OCC system speeds up in-app communications for Uber drivers and riders.
problem Improving efficiency in in-app messaging between drivers and riders.
method Intents are detected using unsupervised embedding and nearest-neighbor classifier. Replies are retrieved based on popularity in historical data.
result System achieves 76% accuracy in intent detection and 71% adoption rate of smart replies.
While mobile social apps have become increasingly important in people's daily life, we have limited understanding on what motivates users to engage with these apps. In this paper, we answer the question whether users' in-app activity patterns help inform their future app engagement (e.g., active days in a future time w…
Study combines chit-chat and goal-oriented dialogue in fantasy games.
problem Combining naturalistic chit-chat with goal-oriented tasks in fantasy games.
method Trained a goal-oriented model with reinforcement learning against an imitation-learned chit-chat model using two approaches.
result Both models outperform a baseline and can converse naturally to achieve goals.
DAG-LSTM improves DA classification in group chats.
problem DA classification in multi-party conversations.
method Directed-Acyclic-Graph LSTM (DAG-LSTM) exploiting turn-taking structure.
result DAG-LSTM outperforms existing methods by 0.8% in accuracy and 1.2% in macro-F1 score.
Sequence generative adversarial networks (SeqGAN) have been used to improve conditional sequence generation tasks, for example, chit-chat dialogue generation. To stabilize the training of SeqGAN, Monte Carlo tree search (MCTS) or reward at every generation step (REGS) is used to evaluate the goodness of a generated sub…
Enhances Transformer for hierarchical language understanding.
problem Lack of explicit hierarchical structure in Transformer models.
method Inspired by U-Net, integrates hierarchical processing into Transformers.
result Improved performance in chit-chat dialogue tasks.
DGDS uses documents to center conversations, promising broader AI understanding.
problem DS classification by function is insufficient for complex conversations.
method Classify DS based on document grounding, analyzing classification, architecture, datasets, and models.
result DGDS can better represent current DS development trends and future AI understanding.
RL platform enhances user journeys in healthcare apps.
problem Improving user experience and personalization in healthcare apps.
method Reinforcement learning framework for adaptive interventions.
result Significant increase in basket size through personalized recommendations.
A new protocol corrects confounding effects to measure alignment-induced activation shifts accurately.
problem Confounding effects in measuring alignment-induced activation shifts using naive methods.
method Introduces a four-variant decomposition to separate alignment shift from template effects.
result Correctly measures alignment-induced activation shifts, recovering behaviorally active subspace.
Boosting trees predict Twitch subscriptions from user activity.
problem Predicting Twitch user subscriptions from activity data.
method Used boosting trees and target-encodings for high cardinality categoricals.
result User activity can be better predicted than content alone.
In sequence generation task, many works use policy gradient for model optimization to tackle the intractable backpropagation issue when maximizing the non-differentiable evaluation metrics or fooling the discriminator in adversarial learning. In this paper, we replace policy gradient with proximal policy optimization (…
Understanding player behavior is fundamental in game data science. Video games evolve as players interact with the game, so being able to foresee player experience would help to ensure a successful game development. In particular, game developers need to evaluate beforehand the impact of in-game events. Simulation opti…
We propose a friend recommendation system (an application of link prediction) using edge embeddings on social networks. Most real-world social networks are multi-graphs, where different kinds of relationships (e.g. chat, friendship) are possible between a pair of users. Existing network embedding techniques do not leve…
New models avoid lookahead bias by training on past data only.
problem Lookahead bias in language models.
method Chronologically consistent training on data before a knowledge-cutoff date.
result Elimination of lookahead bias in predictions.
Pronoun resolution is part of coreference resolution, the task of pairing an expression to its referring entity. This is an important task for natural language understanding and a necessary component of machine translation systems, chat bots and assistants. Neural machine learning systems perform far from ideally in th…
The majority of conversations a dialogue agent sees over its lifetime occur after it has already been trained and deployed, leaving a vast store of potential training signal untapped. In this work, we propose the self-feeding chatbot, a dialogue agent with the ability to extract new training examples from the conversat…
How can we measure whether a natural language generation system produces both high quality and diverse outputs? Human evaluation captures quality but not diversity, as it does not catch models that simply plagiarize from the training set. On the other hand, statistical evaluation (i.e., perplexity) captures diversity b…
Study finds neural dialog models struggle with conversational tasks.
problem Insufficient understanding of dialog by neural models.
method Analysis of internal representations and evaluation of model performance.
result Neural dialog models lack key conversational skills like answering questions and inferring contradiction.
To help enforce data-protection regulations such as GDPR and detect unauthorized uses of personal data, we develop a new \emph{model auditing} technique that helps users check if their data was used to train a machine learning model. We focus on auditing deep-learning models that generate natural-language text, includi…
Nowadays, video game developers record every virtual action performed by their players. As each player can remain in the game for years, this results in an exceptionally rich dataset that can be used to understand and predict player behavior. In particular, this information may serve to identify the most valuable playe…
Study examines flaws in probing LLMs' knowledge and introduces a new method.
problem Flaws in existing methods for probing the veracity of LLMs' internal knowledge.
method sAwMIL (Sparse-Aware Multiple-Instance Learning) combining multiple-instance learning with conformal prediction.
result LLMs encode a third type of signal distinct from true and false.
There has been growing interest in using neural networks and deep learning techniques to create dialogue systems. Conversational recommendation is an interesting setting for the scientific exploration of dialogue with natural language as the associated discourse involves goal-driven dialogue that often transforms natur…
Authorship verification (AV) is a research subject in the field of digital text forensics that concerns itself with the question, whether two documents have been written by the same person. During the past two decades, an increasing number of proposed AV approaches can be observed. However, a closer look at the respect…
This paper tests LLMs in finance to assess ethical behavior.
problem Aligning AI with ethical and legal standards in finance.
method Prompted LLMs to simulate CEO behavior, analyzed with logistic regression.
result Significant heterogeneity in LLMs' unethical behavior propensity.
DataInf efficiently approximates data influence in large models, improving transparency and identifying mislabeled data.
problem Efficiently estimating data influence in large-scale models like LoRA-tuned LLMs and diffusion models.
method DataInf uses a closed-form expression to approximate influence scores efficiently.
result DataInf outperforms existing methods in computational and memory efficiency, accurately identifying influential data points.
Generates biomedical abstracts from titles, years, and keywords.
problem Difficulties in understanding biomedical research papers due to specialized language.
method Conditional transformer-based language model with metadata conditioning.
result Generated abstracts are more relevant and non-trivial than GPT-2.
Proposes a new method for conversational agents using deep learning.
problem Building coherent and non-monotonous conversational agents with proper discourse and coverage.
method End-to-end multi-stream deep learning architecture leveraging contextual and syntactic information.
result Significantly improved next sentence prediction task.
This paper proposes an approach to detect emotion from human speech employing majority voting technique over several machine learning techniques. The contribution of this work is in two folds: firstly it selects those features of speech which is most promising for classification and secondly it uses the majority voting…
LLMs struggle to generate random numbers from statistical distributions, leading to biased results in applications.
problem LLMs' inability to generate random numbers accurately from specified distributions.
method Dual-protocol design: Batch Generation and Independent Requests, benchmarking 11 models across 15 distributions.
result Sampling fidelity degrades with distributional complexity and horizon, leading to systematic biases in downstream applications.
Enhances math problem-solving models with multi-turn preference learning.
problem Improving mathematical problem-solving capabilities of large language models.
method Introduces a multi-turn direct preference learning framework for tool-integrated mathematical reasoning tasks.
result Significant performance improvements in model accuracy on math datasets.
AI model automates financial investment research tasks.
problem Manual labor-intensive tasks in financial analysis.
method Fine-tuning language models on diverse financial data.
result Improved model performance in financial tasks.
VB-Score evaluates AI systems without ground truth, revealing robustness.
problem Evaluating AI systems without ground truth labels, especially for entity-centric tasks.
method VB-Score uses variance-bounded evaluation, constraint relaxation, and Monte Carlo sampling.
result VB-Score reveals robustness differences not seen by conventional frameworks.
This research aims to identify how Bitcoin-related news publications and online discourse are expressed in Bitcoin exchange movements of price and volume. Being inherently digital, all Bitcoin-related fundamental data (from exchanges, as well as transactional data directly from the blockchain) is available online, some…
Gray-box attack improves on white-box methods for trading agents.
problem Robustness of Deep RL trading agents against adversarial attacks.
method Hybrid Deep Neural Network policy for gray-box adversarial attack.
result Adversary can reduce trading agent's reward by 214.17%.
SC unifies ICL calibration methods and improves LLM performance.
problem Systematic biases in LLM predictions leading to unstable performance.
method Supervised Calibration (SC) learns optimal affine transformations in logit space.
result SC delivers state-of-the-art performance across multiple datasets.
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.
Train a lightweight carry-on model on existing LLMs for faster customization.
problem Customizing large language models for specific tasks is computationally expensive.
method Train an additional branch of transformer blocks on the final-layer embedding of pretrained LLMs, then merge them with a carry-on module.
result Training a 100M carry-on layer requires less than 1GB GPU memory, making it scalable and affordable.
Silent abandonment reduces contact center efficiency by 5%-15%.
problem Measuring customer abandonment and patience in text-based contact centers is challenging due to uncertainty.
method Developed methodologies to identify silent-abandonment customers and estimate customer patience using text analysis and queueing models.
result Silent abandonment accounts for 30%-67% of customer abandonments and reduces system efficiency by 5%-15%.
Simple adaptive attacks can jailbreak state-of-the-art safety-aligned language models.
problem Vulnerability of safety-aligned language models to simple adaptive attacks.
method Designing adversarial prompt templates, random search on suffixes, and transfer/prefilling attacks.
result 100% attack success rate on various models including GPT-4, Vicuna, Mistral, and Claude.
Deleting refusal directions from models leads to systematically more optimistic decisions.
problem The impact of removing refusal directions from models on decision-making outcomes.
method Ablation study using a frozen pipeline of 21,600 weekly equity decisions.
result Ablation of refusal directions makes models more optimistic and justifies themselves more, but also reduces confidence.