New method prevents RLHF alignment collapse by accounting for policy's influence on reward model updates.
problem Iterative RLHF leads to alignment collapse where policies exploit RM's blind spots.
method Foresighted policy optimization (FPO) restores missing steering term via regularization.
result FPO prevents alignment collapse on LLM alignment pipelines using Llama-3.2-1B.
This paper extends neural collapse to regression problems, revealing key features and structures.
problem Understanding the structure learned by deep neural networks in regression tasks.
method Established Neural Regression Collapse (NRC) across different models, analyzing feature and weight alignments.
result Deep neural regression models exhibit a collapsed feature space, aligning with target dimensions and covariances.
Improved RL training for DMs reduces mode collapse and preserves diversity.
problem Mode collapse and training instability in RL fine-tuned diffusion models.
method Dynamic hierarchical RL training with sliding-window parameter regularisation.
result Models trained with HRF achieve better preservation of diversity in downstream tasks.
REPAIR mitigates variance collapse to enable linear interpolation between SGD solutions.
problem Linear interpolation between SGD solutions is difficult due to variance collapse in permuted activations.
method REPAIR rescales preactivations of interpolated networks to mitigate variance collapse.
result 60%-100% relative barrier reduction across various architectures and tasks.
Deep learning models show bias and variance are aligned, not in trade-off.
problem The classical bias-variance trade-off in deep learning models.
method Empirical evidence and theoretical analysis of bias and variance in deep learning models.
result Squared bias is approximately equal to variance for correctly classified sample points in deep learning models.
Theoretical work on mode collapse in variational inference models.
problem Mode collapse in variational inference models, where models focus on a few modes instead of all possible ones.
method Theoretical investigation of mode collapse in Gaussian mixture models, identifying key low-dimensional statistics and equations governing their evolution.
result Mode collapse is present even in favorable scenarios, driven by mean alignment and vanishing weight mechanisms.
New RLHF approach mitigates bias in aligning LLMs with human preferences.
problem Algorithmic bias in RLHF leading to preference collapse.
method Preference Matching (PM) RLHF, using PM regularizer and conditional variant.
result 29% to 41% improvement in alignment with human preferences.
Reward collapse occurs when ranking-based reward models yield uniform rewards for different prompts.
problem Reward collapse in aligning large language models with human preferences.
method Introduced a prompt-aware optimization scheme to derive closed-form expressions for reward distributions.
result Our prompt-aware utility functions significantly alleviate reward collapse during training.
We analyze deep neural networks using convex duality to reveal hidden layer structures.
problem Understanding the structure of deep neural networks.
method Introducing a convex analytic framework to characterize hidden layer weights.
result Optimal hidden layer weights align with previous layers via duality.
Study shows 'Ordinal Neural Collapse' in deep OR tasks, revealing simple geometric relationships.
problem Understanding neural collapse in deep Ordinal Regression tasks.
method Combining cumulative link models and Unconstrained Feature Model to investigate neural collapse.
result Demonstrates 'Ordinal Neural Collapse' (ONC) with three key properties.
Graph matching consists of aligning the vertices of two unlabeled graphs in order to maximize the shared structure across networks; when the graphs are unipartite, this is commonly formulated as minimizing their edge disagreements. In this paper, we address the common setting in which one of the graphs to match is a bi…
This paper studies aligning knowledge graphs from different sources or languages. Most existing methods train supervised methods for the alignment, which usually require a large number of aligned knowledge triplets. However, such a large number of aligned knowledge triplets may not be available or are expensive to obta…
Feedback alignment methods need to be evaluated for accuracy and gradient cosine similarity.
problem Evaluating feedback alignment methods
method Proposed diagnostic evaluation protocol
result Identified silent failures in standard reporting pair
Deep Gaussian Processes with polynomial kernels can collapse rapidly without proper hyperparameter tuning.
problem The collapse of Deep Gaussian Processes with polynomial kernels without careful hyperparameter tuning.
method Analysis using the Berry-Esseen Theorem and observation of prior behavior.
result The prior of a Deep Gaussian Process collapses rapidly towards zero or places negligible mass on low norm functions without proper hyperparameter tuning.
Model collapse occurs even with minimal synthetic data, and larger models can exacerbate it.
problem Model collapse in large neural networks due to synthetic data.
method Theoretical and empirical investigation of model collapse in various neural network architectures.
result Model collapse is exacerbated by larger models, but not entirely prevented.
Wide neural networks with weight decay exhibit neural collapse.
problem Proving neural collapse in wide neural networks trained with weight decay.
method Generic guarantees on neural collapse for wide networks with weight decay, proving low training error and balancedness, and bounded conditioning.
result First proof of neural collapse in end-to-end training of wide neural networks with weight decay.
In narrow asymptotic settings Gaussian VAE models of continuous data have been shown to possess global optima aligned with ground-truth distributions. Even so, it is well known that poor solutions whereby the latent posterior collapses to an uninformative prior are sometimes obtained in practice. However, contrary to c…
Generative diffusion models gradually memorize training data, losing independent dimensions.
problem Understanding how generative diffusion models memorize training data, especially on low-dimensional manifolds.
method Measuring latent dimensionality via the learned score field, proposing a geometric memorization theory.
result Generative diffusion models experience a smooth collapse of their capacity to vary across independent directions as data become scarce, leading to near point-wise replication of salient features.
KL annealing helps VAEs avoid posterior collapse and overfitting.
problem Posterior collapse and overfitting in VAEs.
method Theoretical analysis of learning dynamics with KL annealing.
result Posterior collapse is inevitable when β exceeds a threshold. Neural collapse occurs in normalized features over a Riemannian manifold.
problem Understanding neural collapse in normalized feature models.
method Simplified multi-class classification task to a nonconvex optimization problem over the Riemannian manifold, analyzing the landscape of critical points.
result The only global minimizers are neural collapse solutions, with all other critical points being strict saddles.
The paper addresses poor calibration in fine-tuned LLMs after preference alignment.
problem Poor calibration in fine-tuned Large Language Models (LLMs) after preference alignment.
method Proposes a calibration-aware fine-tuning approach to restore calibration without compromising model performance.
result Demonstrates the effectiveness of the proposed methods through extensive experiments.
Paper explores limits and possibilities of aligning LLMs with human preferences.
problem Aligning LLMs with diverse human preferences to ensure fairness and informed outcomes.
method Analysis of probabilistic representation of human preferences and preservation of diverse preferences.
result LLMs can't fully align with human preferences using reward-based approaches due to Condorcet cycles, but mixed strategies are statistically possible.
Benchmarking recursive collapse claims with a new framework under false-positive control.
problem Evaluating recursive systems for failure patterns and warning claims.
method Developed Loopzero framework for testing recursive failures, specified claim boundaries in Lean, evaluated under FP constraint, and compared with standard detectors.
result No standard detectors or Loopzero's pre-registered quantile detector achieved the required operating point under the false-positive contract.
NPO method improves LLM unlearning without catastrophic collapse.
problem Efficiently unlearning undesirable data from LLMs without losing model utility.
method Negative Preference Optimization (NPO) method based on alignment.
result NPO-based methods achieve better unlearning results and maintain model utility.
This paper extends neural collapse to class-imbalanced datasets using an unconstrained ReLU feature model.
problem Understanding neural collapse in class-imbalanced datasets with cross-entropy loss.
method Generalized neural collapse to class-imbalanced settings using an unconstrained ReLU feature model.
result Class-means converge to orthogonal vectors with different lengths, and classifier weights align to these vectors.
This paper investigates how large language models achieve neural collapse, a phenomenon linked to generalization.
problem Neural collapse in large language models under imbalanced and token-rich conditions.
method Empirical investigation of scaling and regularization effects on CLMs' progression towards neural collapse.
result Neural collapse properties develop with scale and regularization, linked to generalization in language modeling.
New framework formalizes RLHF trilemma: improving safety, fairness, and robustness is computationally infeasible.
problem Aligning large language models with diverse human values while maintaining computational feasibility and robustness.
method Complexity-theoretic analysis integrating statistical learning theory and robust optimization.
result Achieving both representativeness (epsilon <= 0.01) and robustness (delta <= 0.001) for global-scale populations requires super-polynomial operations.
The paper proposes a new framework to generate synthetic data with human-like imperfections to prevent model collapse.
problem Model collapse due to statistical optimization of synthetic data.
method Introduces Prompt-driven Cognitive Computing Framework (PMCSF) with Cognitive State Decoder (CSD) and Cognitive Text Encoder (CTE).
result The framework generates text with cognitive imperfections, reducing maximum drawdown and delivering defensive alpha.
Classifies surfaces supporting alignable nets with geodesic and conjugate properties.
problem Classifying surfaces with specific geometric properties.
method Cartan's theory of moving frames, coordinate-free classification, explicit immersion formulas.
result Two classes of alignable Voss surfaces, each with two two-parameter families, including one with an isothermal-conjugate geodesic net.
Graph neural networks (GNNs) have become increasingly popular for classification tasks on graph-structured data. Yet, the interplay between graph topology and feature evolution in GNNs is not well understood. In this paper, we focus on node-wise classification, illustrated with community detection on stochastic block m…
New metric measures dynamical richness without relying on accuracy.
problem Lack of a reliable metric for measuring dynamical richness.
method Developed a computationally efficient, performance-independent metric based on low-rank bias.
result Metric recovers neural collapse as a special case and captures known transitions without accuracy.
SAGE enhances reinforcement learning by injecting hints to prevent model stagnation.
problem Sparse rewards cause large language models to stall under relative policy optimization.
method SAGE injects privileged hints during training to increase within-group outcome diversity.
result SAGE consistently outperforms GRPO on 6 benchmarks with LLMs, achieving significant improvements.
Unified framework for self-supervised learning via latent distribution matching.
problem Lack of a unifying theoretical framework for diverse SSL methods.
method Casting SSL as latent distribution matching (LDM): maximizing alignment and uniformity.
result Derives a Bayesian filtering model and proves identifiable latent representations.
The paper shows that causal identification is not essential for efficient portfolios, focusing on geometric sufficiency conditions.
problem The necessity of causal identification for efficient portfolios.
method Re-examination of predictive signals and their impact on portfolio efficiency under structural misspecification.
result Efficiency is governed by geometric sufficiency conditions (directional alignment, ranking preservation, and calibration) rather than causal identification.
Proposes a new RL method to fine-tune flow-based models with arbitrary rewards.
problem Challenges in fine-tuning continuous flow-based generative models with arbitrary reward functions.
method Online Reward-Weighted Conditional Flow Matching with Wasserstein-2 Regularization (ORW-CFM-W2)
result Achieves optimal policy convergence with controllable trade-offs between reward maximization and diversity preservation.
Convolutional neural networks provide visual features that perform remarkably well in many computer vision applications. However, training these networks requires significant amounts of supervision. This paper introduces a generic framework to train deep networks, end-to-end, with no supervision. We propose to fix a se…
Improved co-clustering for robust data analysis.
problem Noise and posterior collapse in co-clustering.
method Fully variational Co-clustering framework with Compositional ELBO.
result Enhanced robustness and accuracy in co-clustering.
POTNet uses penalized optimal transport to generate data without mode collapse.
problem Mode collapse in WGANs leading to poor synthetic data generation.
method POTNet employs marginally-penalized Wasserstein distance for deep generative modeling.
result POTNet effectively captures underlying data structures, including tail behaviors and minor modalities.
Bayesian deep learning faces posterior collapse due to likelihood vs. prior competition.
problem Posterior collapse in Bayesian deep learning models.
method Identified competition between likelihood and prior regularization in a linear latent variable model.
result Posterior collapse is related to neural and dimensional collapse, suggesting a broader learning issue.
Introduces PCG for better counterfactual explanations in vision models.
problem Ambiguity in latent-space optimization methods for counterfactual explanations.
method Constructs counterfactuals by tracing geodesics under a perceptually Riemannian metric.
result PCG outperforms baselines and reveals hidden failure modes.
Proves weakly non-collapsed RCD spaces are strongly non-collapsed.
problem Proving the equivalence of weakly non-collapsed and strongly non-collapsed RCD spaces.
method Analyzes properties of RCD spaces and uses auxiliary results.
result Confirms conjecture about RCD spaces being strongly non-collapsed.
Study on Neural Collapse limits in deep learning.
problem Understanding the limits of Neural Collapse in deep learning.
method Investigated Neural Collapse in the context of generalization and feature learning, refining conjectures and conducting experiments.
result Neural Collapse primarily occurs on the train set and not on the test set, suggesting it is an optimization phenomenon with unclear connections to generalization.
SFAG generates realistic financial data that passes trading tests.
problem Financial generative models often produce unrealistic and unstable trading outcomes.
method Introduces SFAG, a GAN variant that aligns stylized facts and optimizes with adversarial loss.
result SFAG generates synthetic data that preserves stylized facts and supports robust trading strategies.
NE-GMM uses ES and GMM to improve uncertainty quantification.
problem Challenges in estimating mean and variance of complex distributions.
method Integrates Gaussian Mixture Model with Energy Score.
result NE-GMM outperforms in predictive accuracy and uncertainty quantification.
Ricci flow smooths locally collapsing manifolds with controlled curvature.
problem Locally collapsing manifolds with controlled Ricci curvature.
method Ricci flow for a definite period of time, detecting collapsing infranil fiber bundles.
result Topological conditions detect collapsing infranil fiber bundles.
Mathematical analysis shows annealing prevents mode collapse in Gaussian mixtures.
problem Mode collapse in variational inference for multimodal distributions.
method Analyzed annealing strategies for Gaussian mixtures, derived formulas, and tested on neural networks.
result Appropriately chosen annealing schemes can robustly prevent mode collapse.
The study characterizes and rules out collapsing in convex ancient mean curvature flow.
problem Characterizing and ruling out collapsing in convex ancient mean curvature flow.
method Characterization and counterexamples.
result Collapsing occurs if and only if the flow is asymptotic to at least one Grim hyperplane.
Despite excellent progress in recent years, mode collapse remains a major unsolved problem in generative adversarial networks (GANs).In this paper, we present spectral regularization for GANs (SR-GANs), a new and robust method for combating the mode collapse problem in GANs. Theoretical analysis shows that the optimal …