Study shows effectiveness of offline RL in online RL tasks.
problem Improving online RL efficiency using offline RL data.
method Formalized framework for incorporating offline RL as online RL subroutines, introducing techniques to enhance effectiveness.
result Effectiveness of the framework depends on task nature, techniques greatly enhance effectiveness, and existing methods are ineffective.
This work improves online fine-tuning of diffusion models for specific properties.
problem Efficiently fine-tuning diffusion models to maximize specific properties.
method A novel reinforcement learning procedure that efficiently explores feasible samples.
result The method provides a regret guarantee and empirical validation across multiple domains.
Paper proposes a hybrid RL algorithm that combines offline and online data without needing reward info.
problem How to efficiently use online data to improve RL policies using only offline data.
method A three-stage hybrid RL algorithm that uses reward-agnostic exploration and model-based offline RL.
result The hybrid RL algorithm outperforms both pure offline and pure online RL in sample complexity.
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.
BERT fine-tuning is unstable due to optimization issues, not forgetting or dataset size.
problem Stability of fine-tuning BERT-based models across different random seeds.
method Analysis of BERT, RoBERTa, and ALBERT fine-tuned on GLUE datasets, identifying optimization difficulties as the cause of instability.
result Fine-tuning instability is due to optimization difficulties leading to vanishing gradients, not forgetting or dataset size.
Fine-tuning RL with offline data reduces online interactions.
problem Optimizing RL with limited online interactions and offline data.
method Developed algorithm extsc{FTPedel} for MDPs with linear structure.
result Optimally reduces the number of online interactions needed.
We provide an online RLHF workflow for large language models.
problem Lack of online RLHF workflow for large language models.
method Construct preference models using open-source datasets, approximate human feedback, and implement online iterative RLHF.
result Our LLM achieves state-of-the-art performance on various benchmarks.
Chronicals boosts LLM fine-tuning speed by 3.51x over Unsloth.
problem Memory bottleneck in fine-tuning large language models.
method Four optimizations: fused kernels, cross-entropy reduction, LoRA+, and sequence packing.
result 3.51x speedup on A100-40GB with Qwen2.5-0.5B.
LoRA fine-tuning explained with gradient dynamics for low-rank perturbations.
problem Understanding why gradient descent converges to useful low-rank perturbations in LoRA fine-tuning.
method Generalized student-teacher setting with i.i.d. samples and online gradient descent.
result Gradient descent converges to the teacher model in dkO(1) iterations under certain conditions. Improved RL policies from offline data with relaxed BC constraints.
problem Overestimation bias in offline RL due to lack of interaction with environment.
method Introducing a policy constraint via behavioural cloning (BC) and adjusting the balance between RL and BC.
result Refined policies outperform baseline and match/exceed complex alternatives.
The paper introduces a method to learn and apply value envelopes for faster online reinforcement learning.
problem Accelerating online reinforcement learning using offline data with theoretical grounding.
method A two-stage framework: offline data for learning value bounds, online algorithms for applying them.
result Substantial regret reductions in empirical tests on tabular MDPs.
A new policy switching technique improves offline RL performance.
problem Challenges in adapting off-policy algorithms to different datasets and tasks.
method Combines off-policy RL and BC, using epistemic uncertainty for policy switching.
result Outperforms individual algorithms and state-of-the-art methods on benchmarks.
GALA adapts learning rates online by aligning gradients, improving deep learning model performance.
problem Fine-tuning learning rates for deep learning models requires extensive grid search.
method GALA dynamically adjusts learning rates by tracking gradient alignment and local curvature.
result GALA produces a flexible, adaptive learning rate schedule that increases when gradients align.
Paper addresses inconsistency between offline and online LTR performance.
problem Inconsistency between offline and online LTR performance in E-commerce.
method Proposes an evaluator-generator framework to maximize evaluator score using reinforcement learning.
result Significant improvement in Conversion Rate (CR) over existing models.
Paper resolves bias in ALFT training using generalized alignment games.
problem Systematic bias in estimating logarithmic rewards from small batches.
method Generalized Distributional Alignment Games, U-statistics, minimax polynomial estimators, Variance-Optimal Augmented Polynomial Optimization Program (AQP) Estimator.
result Proves optimal bias and accelerated convergence in ALFT training.
New methods improve LLM preference optimization by intelligently weighting multiple reference models.
problem Improving LLM preference optimization with multiple reference models.
method Introducing four new weighting strategies for multiple-reference preference optimization.
result All four new weighting strategies outperform current methods on preference accuracy.
LOLA uses LLMs to optimize content delivery, outperforming traditional methods.
problem Identifying the most engaging headlines for user engagement.
method LOLA integrates LLMs with adaptive experimentation to optimize content delivery.
result LOLA outperforms traditional methods in optimizing user engagement.
New approach uses inverse reinforcement learning to improve language model training.
problem Training large language models using imitation learning methods.
method Developed a new method of inverse reinforcement learning to optimize sequences directly.
result IRL-based fine-tuning leads to better performance and diversity in language generation.
New method learns to answer questions from correct demonstrations without assuming bounded complexity of the demonstrator.
problem Learning to generate answers from correct demonstrations with multiple correct answers.
method Formalizes imitation learning in contextual bandits, relying on reward model complexity, not policy complexity.
result Achieves nearly optimal performance with sample complexity logarithmic in reward class cardinality.
SUPE combines unlabeled data with RL to efficiently explore tasks.
problem Efficient exploration in reinforcement learning with sparse rewards.
method Extract low-level skills using VAE, pseudo-label unlabeled data, and use as off-policy data for online RL.
result SUPE outperforms prior methods across 42 long-horizon tasks.
New measure FTC quantifies how much a ReLU network can fine-tune.
problem Analyzing memorization capacity in fine-tuned neural networks.
method Defined Fine-Tuning Capacity (FTC) for additive fine-tuning of ReLU networks.
result Upper and lower bounds on FTC for 2 and 3-layer ReLU networks.
AWAC combines offline and online data to accelerate RL learning.
problem Challenges in applying RL to real-world robotic control due to exploration and sample complexity.
method Combines sample-efficient dynamic programming with maximum likelihood policy updates.
result AWAC enables rapid learning of robotic skills with prior data and online experience.
This paper re-evaluates hyperparameters for fine-tuning pre-trained models.
problem Current hyperparameter settings for fine-tuning are often ad-hoc and fixed.
method Empirical evaluation of learning rate, batch size, and momentum for fine-tuning.
result Optimal hyperparameters are not only dataset-dependent but also sensitive to domain similarity.
New method quantifies uncertainty in fine-tuned LLMs using LoRA ensembles.
problem Uncertainty in fine-tuned LLMs and how to trust their predictions.
method Posterior approximations using low-rank adaptation ensembles.
result Unexpected retention of acquired knowledge during fine-tuning in overfitting regime.
FinDPO uses preference optimization to improve financial sentiment analysis models.
problem Financial sentiment analysis models often fail to generalize to unseen data.
method FinDPO uses Direct Preference Optimization (DPO) to align LLMs with human preferences.
result FinDPO achieves state-of-the-art performance and maintains positive returns under realistic trading conditions.
Improved code translation by preserving structure with composed fine-tuning.
problem Improving code translation accuracy with unlabeled code outputs.
method Pre-trained denoiser to capture output structure, composed fine-tuning to fine-tune predictor.
result Composed fine-tuning significantly improves generalization over standard fine-tuning.
Estimates mode from partial feedback, improving AI learning pipelines.
problem Estimating the mode of a distribution with partial feedback.
method Entropy coding, coarse sufficient statistics, bandit algorithms.
result Statistically and computationally efficient solution to mode estimation.
The paper introduces a Hessian-based method to improve generalization in fine-tuned deep neural networks.
problem Improving generalization in fine-tuned deep neural networks, especially in noisy conditions.
method PAC-Bayesian analysis to identify a Hessian-based distance measure, proving generalization bounds, and developing an algorithm with a generalization error guarantee.
result Hessian-based distance measure correlates well with observed generalization gaps and can match the scale of these gaps in practice.
Fine-tuning LLMs improves capability but harms safety, study finds.
problem Balancing capability and safety in LLM fine-tuning.
method Theoretical framework and numerical experiments for two safety-aware fine-tuning strategies.
result Characterization of fundamental limits of safety-capability trade-off in LLM fine-tuning.
Optimizes sparse fine-tuning for privacy in neural networks.
problem Performance gap between DP-SGD and non-private fine-tuning.
method Optimization-based approach using private gradient information for selecting trainable weights.
result Our selection method leads to better prediction accuracy compared to existing approaches.
Fine-tuning harms in-context learning, but restricting updates to the value matrix improves zero-shot performance.
problem Fine-tuning harms in-context learning, reducing zero-shot performance on unseen tasks.
method Theoretical analysis of linear attention models, identifying conditions for degraded few-shot performance.
result Restricting updates to the value matrix improves zero-shot performance while preserving in-context learning.
Measures consistency of tabular LLM predictions under fine-tuning multiplicity.
problem Conflicting predictions from fine-tuned tabular LLMs.
method Local stability measure in embedding space.
result Probabilistic guarantees on prediction consistency under multiplicity.
Self-play fine-tuning improves diffusion models for text-to-image generation.
problem Plateauing performance of diffusion models after data saturation.
method Self-play fine-tuning (SPIN-Diffusion) using competition among model versions.
result Significantly improved model performance and human preference alignment.
The paper develops a theory linking pretraining and fine-tuning in neural networks.
problem Understanding how initialization choices impact feature learning and generalization in neural networks.
method Analytical theory of diagonal linear networks, deriving generalization error as a function of initialization parameters and task statistics.
result Different initialization choices place networks into four fine-tuning regimes with varying abilities to support feature learning and generalization.
New method reduces fine-tuning cost for reused models.
problem Repeating fine-tuning costs with outdated foundation models.
method Portable Reward Tuning (PRT) trains a reward model to maximize the same loss function as fine-tuning.
result PRT achieves comparable accuracy to inference-time tuning with less inference cost.
Compact models match or exceed GPT's performance in financial news sentiment analysis.
problem Improving financial sentiment analysis models without large computational costs.
method Fine-tuning non-generative, small-sized models (FinBERT, FinDRoBERTa) on a novel market score database.
result Fine-tuned models outperform GPT-3.5 and GPT-4 in zero-shot learning for financial news sentiment analysis.
Unsupervised pre-training improves model generalization, but lacks theoretical understanding.
problem Lack of theoretical understanding of unsupervised pre-training's impact on model generalization.
method Introduces a novel theoretical framework to analyze and enhance generalization.
result Enhances understanding of unsupervised pre-training and fine-tuning, proposing a new regularization method.
LP-FT improves personalized model training in FL by balancing generalization and personalization.
problem Federated Learning struggles with balancing global generalization and local personalization due to non-identical data distributions.
method Adapting Linear Probing followed by full Fine-Tuning (LP-FT) to the FL setting.
result LP-FT outperforms standard fine-tuning in balancing personalization and generalization across various datasets and PFT variants.
UBM transfers bias mitigation from upstream to downstream tasks efficiently.
problem Bias in fine-tuned language models across various tasks.
method Apply bias mitigation to an upstream model, then fine-tune a downstream model on this mitigated model.
result UBM effects transfer to new downstream tasks, creating less biased models.
Improved fine-tuning with regularization and robustness for noisy labels.
problem Fine-tuning pre-trained models on small datasets can lead to overfitting and memorization.
method PAC-Bayes generalization bound analysis, layer-wise regularization, self-label-correction, label-reweighting.
result Improves performance by 1.76% on average for image classification tasks and 0.75% for few-shot classification.
Simplified trust region method reduces representation change during fine-tuning.
problem Stability and representational collapse in fine-tuning pre-trained models.
method Replaces adversarial objectives with parametric noise in trust region theory.
result Matches or exceeds previous trust region methods in performance and speed.
Adjoint Matching improves flow and diffusion models with reward fine-tuning.
problem Improving generative models with reward fine-tuning.
method Casting reward fine-tuning as stochastic optimal control (SOC) and enforcing a specific noise schedule.
result Adjoint Matching outperforms existing SOC algorithms.
Fine-tunes LLMs to correct bias in predictions.
problem LLMs exhibit bias in predictions from data.
method Supervised fine-tuning with Low-Rank Adaptation (LoRA).
result Fine-tuning corrects bias in both controlled and real-world settings.
Transfer learning, which allows a source task to affect the inductive bias of the target task, is widely used in computer vision. The typical way of conducting transfer learning with deep neural networks is to fine-tune a model pre-trained on the source task using data from the target task. In this paper, we propose an…
Fine-tuning large language models requires minimal data, making them efficient.
problem Achieving state-of-the-art performance with large language models.
method Using BERT as an example, fine-tuning only the most critical layers of the pre-trained model.
result Fine-tuned models are close in parameter space to the pre-trained model, with many good solutions found in sparsified versions.
FinLlama uses a fine-tuned Llama 2 model for financial sentiment analysis.
problem Accurate financial sentiment analysis for better trading decisions.
method Fine-tuning Llama 2 7B model on financial sentiment data, using a generator-classifier scheme.
result FinLlama provides nuanced insights into financial news articles, enhancing portfolio management.
Fine-tuning improves meta-learning by leveraging shared representations.
problem Meta-learning's challenge in rapidly learning new tasks.
method Theoretical framework and risk bounds on gradient descent fine-tuning.
result Fine-tuning-based methods can provably leverage shared structure.
New method enhances model fine-tuning with minimal data.
problem Improving model performance on new tasks with limited data.
method Introducing α-LoRA, a reparameterization method for fine-tuning. result Enhanced generalization ability of fine-tuned models.