Jessica Hoffmann

Maximilian Mozes, Jessica Hoffmann, Katrin Tomanek et al.

Text-based safety classifiers are widely used for content moderation and increasingly to tune generative language model behavior - a topic of growing concern for the safety of digital assistants and chatbots. However, different policies require different classifiers, and safety policies themselves improve from iteration and adaptation. This paper introduces and evaluates methods for agile text classification, whereby classifiers are trained using small, targeted datasets that can be quickly developed for a particular policy. Experimenting with 7 datasets from three safety-related domains, comprising 15 annotation schemes, led to our key finding: prompt-tuning large language models, like PaLM 62B, with a labeled dataset of as few as 80 examples can achieve state-of-the-art performance. We argue that this enables a paradigm shift for text classification, especially for models supporting safer online discourse. Instead of collecting millions of examples to attempt to create universal safety classifiers over months or years, classifiers could be tuned using small datasets, created by individuals or small organizations, tailored for specific use cases, and iterated on and adapted in the time-span of a day.

Ann Yuan, Asma Ghandeharioun, Carter Blum et al.

While existing evaluations of large language models (LLMs) measure deception rates, the underlying conditions that give rise to deceptive behavior are poorly understood. We investigate this question using a novel dataset of realistic moral trade-offs where honesty incurs variable costs. Contrary to humans, who tend to become less honest given time to deliberate (Capraro, 2017; Capraro et al., 2019), we find that reasoning consistently increases honesty across scales and for several LLM families. This effect is not only a function of the reasoning content, as reasoning traces are often poor predictors of final behaviors. Rather, we show that the underlying geometry of the representational space itself contributes to the effect. Namely, we observe that deceptive regions within this space are metastable: deceptive answers are more easily destabilized by input paraphrasing, output resampling, and activation noise than honest ones. We interpret the effect of reasoning in this vein: generating deliberative tokens as part of moral reasoning entails the traversal of a biased representational space, ultimately nudging the model toward its more stable, honest defaults.

Tianlin Liu, Shangmin Guo, Leonardo Bianco et al.

Aligning language models with human preferences is crucial for reducing errors and biases in these models. Alignment techniques, such as reinforcement learning from human feedback (RLHF), are typically cast as optimizing a tradeoff between human preference rewards and a proximity regularization term that encourages staying close to the unaligned model. Selecting an appropriate level of regularization is critical: insufficient regularization can lead to reduced model capabilities due to reward hacking, whereas excessive regularization hinders alignment. Traditional methods for finding the optimal regularization level require retraining multiple models with varying regularization strengths. This process, however, is resource-intensive, especially for large models. To address this challenge, we propose decoding-time realignment (DeRa), a simple method to explore and evaluate different regularization strengths in aligned models without retraining. DeRa enables control over the degree of alignment, allowing users to smoothly transition between unaligned and aligned models. It also enhances the efficiency of hyperparameter tuning by enabling the identification of effective regularization strengths using a validation dataset.

Tyler A. Chang, Katrin Tomanek, Jessica Hoffmann et al.

We explore a strategy to handle controversial topics in LLM-based chatbots based on Wikipedia's Neutral Point of View (NPOV) principle: acknowledge the absence of a single true answer and surface multiple perspectives. We frame this as retrieval augmented generation, where perspectives are retrieved from a knowledge base and the LLM is tasked with generating a fluent and faithful response from the given perspectives. As a starting point, we use a deterministic retrieval system and then focus on common LLM failure modes that arise during this approach to text generation, namely hallucination and coverage errors. We propose and evaluate three methods to detect such errors based on (1) word-overlap, (2) salience, and (3) LLM-based classifiers. Our results demonstrate that LLM-based classifiers, even when trained only on synthetic errors, achieve high error detection performance, with ROC AUC scores of 95.3% for hallucination and 90.5% for coverage error detection on unambiguous error cases. We show that when no training data is available, our other methods still yield good results on hallucination (84.0%) and coverage error (85.2%) detection.

Hakim Sidahmed, Samrat Phatale, Alex Hutcheson et al. · deepmind

While Reinforcement Learning from Human Feedback (RLHF) effectively aligns pretrained Large Language and Vision-Language Models (LLMs, and VLMs) with human preferences, its computational cost and complexity hamper its wider adoption. To alleviate some of the computational burden of fine-tuning, parameter efficient methods, like LoRA were introduced. In this work, we empirically evaluate the setup of Parameter Efficient Reinforcement Learning from Human Feedback (PE-RLHF) that leverages LoRA fine-tuning for Reward Modeling, and Reinforcement Learning. We benchmark the PE-RLHF setup on six diverse datasets spanning summarization, harmless/helpful response generation, UI automation, and visual question answering in terms of effectiveness of the trained models, and the training resources required. Our findings show, for the first time, that PE-RLHF achieves comparable performance to RLHF, while significantly reducing training time (up to 90% faster for reward models, and 30% faster for RL), and memory footprint (up to 50% reduction for reward models, and 27% for RL). We provide comprehensive ablations across LoRA ranks, and model sizes for both reward modeling and reinforcement learning. By mitigating the computational burden associated with RLHF, we push for a broader adoption of PE-RLHF as an alignment technique for LLMs and VLMs.

Carter Blum, Katja Filippova, Ann Yuan et al. · deepmind

Large language models (LLMs) struggle with cross-lingual knowledge transfer: they hallucinate when asked in one language about facts expressed in a different language during training. This work introduces a controlled setting to study the causes and dynamics of this phenomenon by training small Transformer models from scratch on synthetic multilingual datasets. We identify a learning phase wherein a model develops either separate or unified representations of the same facts across languages, and show that unification is essential for cross-lingual transfer. We also show that the degree of unification depends on mutual information between facts and training data language, and on how easy it is to extract that language. Based on these insights, we develop methods to modulate the level of cross-lingual transfer by manipulating data distribution and tokenization, and we introduce metrics and visualizations to formally characterize their effects on unification. Our work shows how controlled settings can shed light on pre-training dynamics and suggests new directions for improving cross-lingual transfer in LLMs.

Jessica Hoffmann, Christiane Ahlheim, Zac Yu et al.

The paper shows that parameter-efficient reinforcement learning (PE-RL) is a highly effective training regime to improve large language models' (LLMs) ability to answer queries on sensitive topics with a Neutral Point of View (NPOV), i.e. to provide significantly more informative, diverse and impartial answers. This is shown by evaluating PE-RL and multiple strong baselines-including LoRA finetuning (strongest baseline), SFT and RLHF. PE-RL not only improves on overall NPOV quality compared to the strongest baseline ($97.06\%\rightarrow 99.08\%$), but also scores much higher on features linguists identify as key to separating sufficient answers from "great'' answers ($60.25\%\rightarrow 85.21\%$ for presence of supportive details, $68.74\%\rightarrow 91.43\%$ for absence of oversimplification). A qualitative analysis corroborates this. Moreover, our evaluation also finds a key property of PE-RL for this task: unlike methods that update all parameters, it generalises out of topic. Finally, to enable further studies we also release the dataset, SHQ-NPOV, and provide a methodology to create such datasets through iterative rounds of human peer-critique and annotator training.

Matthew Faw, Constantine Caramanis, Jessica Hoffmann

Unconscious bias has been shown to influence how we assess our peers, with consequences for hiring, promotions and admissions. In this work, we focus on affinity bias, the component of unconscious bias which leads us to prefer people who are similar to us, despite no deliberate intention of favoritism. In a world where the people hired today become part of the hiring committee of tomorrow, we are particularly interested in understanding (and mitigating) how affinity bias affects this feedback loop. This problem has two distinctive features: 1) we only observe the biased value of a candidate, but we want to optimize with respect to their real value 2) the bias towards a candidate with a specific set of traits depends on the fraction of people in the hiring committee with the same set of traits. We introduce a new bandits variant that exhibits those two features, which we call affinity bandits. Unsurprisingly, classical algorithms such as UCB often fail to identify the best arm in this setting. We prove a new instance-dependent regret lower bound, which is larger than that in the standard bandit setting by a multiplicative function of $K$. Since we treat rewards that are time-varying and dependent on the policy's past actions, deriving this lower bound requires developing proof techniques beyond the standard bandit techniques. Finally, we design an elimination-style algorithm which nearly matches this regret, despite never observing the real rewards.

Ajil Jalal, Sushrut Karmalkar, Jessica Hoffmann et al.

This work tackles the issue of fairness in the context of generative procedures, such as image super-resolution, which entail different definitions from the standard classification setting. Moreover, while traditional group fairness definitions are typically defined with respect to specified protected groups -- camouflaging the fact that these groupings are artificial and carry historical and political motivations -- we emphasize that there are no ground truth identities. For instance, should South and East Asians be viewed as a single group or separate groups? Should we consider one race as a whole or further split by gender? Choosing which groups are valid and who belongs in them is an impossible dilemma and being "fair" with respect to Asians may require being "unfair" with respect to South Asians. This motivates the introduction of definitions that allow algorithms to be \emph{oblivious} to the relevant groupings. We define several intuitive notions of group fairness and study their incompatibilities and trade-offs. We show that the natural extension of demographic parity is strongly dependent on the grouping, and \emph{impossible} to achieve obliviously. On the other hand, the conceptually new definition we introduce, Conditional Proportional Representation, can be achieved obliviously through Posterior Sampling. Our experiments validate our theoretical results and achieve fair image reconstruction using state-of-the-art generative models.

Moein Khajehnejad, Ahmad Asgharian Rezaei, Mahmoudreza Babaei et al.

Influence maximization is a widely studied topic in network science, where the aim is to reach the maximum possible number of nodes, while only targeting a small initial set of individuals. It has critical applications in many fields, including viral marketing, information propagation, news dissemination, and vaccinations. However, the objective does not usually take into account whether the final set of influenced nodes is fair with respect to sensitive attributes, such as race or gender. Here we address fair influence maximization, aiming to reach minorities more equitably. We introduce Adversarial Graph Embeddings: we co-train an auto-encoder for graph embedding and a discriminator to discern sensitive attributes. This leads to embeddings which are similarly distributed across sensitive attributes. We then find a good initial set by clustering the embeddings. We believe we are the first to use embeddings for the task of fair influence maximization. While there are typically trade-offs between fairness and influence maximization objectives, our experiments on synthetic and real-world datasets show that our approach dramatically reduces disparity while remaining competitive with state-of-the-art influence maximization methods.

Jessica Hoffmann, Soumya Basu, Surbhi Goel et al.

We consider the problem of learning the weighted edges of a balanced mixture of two undirected graphs from epidemic cascades. While mixture models are popular modeling tools, algorithmic development with rigorous guarantees has lagged. Graph mixtures are apparently no exception: until now, very little is known about whether this problem is solvable. To the best of our knowledge, we establish the first necessary and sufficient conditions for this problem to be solvable in polynomial time on edge-separated graphs. When the conditions are met, i.e., when the graphs are connected with at least three edges, we give an efficient algorithm for learning the weights of both graphs with optimal sample complexity (up to log factors). We give complimentary results and provide sample-optimal (up to log factors) algorithms for mixtures of directed graphs of out-degree at least three, for mixture of undirected graphs of unbalanced and/or unknown priors.

Jessica Hoffmann, Constantine Caramanis

We consider the problem of learning the weighted edges of a graph by observing the noisy times of infection for multiple epidemic cascades on this graph. Past work has considered this problem when the cascade information, i.e., infection times, are known exactly. Though the noisy setting is well motivated by many epidemic processes (e.g., most human epidemics), to the best of our knowledge, very little is known about when it is solvable. Previous work on the no-noise setting critically uses the ordering information. If noise can reverse this -- a node's reported (noisy) infection time comes after the reported infection time of some node it infected -- then we are unable to see how previous results can be extended. We therefore tackle two versions of the noisy setting: the limited-noise setting, where we know noisy times of infections, and the extreme-noise setting, in which we only know whether or not a node was infected. We provide a polynomial time algorithm for recovering the structure of bidirectional trees in the extreme-noise setting, and show our algorithm almost matches lower bounds established in the no-noise setting, and hence is optimal up to log-factors. We extend our results for general degree-bounded graphs, where again we show that our (poly-time) algorithm can recover the structure of the graph with optimal sample complexity. We also provide the first efficient algorithm to learn the weights of the bidirectional tree in the limited-noise setting. Finally, we give a polynomial time algorithm for learning the weights of general bounded-degree graphs in the limited-noise setting. This algorithm extends to general graphs (at the price of exponential running time), proving the problem is solvable in the general case. All our algorithms work for any noise distribution, without any restriction on the variance.

Ashish Katiyar, Jessica Hoffmann, Constantine Caramanis

Consider jointly Gaussian random variables whose conditional independence structure is specified by a graphical model. If we observe realizations of the variables, we can compute the covariance matrix, and it is well known that the support of the inverse covariance matrix corresponds to the edges of the graphical model. Instead, suppose we only have noisy observations. If the noise at each node is independent, we can compute the sum of the covariance matrix and an unknown diagonal. The inverse of this sum is (in general) dense. We ask: can the original independence structure be recovered? We address this question for tree structured graphical models. We prove that this problem is unidentifiable, but show that this unidentifiability is limited to a small class of candidate trees. We further present additional constraints under which the problem is identifiable. Finally, we provide an O(n^3) algorithm to find this equivalence class of trees.