what functions can be represented by different neural process architectures.
method Analyzing four types of neural process architectures: CNPs, ANPs, TNPs, and their latent variants.
result Prove these architectures form a strict hierarchy and characterize their representational capabilities.
Neural processes (NPs) learn stochastic processes and predict the distribution of target output adaptively conditioned on a context set of observed input-output pairs. Furthermore, Attentive Neural Process (ANP) improved the prediction accuracy of NPs by incorporating attention mechanism among contexts and targets. In …
Paper analyzes infinite-width attention layers using Tensor Programs.
problem Capturing the infinite-width limit of attention layers.
method Tensor Programs framework to rigorously identify the limit distribution.
result Derives exact form of infinite-width limit distribution without Gaussian approximations.
We demystify attention patterns in multi-head softmax models for linear data.
problem Understanding the training dynamics and emergent patterns in multi-head softmax attention models.
method Extensive empirical experiments and rigorous theoretical analysis.
result Multi-head softmax attention models approximate a debiased gradient descent predictor, outperforming single-head attention and achieving near-Bayesian optimality.
Neural Processes (NPs) (Garnelo et al 2018a;b) approach regression by learning to map a context set of observed input-output pairs to a distribution over regression functions. Each function models the distribution of the output given an input, conditioned on the context. NPs have the benefit of fitting observed data ef…
Studying a softmax-attention model, we show that the learned query converges to the latent signal subspace spanned by the informative direction.
problem Understanding the theoretical principles of attention mechanisms in large-scale token collections.
method Deriving a population objective and analyzing the limiting ordinary differential equation of the learning dynamics.
result The learned query asymptotically recovers the latent signal up to the intrinsic sign ambiguity.
CutMix training technique improves spatial locality in Vision Transformers.
problem Improving spatial locality in Vision Transformers trained from scratch.
method Comparison of Baseline and Modern training protocols on CIFAR-10, CIFAR-100, and Tiny-ImageNet.
result CutMix training component significantly reduces Mean Attention Distance (MAD) in early layers of Vision Transformers.
Sparse Transformers can approximate dense Transformers with only O(n) connections.
problem Can sparse Transformers approximate arbitrary sequence-to-sequence functions?
method Proposed sufficient conditions for universal approximation and proved that sparse Transformers with O(n) connections can approximate dense models.
result Sparse Transformers with O(n) connections can approximate the same function class as dense models with n^2 connections.
Uniform scaling limits in AdamW-trained transformers converge to ODEs.
problem Understanding the dynamics of large-depth transformers trained with AdamW.
method Modeling transformer dynamics as an interacting particle system coupled through attention, proving convergence to ODEs.
result The joint dynamics of hidden states and backpropagated variables converge uniformly to an ODE system.
Develops a framework for identifying mispriced assets through attention factors for statistical arbitrage.
problem Identifying mispriced assets in statistical arbitrage trading.
method Uses conditional latent factors learned from firm characteristic embeddings to identify time-series signals and form a trading strategy.
result Achieves an out-of-sample Sharpe ratio above 4 on the largest U.S. equities over a 24-year period.
Multistep traffic forecasting on road networks is a crucial task in successful intelligent transportation system applications. To capture the complex non-stationary temporal dynamics and spatial dependency in multistep traffic-condition prediction, we propose a novel deep learning framework named attention graph convol…