Search Results for author:
Found 54 papers, 20 papers with code
An Analysis of State-of-the-Art Models for Situated Interactive MultiModal Conversations (SIMMC)
There is a growing interest in virtual assistants with multimodal capabilities, e. g., inferring the context of a conversation through scene understanding.
Inducing Group Fairness in LLM-Based Decisions
Prompting Large Language Models (LLMs) has created new and interesting means for classifying textual data.
Safety Alignment Should Be Made More Than Just a Few Tokens Deep
We refer to this issue as shallow safety alignment.
Robust Preference Optimization through Reward Model Distillation
Moreover, to account for uncertainty in the reward model we are distilling from, we optimize against a family of reward models that, as a whole, is likely to include at least one reasonable proxy for the preference distribution.
Mitigating Object Hallucination via Data Augmented Contrastive Tuning
For a given factual token, we create a hallucinated token through generative data augmentation by selectively altering the ground-truth information.
Reuse Your Rewards: Reward Model Transfer for Zero-Shot Cross-Lingual Alignment
In this work, we evaluate a simple approach for zero-shot cross-lingual alignment, where a reward model is trained on preference data in one source language and directly applied to other target languages.
Asymptotics of Language Model Alignment
The goal of language model alignment is to alter $p$ to a new distribution $\phi$ that results in a higher expected reward while keeping $\phi$ close to $p.$ A popular alignment method is the KL-constrained reinforcement learning (RL), which chooses a distribution $\phi_\Delta$ that maximizes $E_{\phi_{\Delta}} r(y)$ subject to a relative entropy constraint $KL(\phi_\Delta || p) \leq \Delta.$ Another simple alignment method is best-of-$N$, where $N$ samples are drawn from $p$ and one with highest reward is selected.
Block Verification Accelerates Speculative Decoding
Speculative decoding is an effective method for lossless acceleration of large language models during inference.
Gradient-Based Language Model Red Teaming
Red teaming is a common strategy for identifying weaknesses in generative language models (LMs), where adversarial prompts are produced that trigger an LM to generate unsafe responses.
Theoretical guarantees on the best-of-n alignment policy
A commonly used analytical expression in the literature claims that the KL divergence between the best-of-$n$ policy and the base policy is equal to $\log (n) - (n-1)/n.$ We disprove the validity of this claim, and show that it is an upper bound on the actual KL divergence.
Helping or Herding? Reward Model Ensembles Mitigate but do not Eliminate Reward Hacking
However, even pretrain reward ensembles do not eliminate reward hacking: we show several qualitative reward hacking phenomena that are not mitigated by ensembling because all reward models in the ensemble exhibit similar error patterns.
Multi-Group Fairness Evaluation via Conditional Value-at-Risk Testing
Here, the sample complexity for estimating the worst-case performance gap across groups (e. g., the largest difference in error rates) increases exponentially with the number of group-denoting sensitive attributes.
FRAPPE: A Group Fairness Framework for Post-Processing Everything
Despite achieving promising fairness-error trade-offs, in-processing mitigation techniques for group fairness cannot be employed in numerous practical applications with limited computation resources or no access to the training pipeline of the prediction model.
Improving Robustness via Tilted Exponential Layer: A Communication-Theoretic Perspective
State-of-the-art techniques for enhancing robustness of deep networks mostly rely on empirical risk minimization with suitable data augmentation.
Controlled Decoding from Language Models
KL-regularized reinforcement learning (RL) is a popular alignment framework to control the language model responses towards high reward outcomes.
Language Modelling
Multi-Objective Reinforcement Learning
+1
Break it, Imitate it, Fix it: Robustness by Generating Human-Like Attacks
We demonstrate the advantages of this system on the ANLI and hate speech detection benchmark datasets - both collected via an iterative, adversarial human-and-model-in-the-loop procedure.
Improving Few-shot Generalization of Safety Classifiers via Data Augmented Parameter-Efficient Fine-Tuning
As large language models (LLMs) are widely adopted, new safety issues and policies emerge, to which existing safety classifiers do not generalize well.
Improving Diversity of Demographic Representation in Large Language Models via Collective-Critiques and Self-Voting
A crucial challenge for generative large language models (LLMs) is diversity: when a user's prompt is under-specified, models may follow implicit assumptions while generating a response, which may result in homogenization of the responses, as well as certain demographic groups being under-represented or even erased from the generated responses.
SpecTr: Fast Speculative Decoding via Optimal Transport
We show that the optimal draft selection algorithm (transport plan) can be computed via linear programming, whose best-known runtime is exponential in $k$.
Enhancing Group Fairness in Online Settings Using Oblique Decision Forests
In the online setting, where the algorithm has access to a single instance at a time, estimating the group fairness objective requires additional storage and significantly more computation (e. g., forward/backward passes) than the task-specific objective at every time step.
A Systematic Survey of Prompt Engineering on Vision-Language Foundation Models
This paper aims to provide a comprehensive survey of cutting-edge research in prompt engineering on three types of vision-language models: multimodal-to-text generation models (e. g. Flamingo), image-text matching models (e. g.
Towards A Scalable Solution for Improving Multi-Group Fairness in Compositional Classification
Despite the rich literature on machine learning fairness, relatively little attention has been paid to remediating complex systems, where the final prediction is the combination of multiple classifiers and where multiple groups are present.
Let's Do a Thought Experiment: Using Counterfactuals to Improve Moral Reasoning
Language models still struggle on moral reasoning, despite their impressive performance in many other tasks.
Uncovering the Hidden Dynamics of Video Self-supervised Learning under Distribution Shifts
Video self-supervised learning (VSSL) has made significant progress in recent years.
Towards Robust Prompts on Vision-Language Models
With the advent of vision-language models (VLMs) that can perform in-context and prompt-based learning, how can we design prompting approaches that robustly generalize to distribution shift and can be used on novel classes outside the support set of the prompts?
Robust Conversational Agents against Imperceptible Toxicity Triggers
Existing work to generate such attacks is either based on human-generated attacks which is costly and not scalable or, in case of automatic attacks, the attack vector does not conform to human-like language, which can be detected using a language model loss.
Database Search Results Disambiguation for Task-Oriented Dialog Systems
We find that training on our augmented dialog data improves the model's ability to deal with ambiguous scenarios, without sacrificing performance on unmodified turns.
Know Thy Strengths: Comprehensive Dialogue State Tracking Diagnostics
Recent works that revealed the vulnerability of dialogue state tracking (DST) models to distributional shifts have made holistic comparisons on robustness and qualitative analyses increasingly important for understanding their relative performance.
Ranked #4 on
Multi-domain Dialogue State Tracking
on MULTIWOZ 2.1
(using extra training data)
Dialogue State Tracking
Multi-domain Dialogue State Tracking
+1
Information-Theoretic Bayes Risk Lower Bounds for Realizable Models
For models that are (roughly) lower Lipschitz in their parameters, we bound the rate distortion function from below, whereas for VC classes, the mutual information is bounded above by $d_\mathrm{vc}\log(n)$.
Federated Learning with Heterogeneous Differential Privacy
We propose a new algorithm for FL with heterogeneous DP, named FedHDP, which employs personalization and weighted averaging at the server using the privacy choices of clients, to achieve better performance on clients' local models.
Robustness through Data Augmentation Loss Consistency
Our experiments show that DAIR consistently outperforms ERM and DA-ERM with little marginal computational cost and sets new state-of-the-art results in several benchmarks involving covariant data augmentation.
Ranked #1 on
Multi-domain Dialogue State Tracking
on MULTIWOZ 2.2
Multi-domain Dialogue State Tracking
Visual Question Answering
On Tilted Losses in Machine Learning: Theory and Applications
Finally, we demonstrate that TERM can be used for a multitude of applications in machine learning, such as enforcing fairness between subgroups, mitigating the effect of outliers, and handling class imbalance.
Annotation Inconsistency and Entity Bias in MultiWOZ
In this work, we identify an overlooked issue with dialog state annotation inconsistencies in the dataset, where a slot type is tagged inconsistently across similar dialogs leading to confusion for DST modeling.
A Stochastic Optimization Framework for Fair Risk Minimization
We consider the problem of fair classification with discrete sensitive attributes and potentially large models and data sets, requiring stochastic solvers.
DVD: A Diagnostic Dataset for Multi-step Reasoning in Video Grounded Dialogue
A video-grounded dialogue system is required to understand both dialogue, which contains semantic dependencies from turn to turn, and video, which contains visual cues of spatial and temporal scene variations.
Coded Machine Unlearning
We also present the corresponding unlearning protocol and show that it satisfies the perfect unlearning criterion.
Ditto: Fair and Robust Federated Learning Through Personalization
Fairness and robustness are two important concerns for federated learning systems.
Overview of the Ninth Dialog System Technology Challenge: DSTC9
Interactive evaluation of dialog, and 4.
Resource Constrained Dialog Policy Learning via Differentiable Inductive Logic Programming
We explore the tasks of one-shot learning and zero-shot domain transfer with DILOG on SimDial and MultiWoZ.
Tilted Empirical Risk Minimization
Empirical risk minimization (ERM) is typically designed to perform well on the average loss, which can result in estimators that are sensitive to outliers, generalize poorly, or treat subgroups unfairly.
Competitive Balance in Team Sports Games
Competition is a primary driver of player satisfaction and engagement in multiplayer online games.
Situated and Interactive Multimodal Conversations
Next generation virtual assistants are envisioned to handle multimodal inputs (e. g., vision, memories of previous interactions, in addition to the user's utterances), and perform multimodal actions (e. g., displaying a route in addition to generating the system's utterance).
Rényi Fair Inference
In this paper, we use R\'enyi correlation as a measure of fairness of machine learning models and develop a general training framework to impose fairness.
On Multi-Agent Learning in Team Sports Games
In recent years, reinforcement learning has been successful in solving video games from Atari to Star Craft II.
Towards Interactive Training of Non-Player Characters in Video Games
We propose to create such NPC behaviors interactively by training an agent in the target environment using imitation learning with a human in the loop.
Fair Resource Allocation in Federated Learning
Federated learning involves training statistical models in massive, heterogeneous networks.
Winning Isn't Everything: Enhancing Game Development with Intelligent Agents
We discuss two fundamental metrics based on which we measure the human-likeness of agents, namely skill and style, which are multi-faceted concepts with practical implications outlined in this paper.
Information Bottleneck Methods for Distributed Learning
We consider this problem for unsupervised learning for batch and sequential data.
A Characterization of Guesswork on Swiftly Tilting Curves
We show that the tilt operation on a memoryless string-source parametrizes an exponential family of memoryless string-sources, which we refer to as the tilted family.
Information Theory Information Theory
On Data-Dependent Random Features for Improved Generalization in Supervised Learning
The randomized-feature approach has been successfully employed in large-scale kernel approximation and supervised learning.
On Optimal Generalizability in Parametric Learning
Such bias is measured by the cross validation procedure in practice where the data set is partitioned into a training set used for training and a validation set, which is not used in training and is left to measure the out-of-sample performance.
Rate-Distortion Bounds on Bayes Risk in Supervised Learning
We provide lower and upper bounds on the rate-distortion function, using $L_p$ loss as the distortion measure, of a maximum a priori classifier in terms of the differential entropy of the posterior distribution and a quantity called the interpolation dimension, which characterizes the complexity of the parametric distribution family.
Performance Trade-Offs in Multi-Processor Approximate Message Passing
We prove that the achievable region of this MOP is convex, and conjecture how the combined cost of computation and communication scales with the desired mean squared error.
