Hölder-DPO aligns models robustly with noisy human feedback.
problem No existing alignment methods can handle severe label noise.
method Proposes Hölder-DPO, a principled alignment loss with provable redescending property.
result Hölder-DPO enables scalable human feedback valuation and improves model alignment.
MKA incorporates manifold geometry into kernel alignment for more robust representation comparison.
problem Inadequate accounting for manifold geometry in kernel alignment metrics.
method Derives a theoretical framework for Manifold Approximated Kernel Alignment (MKA).
result MKA provides a more robust foundation for measuring representations.
AI models aligned with human vision perform well on few data tasks.
problem Few-shot learning performance with limited data.
method Information-theoretic analysis and empirical testing of 491 models.
result Highly aligned models show better robustness to attacks and domain shifts.
Paper addresses online alignment of large language models under uncertain preference feedback.
problem Online alignment of large language models with misspecified preference feedback.
method Formulates an oracle-robust objective as a worst-case optimization problem for log-linear policies, and develops projected stochastic composite updates.
result Shows that the robust objective admits an exact closed-form decomposition and achieves O ~ ( ε − 2 ) \widetilde{O}(\varepsilon^{-2}) O ( ε − 2 ) oracle complexity. For a standard convolutional neural network, optimizing over the input pixels to maximize the score of some target class will generally produce a grainy-looking version of the original image. However, Santurkar et al. (2019) demonstrated that for adversarially-trained neural networks, this optimization produces images …
Proposes a robust algorithm for aligning large language models with human preferences.
problem Misspecification in preference models, reference policies, and reward functions.
method Doubly robust preference optimization algorithm.
result Superior and more robust performance compared to state-of-the-art algorithms.
Robustly aligns datasets with partial GW distance to handle contamination.
problem Aligning contaminated datasets using Gromov-Wasserstein distances.
method Proposes a partial GW distance estimator to minimize distortion from outliers.
result The partial GW distance estimator is minimax optimal and near-optimal in finite samples.
Novel methods robustify Gromov-Wasserstein distance for cross-domain alignment.
problem Robustifying Gromov-Wasserstein distance for cross-domain alignment.
method Three novel techniques derived from robust statistics to improve GW and its variants.
result Empirical validation shows superior resilience to contamination.
DRO-REBEL improves LLM alignment by robustly updating models online.
problem Overfitting and drifting of LLMs during RLHF.
method DRO-REBEL uses type- p p p Wasserstein, KL, and χ 2 χ^2 χ 2 ambiguity sets for robust online updates. result DRO-REBEL achieves faster convergence and better performance than prior methods.
PGD-trained models have a preferential direction in their gradients, which improves robustness.
problem Mathematical lack of clarity in the direction of preferential gradient alignment after adversarial training.
method Proposed a novel definition of preferential direction and evaluated it using a metric based on GANs.
result PGD-trained models have higher alignment with the proposed preferential direction than baseline models.
Aligns attention distributions for improved accuracy and robustness.
problem Improving the accuracy and robustness of neural networks using attention mechanisms.
method Alignment attention that encourages key and query distributions to match within each head.
result Alignment attention leads to better accuracy, uncertainty estimation, and robustness across various tasks.
Unbalanced COOT improves feature alignment robustly to outliers.
problem Optimal transport methods are sensitive to outliers in real-world data.
method COOT infers alignment between features and samples, unbalanced COOT adds robustness.
result Unbalanced COOT is robust to noise in real-world datasets.
EBRM improves robustness and generalization of language model rewards.
problem Challenges in capturing complex human preferences and generalizing to unseen data in reward models.
method Energy-Based Reward Model (EBRM) that models reward distribution explicitly, using conflict-aware data filtering, label-noise-aware contrastive training, and hybrid initialization.
result Significant improvements in robustness and generalization, up to 5.97% improvement in safety-critical alignment tasks.
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.
New method for robust distribution alignment using log-likelihood ratio and normalizing flows.
problem Distribution alignment challenges in deep learning.
method Log-likelihood ratio statistic and normalizing flows.
result Minimizing the proposed objective yields robust domain alignment.
Survey of alignment techniques for large language models.
problem Ensuring large language models align with human values.
method Analysis of diverse alignment methods and training paradigms.
result Preference-based methods offer more flexibility for nuanced alignment.
Develops an ordinal-similarity framework for scalable and interpretable representation alignment.
problem Measuring representation similarity in large datasets.
method Triplet and Quadruplet Similarity Indices.
result Demonstrates inherent interpretability, robustness to outliers, and computational efficiency.
Traditional pairwise sequence alignment is based on matching individual samples from two sequences, under time monotonicity constraints. However, in many application settings matching subsequences (segments) instead of individual samples may bring in additional robustness to noise or local non-causal perturbations. Thi…
Proposes neuron alignment to optimize mode connectivity in neural networks.
problem Understanding and optimizing mode connectivity in deep neural networks.
method Introduces neuron alignment to approximate optimal weight permutations and improve mode connectivity.
result Neuron alignment significantly alleviates robust loss barriers and improves model robustness and accuracy.
This study presents the results of a series of simulation experiments that evaluate and compare four different manifold alignment methods under the influence of noise. The data was created by simulating the dynamics of two slightly different double pendulums in three-dimensional space. The method of semi-supervised fea…
SARA uses similarity to learn rewards robustly and adaptively.
problem Robustness to labeler errors and adaptability to diverse feedback formats.
method Contrastive framework that learns latent representations and computes rewards as similarities.
result Strong performance on offline RL benchmarks and diverse applications.
This work explores how neural network architecture affects robustness to noisy labels.
problem The impact of neural network architecture on robustness to noisy labels.
method Formal framework connecting robustness to architecture alignments, measured by predictive power in representations.
result Network robustness to noisy labels improves when its architecture is more aligned with the target function.
This paper improves GNN robustness by aligning feature and adjacency matrix learning.
problem Improving robustness of graph neural networks (GNN) in noisy graph data.
method Proposes a novel regularized GSL approach that aligns feature information and graph information, incorporating sparse dimensional reduction.
result Demonstrates superior performance in noisy graph structures compared to competitive baselines.
This work frames reward modelling from preferences as a causal problem.
problem Reward modelling from preference data for AI alignment.
method Causal inference approach to identify challenges and assumptions.
result Causally-inspired approaches improve model robustness.
GrokAlign aligns Jacobians to accelerate grokking in deep networks.
problem Accelerating the training dynamics of deep networks to avoid delayed generalisation and robustness.
method Aligning the Jacobians of a deep network with the training data to ensure grokking under a low-rank assumption.
result GrokAlign regularizes Jacobians to induce grokking sooner than conventional methods.
Robust GW distance improves graph data alignment.
problem Outliers in GW distance lead to inaccurate comparisons.
method Optimistically perturbed marginal constraints within a Kullback-Leibler divergence-based ambiguity set.
result RGW reduces inaccuracies in graph data alignment.
Recent studies on the adversarial vulnerability of neural networks have shown that models trained to be more robust to adversarial attacks exhibit more interpretable saliency maps than their non-robust counterparts. We aim to quantify this behavior by considering the alignment between input image and saliency map. We h…
Sequence-to-sequence text-to-speech (TTS) is dominated by soft-attention-based methods. Recently, hard-attention-based methods have been proposed to prevent fatal alignment errors, but their sampling method of discrete alignment is poorly investigated. This research investigates various combinations of sampling methods…
New findings link causal models to strategic classification, improving robustness and alignment.
problem Strategic adaptation by users in classification tasks.
method Causal models to bound worst-case out-of-distribution risk.
result Causal classification optimizes classification error after adaptation under certain noise conditions.
Paper proposes MDAT to stabilize domain alignment in label-scarce settings.
problem Stable and comprehensive domain alignment in label-scarce settings.
method Max-margin Domain-Adversarial Training (MDAT) with Adversarial Reconstruction Network (ARN).
result MDAT stabilizes gradient reversing and achieves strong robustness to hyper-parameters.
High-dimensional neural network manifolds misalign with human perception, causing adversarial examples.
problem Adversarial attacks fool neural networks, but their origin is unclear.
method Defined and analyzed a network's perceptual manifold (PM) for a class concept.
result Neural network PMs have orders of magnitude higher dimensions than natural human concepts, suggesting exponential misalignment.
We introduce a kernel method for manifold alignment (KEMA) and domain adaptation that can match an arbitrary number of data sources without needing corresponding pairs, just few labeled examples in all domains. KEMA has interesting properties: 1) it generalizes other manifold alignment methods, 2) it can align manifold…
RESTA defends LLMs against jailbreaking attacks by adding random noise to embeddings.
problem Vulnerability of LLMs to jailbreaking attacks that generate harmful outputs.
method Adds random noise to embedding vectors and aggregates during token generation.
result RESTA achieves superior robustness versus utility tradeoffs compared to baseline defenses.
Proposes DFDG for robust domain generalization without source domain labels.
problem Robustness of deep learning models in real-world applications where train and test distributions differ.
method Model-agnostic, class-aware alignment of class relationships through saliency maps.
result Competitive performance on time series sensor and image classification datasets.
Improves domain adaptation by aligning source and target distributions and mitigating noisy labels.
problem Improving performance on target images with different acquisition conditions.
method Combines optimal transport, MixUp regularization, and robust loss for noisy labels.
result Improves domain adaptation performance on various benchmarks and real-world problems.
HSACC improves multi-view clustering of incomplete data.
problem Challenges in clustering incomplete multi-view data.
method Hierarchical Semantic Alignment and Cooperative Completion framework.
result HSACC outperforms state-of-the-art methods on benchmark datasets.
Gradient-trained shallow networks can generalize well but are vulnerable to small-radius adversarial attacks.
problem Adversarial robustness of gradient-trained shallow networks.
method Analysis of neuron alignment and polynomial ReLU activation.
result Gradient-trained shallow networks with polynomial ReLU activation are robust to small-radius adversarial attacks.
New method defends deep nets against large perturbations perceptible to humans.
problem Vulnerability of deep nets to adversarial attacks with perceptible but not changing predictions.
method Oracle-Aligned Adversarial Training (OA-AT) to align network predictions with Oracle's.
result Achieves state-of-the-art performance at large perturbation bounds (L-inf of 16/255 on CIFAR-10).
The paper explores game-theoretic alignment of LLMs with human preferences, finding limitations and conditions.
problem Aligning LLMs with human preferences using game theory.
method Systematic study of payoff choices in a two-player zero-sum game for desirable alignment properties.
result Impossibility of preference matching in game-theoretic LLM alignment under standard assumptions.
SONA improves conditional generation by balancing authenticity and alignment.
problem Challenges in balancing authenticity and conditional alignment in conditional generative models.
method SONA integrates unconditional discrimination, matching-aware supervision, and adaptive weighting to balance authenticity and alignment.
result SONA achieves superior sample quality and conditional alignment compared to state-of-the-art methods.
Improved graph matching algorithm robust to noise.
problem Finding a bijection between vertex sets of two graphs.
method Uses multistage signature vectors to match vertices.
result Recover matching exactly with high probability for α ≤ 1 / ( log log n ) C α\le 1 / (\log \log n)^C α ≤ 1/ ( log log n ) C . OOM-RL uses financial market losses to align AI agents in autonomous systems.
problem Constrained alignment of autonomous software agents in live financial markets.
method Deploying agents in live financial markets to enforce strict test-driven workflows.
result Final OOM-RL-aligned system achieved a stable equilibrium with an annualized Sharpe ratio of 2.06.
RePULSe improves language model alignment by reducing undesired outputs without sacrificing overall performance.
problem Aligning language models with human preferences while minimizing undesired outputs.
method Integrates probabilistic inference into RL training to reduce undesired outputs.
result RePULSe achieves a better balance between expected reward and undesired output probability.
Unified framework simplifies DPO algorithms for LLM alignment.
problem Vast number of DPO variants complicates model alignment.
method Mutual information inspired unifying framework with flexible priors.
result Many DPO variants can be derived from the new framework.
Develops methods to correct bias in AI feedback for more accurate alignment.
problem Systematic bias in AI feedback compared to human labels.
method Two debiased alignment methods: DDPO and DIPO.
result Methods improve alignment efficiency and performance close to human-labeled data.
Enhances deep learning robustness to noise without sacrificing clean data accuracy.
problem Robustness of deep neural networks to input noise.
method Discriminative loss at penultimate layer and class-wise feature alignment with Gaussian noise.
result Improves robustness to various perturbations without degrading clean data accuracy.
Proposes a new method for manifold alignment using geometry-regularized twin autoencoders.
problem Traditional MA methods lack out-of-sample extension and real-world applicability.
method Guided representation learning with geometry-regularized twin autoencoders.
result Improves cross-domain generalization and robustness while maintaining alignment fidelity.
UDA improves ABI robustness but fails under certain prior misspecifications.
problem Robustness of ABI in noisy real-world data.
method Systematic evaluation of UDA across various misspecification scenarios.
result UDA aligns summary spaces but can fail under prior misspecifications.