A new robust prefix-tuning framework improves model robustness against adversarial attacks.
problem Lack of robustness in prefix-tuning for adversarial attacks.
method Leveraging layerwise activations of pretrained models for additional prefix finetuning during the test phase.
result Framework substantially improves robustness over strong baselines while maintaining comparable accuracy on clean texts.
PPT optimizes transformer behavior by steering its latent posterior using prior samples.
problem Eliciting desired behavior from transformers without backpropagation.
method Posterior Prefix Tuning (PPT) uses predictive Monte Carlo (PMC) samples and importance sampling to optimize the latent posterior.
result PPT optimizes transformer behavior without backpropagation, achieving high utility across different utility functions.
Improved text generation with discourse-aware soft prompts.
problem Efficient fine-tuning methods don't generalize across all generation tasks.
method Hierarchical blocking and attention sparsity on prefix parameters.
result Structured design of prefix parameters yields more coherent generations.
This paper explains how to optimize prompts for model adaptation.
problem Understanding and optimizing prompt tuning for model adaptation.
method Bayesian view and meta-learning to explain prompt optimization.
result Optimal prompting can be studied formally as conditioning Bayesian predictors.
Improved financial VA intent classification accuracy.
problem Determining user intents for unseen open intents.
method Supervised pre-training of intent representations using prefix-tuning and fine-tuning.
result 1.63% - 2.07% higher accuracy on banking77 benchmark.
CROP verifies clean prefixes in reasoning traces, improving downstream repair accuracy.
problem Uncertainty in reasoning traces prevents full certification of entire responses.
method CROP selects a calibrated threshold to certify the longest prefix with low risk proxies.
result CROP improves downstream repair accuracy by preserving valid reasoning and discarding misleading suffixes.
The paper identifies patterns in language model weights used for memorizing paragraphs.
problem Locating the specific mechanisms and weights used by language models to memorize paragraphs.
method Examined gradients and attention patterns in language models to identify memorized paragraphs.
result Gradients of memorized paragraphs have a distinguishable spatial pattern, and localized attention heads are involved in paragraph memorization.
Prefix consistency improves model reliability by weighting answers based on their reproducibility.
problem Improving the reliability of large language models' reasoning traces.
method Use prefix consistency to weight candidate answers based on their reproducibility during regeneration.
result Prefix consistency is the best correctness predictor, reaching Standard MV plateau accuracy with up to 21x fewer tokens.
ReOPD uses pre-collected teacher trajectories to distill knowledge from multi-turn interactions.
problem The cost of fully online on-policy distillation for multi-turn interactions.
method ReOPD, an off-environment alternative that reuses pre-collected teacher trajectories as replayed prefixes, addressing the prefix trap and distribution shift.
result ReOPD preserves or improves OPD-level accuracy, uses zero tool calls, and is at least 4imes faster per training step. New methods for tuning alpha in Gibbs posteriors improve speed and accuracy.
problem Inconsistency in Bayesian inference and lack of fast tuning methods for alpha.
method Proposed two data-driven methods: sample-splitting and bootstrapping. Formulated alpha-posteriors for three models.
result Sample-splitting outperforms SafeBayes in speed and accuracy, especially in complex models.
Cold posteriors improve Bayesian neural networks by reducing overestimation of aleatoric uncertainty.
problem Overestimation of aleatoric uncertainty in Bayesian neural networks.
method Tuning the temperature of the posterior on a validation set.
result Reducing temperature leads to better reflection of true prior beliefs.
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.
PFN-TS uses Thompson sampling with PFNs to improve contextual bandit performance.
problem Improving contextual bandit performance using Thompson sampling with prior-data fitted networks.
method PFN-TS converts PFN posterior predictives into mean-reward samples using a subsampled predictive central limit theorem.
result PFN-TS achieves the best average rank across nonlinear synthetic and OpenML classification-to-bandit benchmarks.
Exposure bias has been regarded as a central problem for auto-regressive language models (LM). It claims that teacher forcing would cause the test-time generation to be incrementally distorted due to the training-generation discrepancy. Although a lot of algorithms have been proposed to avoid teacher forcing and theref…
KVCOMM optimizes multi-agent LLM systems by reusing KV-caches, reducing redundant processing.
problem Substantial overhead from reprocessing overlapping contexts across multi-agent systems.
method KVCOMM reuses KV-caches and adjusts offsets for shared content using a pool of cached examples (anchors).
result Achieves over 70% reuse rate across diverse multi-agent tasks, up to 7.8x speedup.
Proposes a method to quantify uncertainty in PFNs.
problem Lack of uncertainty quantification in PFNs.
method Martingale posteriors for efficient, tuning-free sampling.
result Proves convergence of proposed sampling procedure.
A new method improves uncertainty quantification in Bayesian inference.
problem Poor uncertainty quantification in traditional Gibbs posteriors.
method Sequential Gibbs posteriors with a Bernstein-von Mises theorem.
result Sequential Gibbs posteriors provide better frequentist coverage.
New algorithm tunes SGMCMC hyperparameters for scalable Bayesian inference.
problem Tuning hyperparameters for SGMCMC is challenging due to lack of principled methods.
method Proposes a bandit-based algorithm using Stein discrepancies to tune hyperparameters.
result The method effectively tunes SGMCMC hyperparameters for various applications.
BBVI relies on adaptive stochastic optimization algorithms for posterior approximation, but these require extensive tuning.
problem BBVI posterior approximation
method Adaptive stochastic optimization algorithms
result No single method dominates, but a selection of 5 suffices.
Neural networks with learned biases can approximate any function.
problem Whether neural networks with only learned biases can approximate any continuous function.
method Theoretical and numerical analysis of random weights and learned biases in neural networks.
result Feedforward and recurrent neural networks with random weights can approximate any continuous function and dynamical systems.
A new method combines Gaussian Processes to optimize under uncertainty.
problem Bayesian Optimization's weakness in fitting Gaussian Processes.
method Wasserstein Barycenter Gaussian Process (WBGP) approach.
result WBGP-BO converges to the optimum, improving on vanilla BO.
Generative approach speeds hyperparameter tuning for machine learning models.
problem Computational infeasibility of cross-validation and difficulty of fully Bayesian hyper-parameter learning.
method Combines optimization-based approximations and amortization techniques.
result Rapid evaluation of hyper-parameters over grids or ranges, supporting predictive tuning and uncertainty quantification.
Autostackability for finitely generated groups is defined via a topological property of the associated Cayley graph which can be encoded in a finite state automaton. Autostackable groups have solvable word problem and an effective inductive procedure for constructing van Kampen diagrams with respect to a canonical fini…
Study risk-constrained Kelly optimization for mutually exclusive outcomes, proving support invariance and developing a structured algorithm.
problem Risk-constrained Kelly optimization for mutually exclusive outcomes with explicit state prices.
method Analyzes the finite mutually exclusive outcome version of risk-constrained Kelly optimization with explicit state prices, proving support invariance and developing a structured algorithm.
result Support is invariant across CRRA parameter and drawdown-surrogate parameter in the overround regime.
SMART-FAN-Lasso fine-tunes neural networks for high-dimensional nonparametric regression.
problem Fine-tuning neural networks for high-dimensional nonparametric regression with variable selection.
method Source-model-augmented residual tuning (SMART) framework for neural Lasso.
result SMART-FAN-Lasso achieves statistical acceleration over single-task learning under precise conditions.
New method for efficient probabilistic inference using masked language modeling.
problem Efficient posterior inference in probabilistic programs with many hyper-parameters.
method Formulate inference as masked language modeling, train a neural network to unmask random values.
result Foundation posterior for zero-shot inference and fine-tuning across a range of programs.
New algorithm reduces regret in private online learning with optimal gap-dependent rate.
problem Optimal gap-dependent regret rate for private stochastic decision-theoretic online learning.
method Horizon-free pure-DP algorithm with exponential block partitioning and softmax selection.
result Explicit regret bound of 1000⋅(ΔminlogK+εlogK). Improved Bayesian FL method calibrates predictions for federated learning.
problem Overconfident predictions in Bayesian FL methods for federated learning.
method β-Predictive Bayes algorithm interpolates between mixture and product of local predictive posteriors, tuning parameter β for better calibration.
result Demonstrated superior calibration compared to other baselines, even with increased data heterogeneity.
Optimized α-posteriors reduce KL divergence from true posterior in parametric misspecification.
problem Reduction of KL divergence from true posterior in parametric model misspecification.
method Derivation of Bernstein-von Mises theorem and optimization of α-posteriors. result Optimized α-posteriors minimize KL divergence from true posterior, especially in severe misspecification. New method reduces computational cost for Bayesian inference.
problem High computational cost in Bayesian inference.
method Coreset MCMC, a Markov chain targeting coreset posterior.
result Coreset MCMC provides higher quality posterior approximations and reduced computational cost.
FlowVAT improves variational inference for multi-modal distributions.
problem Mode-seeking behavior and collapse in variational inference for complex posteriors.
method Conditional tempering approach for normalizing flow variational inference.
result FlowVAT outperforms traditional and adaptive annealing methods in multi-modal distributions, finding more modes and achieving better ELBO values.
AutoGraph uses transformers to efficiently generate graphs as sequences.
problem Efficiently generating large, sparse graphs without expensive node features.
method Flattening graphs into sequences and using decoder-only transformers.
result AutoGraph achieves state-of-the-art performance on synthetic and molecular benchmarks.
Improved bounds on acylindricity for right-angled Artin groups.
problem Bounding the acylindrical action of right-angled Artin groups on their extension graphs.
method Exploring lattice properties, studying prefixes of powers, and extending quasi-root uniqueness.
result Cardinality of r-quasi-stabilizer is bounded by a linear function of r. A new method improves language model extrapolation without changing training.
problem Improving language model performance on sequences not seen during training.
method Perturbation-based training procedure for language models.
result The method consistently improves out-of-support prediction.
We analyze SGAs for statistical inference via asymptotics, improving tuning methods.
problem Improper tuning of SGAs for optimization and sampling.
method Characterize large-sample asymptotics of SGAs via step-size and sample-size scaling limits.
result Iterate averaging with large step size is robust and asymptotically has covariance proportional to MLE's.
Penalized regression models are popularly used in high-dimensional data analysis to conduct variable selection and model fitting simultaneously. Whereas success has been widely reported in literature, their performances largely depend on the tuning parameters that balance the trade-off between model fitting and model s…
As deep learning techniques advance more than ever, hyper-parameter optimization is the new major workload in deep learning clusters. Although hyper-parameter optimization is crucial in training deep learning models for high model performance, effectively executing such a computation-heavy workload still remains a chal…
We consider the problem of path inference: given a path prefix, i.e., a partially observed sequence of nodes in a graph, we want to predict which nodes are in the missing suffix. In particular, we focus on natural paths occurring as a by-product of the interaction of an agent with a network---a driver on the transporta…
New method tunes prior IP to data for flexible predictive distributions.
problem Challenges in approximate inference for large models with high parameter dependencies.
method Inducing-point representation of prior IP to approximate posterior process.
result Scalable method that tunes prior IP to data and provides accurate non-Gaussian predictive distributions.
We improve private training accuracy with learning rate schedules and matrix factorizations.
problem Private training with learning rate schedules and correlated noise.
method General upper and lower bounds for learning rate schedules, memory-efficient constructions, and schedule-aware factorizations.
result Schedule-aware factorizations improve accuracy in private training.
DTM improves dLLM fine-tuning stability and performance.
problem Intractable sequence-level marginal likelihoods for masked diffusion models.
method Discrete Tilt Matching (DTM) recasts dLLM fine-tuning as state-level matching of local unmasking posteriors under reward tilting.
result DTM yields strong gains on Sudoku and Countdown while remaining competitive on MATH500 and GSM8K.
Bayesian approach speeds up SPN learning and inference.
problem Increased learning-time complexity for large SPNs.
method Bayesian posterior sampling with hyperparameter tuning.
result Significant improvement in learning-time complexity and predictive performance.
New priors improve Bayesian neural networks without cooling.
problem Bayesian neural networks underfit on clean datasets.
method Introduce DirClip and confidence priors to replace cooling.
result DirClip and confidence priors outperform cold posterior.
MFMs enable efficient reward alignment for generative models.
problem Computational bottleneck in controlling generative models.
method Meta Flow Maps (MFMs) extend consistency models and flow maps to stochastic regime for efficient value function estimation.
result MFMs enable inference-time steering and unbiased, off-policy fine-tuning to general rewards efficiently.
Enhanced Gaussian process models accelerate optimization and posterior approximation.
problem Improving the accuracy and speed of Gaussian process models for optimization and inference.
method Introduces a random exploration step to classical GP-UCB algorithms, facilitating faster convergence.
result New algorithms achieve nearly optimal convergence rates and provide bounds for Hellinger distance.
Elliptical slice sampling converges geometrically, providing reliable sampling for Bayesian learning.
problem Sampling from posterior distributions in Bayesian learning.
method Elliptical slice sampling, geometric ergodicity.
result Elliptical slice sampling yields geometric convergence guarantees under weak regularity assumptions.
Exact posterior score estimation for solving linear inverse problems
problem Solving linear inverse problems
method Derive the exact posterior score and use it as a denoising training objective
result EPS outperforms training-free and training-based baselines on various metrics
The following discourse is inspired by the works on hyperbolic groups of Epstein, and Neumann/Reeves. Epstein showed that geometrically finite hyperbolic groups are biautomatic. Neumann/Reeves showed that virtually central extensions of word hyperbolic groups are biautomatic. We prove the following generalisation: Theo…