Automates debiasing for large language model evaluations through Fisher random walk.
problem Rigorous and scalable evaluation of large language models.
method Semiparametric efficient estimator using Fisher random walk for weighted residual balancing.
result Efficient estimation of contextual preference scores for large language models.
This research explores using kernels in the softmax layer for better contextual word classification.
problem Improving contextual word classification accuracy.
method Replacing the inner product in the softmax layer with various kernel functions and comparing their performance.
result Different kernel settings yield varying performance in contextual word classification tasks.
Contextualizing financial news improves stock price predictions.
problem Predicting stock prices from financial news requires understanding historical context.
method Proposed a method using a large language model for main articles and a small model for historical context.
result Historical context significantly improves model performance across methods and time horizons.
Neural models use default reasoning for number and gender assignment tasks.
problem Understanding how neural language models make decisions for grammatical phenomena.
method Generalised Contextual Decomposition (GCD) to isolate semantic, syntactic, and bias-driven components of predictions.
result Models rely on default reasoning for tasks like number and gender assignment.
This study analyzes how RNNs process context in sentiment analysis.
problem Understanding how recurrent neural networks process context in sentiment analysis.
method Developed methods to reverse engineer RNNs, identifying contextual effects and quantifying their strength and timescale.
result Identified inputs that induce contextual effects and quantified their properties.
We propose a simple modification to existing neural machine translation (NMT) models that enables using a single universal model to translate between multiple languages while allowing for language specific parameterization, and that can also be used for domain adaptation. Our approach requires no changes to the model a…
Study compares BERT and XLNet for multi-class categorization of product descriptions.
problem Robustness of multi-class categorization using pre-trained contextualized language models.
method Fine-tuning BERT and XLNet on Amazon product data for multi-class classification.
result Performance decreases linearly with the number of class labels, with BERT consistently outperforming XLNet.
Model identifies urgent radiology reports with high accuracy.
problem Lack of annotated training data for text analysis.
method Self-supervised contextual language representation using BERT.
result Model achieved 97.0% precision, 93.3% recall, and 95.1% F-measure.
Elliptical Attention improves transformer performance by focusing on contextually relevant features.
problem Transformer models suffer from representation collapse and are vulnerable to contaminated samples.
method Uses Mahalanobis distance to define hyper-ellipsoidal neighborhoods for attention weights.
result Elliptical Attention reduces representation collapse and enhances model robustness.
Text documents are structured on multiple levels of detail: individual words are related by syntax, but larger units of text are related by discourse structure. Existing language models generally fail to account for discourse structure, but it is crucial if we are to have language models that reward coherence and gener…
Improved financial sentiment analysis using simple instruction tuning of LLMs.
problem Lack of accurate financial sentiment analysis by large language models.
method Instruction tuning of general-purpose LLMs with a small portion of financial sentiment data.
result Significant improvement in financial sentiment analysis, especially in complex scenarios.
New neural network layer handles OOV words in NLP tasks without pre-training.
problem Handling out-of-vocabulary words in natural language processing.
method Contextual-compositional neural network layer that attends to character sequence and context.
result Improves performance on 23 languages in joint tagging tasks.
Proposes an online model for LLM cascading with adaptive API selection.
problem Adaptive querying and selection of LLM APIs in a context-dependent environment.
method Develops a learning approach combining GMM estimation and UCB-style bounds.
result Achieves cumulative regret of O ~ ( T ) \widetilde O(\sqrt T) O ( T ) over T T T periods. The Widrow-Hoff rule simplifies language data simulation.
problem Simulating language phenomena computationally.
method Implementation and application of the Widrow-Hoff rule.
result The Widrow-Hoff rule offers new perspectives on language simulation.
LLM-Lasso uses LLMs to improve feature selection in Lasso regression.
problem Improving feature selection in Lasso regression with domain-specific knowledge.
method Combines LLMs with Lasso regularization to generate feature weights.
result Outperforms standard Lasso and feature selection baselines in biomedical studies.
LoCo-RLHF models diverse human feedback with contextual information.
problem Heterogeneous human feedback from diverse contexts and preferences.
method Low-rank contextual preference model, PRS policy.
result LoCo-RLHF achieves tighter sub-optimality gap than existing methods.
Study on how attention in prompt-tuning affects large language models.
problem Limited theoretical understanding of prompt-tuning and attention in LLMs.
method Exploration of prompt-tuning for one-layer attention architectures, contextual mixture-models, and self-contained prompt-attention model.
result Softmax-prompt-attention is more expressive than self-attention and linear-prompt-attention under contextual data model.
Efficiently identifies good policies by choosing contexts for human feedback.
problem Efficiently acquiring human feedback for preference alignment in large language models.
method Formalizes active exploration as a dueling bandit problem and proposes an active exploration algorithm with a polynomial worst-case regret bound.
result Proposed method outperforms baselines with limited human preferences on various language models and datasets.
Enhances GNNs with text features for better fake news detection.
problem Detecting disinformation on social media using GNNs.
method Integrates Transformer-based textual features into GNNs.
result Contextual text representations improve GNN performance by 33.8% in Macro F1.
Paper proposes a new RLHF framework for human preference learning.
problem Handling dependent online human preference outcomes with dynamic contexts.
method Two-stage algorithm with ε ε ε -greedy followed by exploitation; anti-concentration inequalities and matrix martingale concentration techniques. result Our method achieves optimal regret bound and asymptotic normality of estimators.
Paper tackles Arabic question similarity, outperforming state-of-the-art.
problem Detecting semantically similar questions in Arabic is challenging.
method Utilizes contextualized word representations (ELMo embeddings) trained on MSA and dialectic sentences, combined with a pairwise similarity layer.
result Achieves 93% F1-score on Modern Standard Arabic benchmark and 82% on dialectical benchmark.
This paper assesses biases in contextualized word representations.
problem Analyzing biases in contextualized word representations.
method Proposes assessing bias at the contextual word level, capturing contextual effects of bias.
result Demonstrates evidence of bias in contextual word models, including racial bias and exacerbated effects for intersectional minorities.
System learns to combine multiple model components for personalized text generation.
problem Adapting and biasing language models for personal preferences.
method Combines model-defined components, learns activation and probability combination from unlabeled text.
result Directly generates text with personalized components from unlabeled data.
Project aims to improve coherence in language generation models.
problem Models often generate inconsistent text that diverges from the prompt.
method Trained a sentence pair coherence classifier and co-trained GPT-2 with this coherence objective.
result Fine-tuned model generates coherent paragraphs without diverging.
Recent Transformer-based contextual word representations, including BERT and XLNet, have shown state-of-the-art performance in multiple disciplines within NLP. Fine-tuning the trained contextual models on task-specific datasets has been the key to achieving superior performance downstream. While fine-tuning these pre-t…
Efficiently identifies good policies by choosing contexts for human feedback.
problem Efficiently identifying good policies in applications with high feedback costs.
method Introduces offline contextual dueling bandit setting and an upper-confidence-bound style algorithm.
result Proves a regret bound and shows superior performance over uniformly sampled contexts.
FINCH dataset enables financial Text-to-SQL tasks, improving model evaluation.
problem Lack of a large-scale financial dataset for Text-to-SQL tasks.
method Curated financial dataset (FINCH), benchmarking reasoning and language models.
result Proposed FINCH Score for more accurate financial model evaluation.
Survey of LLMs in finance tasks, highlighting progress and challenges.
problem Transforming financial practices with advanced LLMs.
method Exploration of various financial tasks, categorization, and analysis of methodologies.
result Unlocking novel opportunities for financial applications with LLMs.
In this work, we describe practical lessons we have learned from successfully using contextual bandits (CBs) to improve key business metrics of the Microsoft Virtual Agent for customer support. While our current use cases focus on single step einforcement learning (RL) and mostly in the domain of natural language proce…
This paper improves neural network explanations by quantifying and visualizing semantic compositions.
problem Improving neural network explanations for natural language processing tasks.
method Proposes a formal way to quantify word and phrase importance, introduces SCD and SOC algorithms.
result Our algorithms outperform prior methods in explaining neural network predictions.
A model integrates CNN and LSTM with LLM for better stock forecasting.
problem Complex stock market prediction with volatile patterns and multifaceted data.
method Two-Level Conv-LSTM Neural Network with LLM for holistic stock advising.
result Improved prediction accuracy and contextual insights.
LR-Robot automates SLRs with AI, expert oversight, and multidimensional analysis.
problem Efficient but contextually limited outputs from existing SLR frameworks.
method Human-in-the-loop process, structured knowledge sources, retrieval-augmented generation.
result Empirical demonstration of AI-driven literature synthesis in option pricing.
Law explains how LLMs learn to predict next tokens.
problem Understanding how LLMs process input data internally.
method Introduced a precise law governing token embeddings in LLMs.
result Each layer equally contributes to next-token prediction accuracy.
Paper proposes efficient RLHF methods for LLMs using active queries.
problem Efficiently aligning LLMs with human preferences using RLHF.
method Formalizes RLHF as a dueling bandit problem, introduces APPO and ADPO algorithms.
result ADPO achieves similar performance to state-of-the-art methods with fewer queries.
In this paper, we propose TopicRNN, a recurrent neural network (RNN)-based language model designed to directly capture the global semantic meaning relating words in a document via latent topics. Because of their sequential nature, RNNs are good at capturing the local structure of a word sequence - both semantic and syn…
Study uses LLMs to improve financial forecasting by integrating textual and numerical data.
problem Challenges in fusing multimodal information and measuring qualitative outputs from LLMs.
method Created context sets by segmenting daily securities reports into key factors and combining them with numerical data. Used dynamic updates and crafted prompts to assign scores to qualitative insights.
result LLMs outperform time-series models in market forecasting, though challenges remain.
This paper investigates how transformers can learn to generalize to unseen examples in context.
problem Understanding how transformers can generalize to unseen examples in a prompt.
method Gradient descent analysis of one-layer multi-head transformers for in-context learning.
result The training loss for a one-layer multi-head transformer converges linearly to a global minimum, effectively learning ridge regression over basis functions.
Study examines biases in clinical word embeddings, revealing performance gaps across groups.
problem Biases in clinical word embeddings leading to performance differences across groups.
method Pretrained BERT models on MIMIC-III, fill-in-the-blank method, fairness evaluation on clinical tasks.
result Classifiers trained from BERT representations exhibit statistically significant differences in performance across groups.
LLMs improve financial sentiment analysis in finance.
problem Defining and measuring financial sentiment.
method Investigation of sentiment measurement methods and LLMs.
result LLMs enhance financial sentiment analysis.
MEANTIME improves sequential recommendation by using multi-temporal embeddings and attention mechanisms.
problem Limited use of timestamp information and information bottleneck in sequential recommendation models.
method MEANTIME employs multiple types of temporal embeddings and attention mechanisms to capture diverse patterns from user behavior sequences.
result MEANTIME outperforms state-of-the-art sequential recommendation methods.
User interaction with voice-powered agents generates large amounts of unlabeled utterances. In this paper, we explore techniques to efficiently transfer the knowledge from these unlabeled utterances to improve model performance on Spoken Language Understanding (SLU) tasks. We use Embeddings from Language Model (ELMo) t…
New framework assesses LLMs' expertise using nonparametric ranking and confidence diagrams.
problem Evaluating and ranking large language models (LLMs) for alignment and performance.
method Nonparametric contextual ranking, confidence diagram, Gaussian multiplier bootstrap.
result Validated confidence diagram for assessing LLMs' domain-specific expertise.
Generative Adapter adapts LMs with a single forward pass, reducing inference overhead.
problem Efficient adaptation of large language models for new contexts.
method Generative Adapter directly maps new contexts to low-rank LM adapters via self-supervised learning.
result Significant reduction in inference overhead with no need for fine-tuning.
The paper proposes a deep learning technique for structured and composable representations.
problem Learning structured and composable representations from input images and discrete labels.
method End-to-end deep learning to learn representations based on distance estimates between class label and contextual information.
result The representations have a clear structure allowing for class and environment decomposition.
Microsoft Research Asia won first place in 8 out of 11 WMT19 language directions.
problem Improving machine translation quality across multiple languages.
method Transformer, back translation, knowledge distillation, MADL, MASS, NAO, SCA.
result Demonstrated significant improvement in 8 out of 11 WMT19 language directions.
Improves topic modeling using LLM embeddings and Poisson process.
problem Traditional topic modeling ignores word context; LLMs offer better contextual embeddings.
method Convert documents to word embeddings, model as Poisson process, estimate topics using flexible algorithm.
result Method integrates LLMs without fine-tuning, offers advantages over traditional methods.
Paper analyzes and improves KL-regularized RL for LLMs with logarithmic regret.
problem Improving efficiency of RL fine-tuning for large language models.
method Optimism-based KL-regularized online contextual bandit algorithm with novel regret analysis.
result Achieves an O ( η log ( N R T ) ⋅ d R ) \mathcal{O}\big(η\log (N_{\mathcal R} T)\cdot d_{\mathcal R}\big) O ( η log ( N R T ) ⋅ d R ) logarithmic regret bound. Combines BERT and GCN for better text classification.
problem Limited global information capture by BERT.
method Integrates BERT with VGCN for improved text classification.
result VGCN-BERT outperforms BERT and GCN alone.