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.
The study uncovers latent capabilities of language models via causal representation learning.
problem Rigorous causal evaluations of language model capabilities are challenging due to confounding effects and computational costs.
method Proposes a causal representation learning framework to identify latent capability factors as causally interrelated after controlling for a common confounder (base model).
result Identifies a three-node linear causal structure explaining performance variations across 1500 models and six benchmarks.
ML Compass helps organizations choose AI models that balance utility, cost, and compliance.
problem Selecting AI models that meet user utility, deployment costs, and compliance requirements.
method Develops ML Compass, a framework for constrained optimization over a capability-cost frontier, using internal measures and empirical data.
result ML Compass produces deployment-aware recommendations that differ from capability-only rankings, clarifying trade-offs between capability, cost, and safety.
A new framework ensures model safety by retaining old model capabilities while improving new tasks.
problem Catastrophic forgetting in continual learning systems.
method Retention-centric framework with data-dependent constraints and an efficient constrained optimization algorithm.
result The method ensures that new models retain important capabilities of old models while improving new tasks.
Reward hacking exploits misspecified rewards, affecting agent capabilities and true performance.
problem Reward hacking in RL models exploiting reward misspecifications.
method Constructed four RL environments with misspecified rewards; analyzed agent capabilities and behavior.
result More capable agents exploit reward misspecifications, achieving higher proxy reward but lower true reward.
RLVR maintains safety while improving reasoning capabilities in LLMs.
problem Safety-capability tradeoff in fine-tuning LLMs.
method Reinforcement Learning with Verifiable Rewards (RLVR) and theoretical analysis.
result RLVR can enhance reasoning while maintaining safety guardrails.
Paper examines LLM capability benchmarks through construct validity, favoring nomological account.
problem Linking theoretical capabilities to empirical measurements in LLMs.
method Contrasts three frameworks: nomological, inferential, and causal.
result Nomological account provides best foundation for LLM research.
BIG-bench benchmarks language models, revealing their strengths and weaknesses.
problem Understanding and quantifying the capabilities of large language models.
method Developed the BIG-bench benchmark with 204 tasks from various domains, evaluated across multiple model sizes and types.
result Model performance improves with scale but remains poor, with breakthrough behaviors often involving multiple steps.
Much of the analysis of economic growth has focused on the study of aggregate output. Here, we deviate from this tradition and look instead at the structure of output embodied in the network connecting countries to the products that they export.We characterize this network using four structural features: the negative r…
New framework transfers latent knowledge from weak to strong models.
problem Aligning superhuman LLMs with human feedback.
method Transfer learning framework using refinement approach.
result Proves weak-to-strong generalization is possible.
LLMs struggle with arithmetic tasks unless they use high numerical precision.
problem Improving arithmetical reasoning capabilities of LLMs.
method Theoretical analysis and empirical experiments on numerical precision.
result LLMs require high numerical precision to efficiently handle arithmetic tasks.
New approach extracts AI model representations for steering and monitoring.
problem Understanding and steering AI models' internal knowledge representation.
method Feature learning advances for scalable linear concept representations.
result Concept representations enable steering and monitoring of AI models.
TIR expands LLM capabilities by enabling problem-solving strategies.
problem Lack of a principled theory explaining why LLMs with tools are more capable.
method Formal proof and Advantage Shaping Policy Optimization (ASPO) algorithm.
result TIR model decisively outperforms pure-text models on challenging benchmarks.
Predicts language model performance from public models without training.
problem Understanding how language model performance varies with scale.
method Observational approach building scaling laws from publicly available models.
result Smooth, sigmoidal behavior of emergent phenomena is predictable from small models.
LaTRO optimizes latent reasoning in LLMs without external reward.
problem Training LLMs to perform complex reasoning tasks.
method Formulates reasoning as latent distribution sampling and optimizes via variational approaches.
result LLMs improve reasoning and evaluation quality through self-improvement.
Transformers can scale both context and task, but MLPs can only scale task.
problem Understanding and scaling In-Context Learning in transformers.
method Simplified transformer architecture, feature map, and MLP combination.
result Simplified transformer can perform ICL and context-scaling but not task-scaling.
A new method learns complex dynamical systems from data efficiently.
problem Learning complex dynamical systems from large-scale data efficiently.
method Low-rank structured variational autoencoding framework for nonlinear Gaussian state-space models.
result Consistently demonstrates better predictive capabilities compared to other models.
Paper proposes a new financial fraud detection system using improved RF and GBM.
problem Detecting financial fraud with high accuracy and efficiency.
method Introduces GBM-SSRF, a novel model combining GBM and SSRF.
result GBM-SSRF model outperforms traditional methods in fraud detection.
Designing a deep neural network (DNN) with good generalization capability is a complex process especially when the weights are severely quantized. Model averaging is a promising approach for achieving the good generalization capability of DNNs, especially when the loss surface for training contains many sharp minima. W…
HeteroFL trains diverse clients with varying capabilities efficiently.
problem Training models on clients with different computational and communication capacities.
method Proposes HeteroFL framework to adaptively distribute subnetworks based on client capabilities.
result Adaptive subnetwork distribution leads to efficient computation and communication.
The DoD needs a robust process to evaluate AI/ML model performance and robustness.
problem AI/ML models are brittle and nonrobust, posing risks in national security.
method Reviews AI/ML development process and best practices for evaluation.
result Recommendations for DoD evaluators to ensure robust AI/ML capabilities.
New method predicts language model scaling risks with limited data.
problem Accurately predicting model capabilities and risks with limited data.
method Introduces a robust estimation framework using beta-binomial distribution and dynamic sampling strategy.
result More accurate predictions of rare risks and capabilities at lower computational cost.
This paper shows ARMs and EBMs are equivalent, revealing ARM lookahead capabilities.
problem Understanding the lookahead capabilities of next-token prediction models.
method Unified view of ARMs and EBMs, establishing a bijection and deriving equivalence.
result ARMs and EBMs are equivalent, revealing ARM lookahead capabilities.
Proposes a framework to identify and correct model-form errors in nonlinear systems.
problem Model-form errors in nonlinear dynamical systems due to unknown or approximated governing equations.
method Uses a hybrid approach combining machine learning and Bayesian filtering to estimate and correct model-form errors.
result Improves the predictive capability of known but approximate governing equations for nonlinear dynamical systems.
LLMs improve financial analysis by processing large data sets.
problem Traditional financial analysis methods struggle with large data volumes.
method Integrating LLMs for enhanced data processing and analysis.
result LLMs offer new capabilities for real-time financial decision-making.
The paper tackles extrapolation in generative models by enforcing independence of mechanisms.
problem How to make generative models extrapolate to new, unseen environments?
method Developed a theoretical framework for independence of mechanisms, demonstrated on toy examples and real-world data.
result Extrapolation capabilities of generative models can be improved by enforcing independence of mechanisms explicitly during training.
Chronos-2 forecasts multivariate and covariate data without task-specific training.
problem Limited applicability of existing time series forecasting models to real-world multivariate and covariate data.
method Chronos-2 uses a group attention mechanism for in-context learning across multiple time series.
result Chronos-2 achieves state-of-the-art performance across comprehensive benchmarks.
Well-established methods for the solution of stochastic partial differential equations (SPDEs) typically struggle in problems with high-dimensional inputs/outputs. Such difficulties are only amplified in large-scale applications where even a few tens of full-order model runs are impracticable. While dimensionality redu…
The paper uses geometry to understand how neural networks learn.
problem Understanding the learning capability of neural networks.
method Statistical and differential geometric analysis of neural networks performing simple regression.
result Neural networks with higher generalization capability have a slower convergence rate.
Bayesian neural network models improve uncertainty quantification in multivariate regression.
problem Uncertainty quantification in multivariate regression models with heteroscedastic noise.
method Proposes Bayesian Last Layer neural network models and EM algorithms for parameter learning.
result Capable of disentangling aleatoric and epistemic uncertainty.
Music Inpainting is the task of filling in missing or lost information in a piece of music. We investigate this task from an interactive music creation perspective. To this end, a novel deep learning-based approach for musical score inpainting is proposed. The designed model takes both past and future musical context i…
New model predicts particle precipitation from magnetosphere to ionosphere.
problem Improving prediction of electron particle precipitation from magnetosphere to ionosphere.
method Compilation of new database, use of machine learning (ML) tools, neural network (PrecipNet).
result PrecipNet achieves >50% reduction in errors and better captures dynamic changes.
SYNTHONY selects tabular synthesizers based on stress profiling and user intent.
problem Non-uniform performance of tabular generative models across datasets.
method Stress profiling and intent-conditioned tabular synthesis selection.
result Meta-features predict synthesizer performance, improving selection accuracy.
DLFM models complex systems with uncertainty, outperforming traditional methods.
problem Modeling highly nonlinear dynamical systems with robust uncertainty quantification.
method Deep latent force model (DLFM) using physics-informed kernels derived from ODEs.
result DLFM achieves comparable performance to non-physics-informed models on univariate tasks and captures dynamics in real-world data.
We present Manifold Alignment Determination (MAD), an algorithm for learning alignments between data points from multiple views or modalities. The approach is capable of learning correspondences between views as well as correspondences between individual data-points. The proposed method requires only a few aligned exam…
We present a comprehensive study on the use of autoencoders for modelling text data, in which (differently from previous studies) we focus our attention on the following issues: i) we explore the suitability of two different models bDA and rsDA for constructing deep autoencoders for text data at the sentence level; ii)…
We present an agent based model of a single asset financial market that is capable of replicating several non-trivial statistical properties observed in real financial markets, generically referred to as stylized facts. While previous models reported in the literature are also capable of replicating some of these stati…
While the Machine Learning (ML) landscape is evolving rapidly, there has been a relative lag in the development of the "learning systems" needed to enable broad adoption. Furthermore, few such systems are designed to support the specialized requirements of scientific ML. Here we present the Data and Learning Hub for sc…
Recurrent neural networks (RNNs) are powerful constructs capable of modeling complex systems, up to and including Turing Machines. However, learning such complex models from finite training sets can be difficult. In this paper we empirically show that RNNs can learn models of computer peripheral devices through input a…
Transformers learn to play games in-context, proving Nash equilibrium.
problem Understanding in-context game-playing capabilities of pre-trained transformers.
method Theoretical guarantees and constructional results for transformer architecture in multi-agent games.
result Pre-trained transformers can learn Nash equilibrium in-context for two-player zero-sum games.
Experts predict significant adoption of decentralized finance by 2034, with traditional finance adapting.
problem Adoption and integration of decentralized finance (DeFi) in financial services.
method Survey analysis using New Institutional Economics and Dynamic Capabilities Theory.
result Experts expect adoption of DeFi to rise from negligible to 43% by 2034, with traditional finance likely to embrace it.
Agents need world models to generalize multi-step tasks.
problem The necessity of world models for flexible, goal-directed behavior.
method Formal analysis and demonstration of the necessity of world models for agents to generalize multi-step tasks.
result World models are necessary for agents to generalize to multi-step goal-directed tasks.
SSMs combined with neural networks match Transformers in dynamic token selection.
problem Understanding the capabilities of SSMs in dynamic token selection.
method Exploring SSMs combined with fully connected neural networks.
result SSMs combined with nonlinear layers can efficiently solve challenging tasks and estimate functions.
Language models can predict numeric values as strings.
problem Regression tasks with numeric predictions.
method Causal sequence decoding models trained for next-token prediction.
result Decoder-based heads perform as well as standard heads in numeric regression tasks.
We present in this paper a model for forecasting short-term power loads based on deep residual networks. The proposed model is able to integrate domain knowledge and researchers' understanding of the task by virtue of different neural network building blocks. Specifically, a modified deep residual network is formulated…
As research into community finding in social networks progresses, there is a need for algorithms capable of detecting overlapping community structure. Many algorithms have been proposed in recent years that are capable of assigning each node to more than a single community. The performance of these algorithms tends to …
Bayesian framework for sequential learning tasks with low-rank approximations.
problem Balancing knowledge retention and adaptability in sequential neural networks.
method Bayesian framework with diagonal plus low-rank approximations of the precision matrix.
result Unlocking capabilities to encode task relationships and incorporate prior knowledge from later tasks.
Transformers solve Gaussian Mixture Models without supervision.
problem Solving Gaussian Mixture Models (GMMs) unsupervised.
method Proposes TGMM, a transformer-based framework for GMM tasks.
result Transformers can effectively solve GMM tasks, improving upon classical methods.