SAIL: Self-Improving Efficient Online Alignment of Large Language Models

Reinforcement Learning from Human Feedback (RLHF) is a key method for aligning large language models (LLMs) with human preferences. However, current offline alignment approaches like DPO, IPO, and SLiC rely heavily on fixed preference datasets, which can lead to sub-optimal performance. On the other hand, recent literature has focused on designing online RLHF methods but still lacks a unified conceptual formulation and suffers from distribution shift issues. To address this, we establish that online LLM alignment is underpinned by bilevel optimization. By reducing this formulation to an efficient single-level first-order method (using the reward-policy equivalence), our approach generates new samples and iteratively refines model alignment by exploring responses and regulating preference labels. In doing so, we permit alignment methods to operate in an online and self-improving manner, as well as generalize prior online RLHF methods as special cases. Compared to state-of-the-art iterative RLHF methods, our approach significantly improves alignment performance on open-sourced datasets with minimal computational overhead.

Is poisoning a real threat to LLM alignment? Maybe more so than you think

Recent advancements in Reinforcement Learning with Human Feedback (RLHF) have significantly impacted the alignment of Large Language Models (LLMs). The sensitivity of reinforcement learning algorithms such as Proximal Policy Optimization (PPO) has led to new line work on Direct Policy Optimization (DPO), which treats RLHF in a supervised learning framework. The increased practical use of these RLHF methods warrants an analysis of their vulnerabilities. In this work, we investigate the vulnerabilities of DPO to poisoning attacks under different scenarios and compare the effectiveness of preference poisoning, a first of its kind. We comprehensively analyze DPO's vulnerabilities under different types of attacks, i.e., backdoor and non-backdoor attacks, and different poisoning methods across a wide array of language models, i.e., LLama 7B, Mistral 7B, and Gemma 7B. We find that unlike PPO-based methods, which, when it comes to backdoor attacks, require at least 4\% of the data to be poisoned to elicit harmful behavior, we exploit the true vulnerabilities of DPO more simply so we can poison the model with only as much as 0.5\% of the data. We further investigate the potential reasons behind the vulnerability and how well this vulnerability translates into backdoor vs non-backdoor attacks.

DIPPER: Direct Preference Optimization to Accelerate Primitive-Enabled Hierarchical Reinforcement Learning

Learning control policies to perform complex robotics tasks from human preference data presents significant challenges. On the one hand, the complexity of such tasks typically requires learning policies to perform a variety of subtasks, then combining them to achieve the overall goal. At the same time, comprehensive, well-engineered reward functions are typically unavailable in such problems, while limited human preference data often is; making efficient use of such data to guide learning is therefore essential. Methods for learning to perform complex robotics tasks from human preference data must overcome both these challenges simultaneously. In this work, we introduce DIPPER: Direct Preference Optimization to Accelerate Primitive-Enabled Hierarchical Reinforcement Learning, an efficient hierarchical approach that leverages direct preference optimization to learn a higher-level policy and reinforcement learning to learn a lower-level policy. DIPPER enjoys improved computational efficiency due to its use of direct preference optimization instead of standard preference-based approaches such as reinforcement learning from human feedback, while it also mitigates the well-known hierarchical reinforcement learning issues of non-stationarity and infeasible subgoal generation due to our use of primitive-informed regularization inspired by a novel bi-level optimization formulation of the hierarchical reinforcement learning problem. To validate our approach, we perform extensive experimental analysis on a variety of challenging robotics tasks, demonstrating that DIPPER outperforms hierarchical and non-hierarchical baselines, while ameliorating the non-stationarity and infeasible subgoal generation issues of hierarchical reinforcement learning.

Transfer Q Star: Principled Decoding for LLM Alignment

Aligning foundation models is essential for their safe and trustworthy deployment. However, traditional fine-tuning methods are computationally intensive and require updating billions of model parameters. A promising alternative, alignment via decoding, adjusts the response distribution directly without model updates to maximize a target reward $r$, thus providing a lightweight and adaptable framework for alignment. However, principled decoding methods rely on oracle access to an optimal Q-function ($Q^*$), which is often unavailable in practice. Hence, prior SoTA methods either approximate this $Q^*$ using $Q^{\pi_{\texttt{sft}}}$ (derived from the reference $\texttt{SFT}$ model) or rely on short-term rewards, resulting in sub-optimal decoding performance. In this work, we propose Transfer $Q^*$, which implicitly estimates the optimal value function for a target reward $r$ through a baseline model $\rho_{\texttt{BL}}$ aligned with a baseline reward $\rho_{\texttt{BL}}$ (which can be different from the target reward $r$). Theoretical analyses of Transfer $Q^*$ provide a rigorous characterization of its optimality, deriving an upper bound on the sub-optimality gap and identifying a hyperparameter to control the deviation from the pre-trained reference $\texttt{SFT}$ model based on user needs. Our approach significantly reduces the sub-optimality gap observed in prior SoTA methods and demonstrates superior empirical performance across key metrics such as coherence, diversity, and quality in extensive tests on several synthetic and real datasets.

On the Safety Concerns of Deploying LLMs/VLMs in Robotics: Highlighting the Risks and Vulnerabilities

In this paper, we highlight the critical issues of robustness and safety associated with integrating large language models (LLMs) and vision-language models (VLMs) into robotics applications. Recent works have focused on using LLMs and VLMs to improve the performance of robotics tasks, such as manipulation, navigation, etc. However, such integration can introduce significant vulnerabilities, in terms of their susceptibility to adversarial attacks due to the language models, potentially leading to catastrophic consequences. By examining recent works at the interface of LLMs/VLMs and robotics, we show that it is easy to manipulate or misguide the robot's actions, leading to safety hazards. We define and provide examples of several plausible adversarial attacks, and conduct experiments on three prominent robot frameworks integrated with a language model, including KnowNo VIMA, and Instruct2Act, to assess their susceptibility to these attacks. Our empirical findings reveal a striking vulnerability of LLM/VLM-robot integrated systems: simple adversarial attacks can significantly undermine the effectiveness of LLM/VLM-robot integrated systems. Specifically, our data demonstrate an average performance deterioration of 21.2% under prompt attacks and a more alarming 30.2% under perception attacks. These results underscore the critical need for robust countermeasures to ensure the safe and reliable deployment of the advanced LLM/VLM-based robotic systems.

Provably Sample Efficient RLHF via Active Preference Optimization

Reinforcement Learning from Human Feedback (RLHF) is pivotal in aligning Large Language Models (LLMs) with human preferences. While these aligned generative models have demonstrated impressive capabilities across various tasks, the dependence on high-quality human preference data poses a costly bottleneck in practical implementation of RLHF. Hence better and adaptive strategies for data collection is needed. To this end, we frame RLHF as a contextual preference bandit problem with prompts as contexts and show that the naive way of collecting preference data by choosing prompts uniformly at random leads to a policy that suffers an $\Omega(1)$ suboptimality gap in rewards. Then we propose $\textit{Active Preference Optimization}$ ($\texttt{APO}$), an algorithm that actively selects prompts to collect preference data. Under the Bradley-Terry-Luce (BTL) preference model, \texttt{APO} achieves sample efficiency without compromising on policy performance. We show that given a sample budget of $T$, the suboptimality gap of a policy learned via $\texttt{APO}$ scales as $O(1/\sqrt{T})$. Next, we propose a compute-efficient batch version of $\texttt{APO}$ with minor modification and evaluate its performance in practice. Experimental evaluations on a human preference dataset validate \texttt{APO}'s efficacy as a sample-efficient and practical solution to data collection for RLHF, facilitating alignment of LLMs with human preferences in a cost-effective and scalable manner.

MaxMin-RLHF: Towards Equitable Alignment of Large Language Models with Diverse Human Preferences

Reinforcement Learning from Human Feedback (RLHF) aligns language models to human preferences by employing a singular reward model derived from preference data. However, such an approach overlooks the rich diversity of human preferences inherent in data collected from multiple users. In this work, we first derive an impossibility result of alignment with single reward RLHF, thereby highlighting its insufficiency in representing diverse human preferences. To provide an equitable solution to the problem, we learn a mixture of preference distributions via an expectation-maximization algorithm and propose a MaxMin alignment objective for policy learning inspired by the Egalitarian principle in social choice theory to better represent diverse human preferences. We elucidate the connection of our proposed approach to distributionally robust optimization and general utility RL, thereby highlighting the generality and robustness of our proposed solution. We present comprehensive experimental results on small-scale (GPT-2) and large-scale language models (with Tulu2-7B) and show the efficacy of the proposed approach in the presence of diversity among human preferences. Our algorithm achieves an average improvement of more than 16% in win-rates over conventional RLHF algorithms and improves the win-rate (accuracy) for minority groups by over 33% without compromising the performance of majority groups, showcasing the robustness and fairness of our approach. We remark that our findings in this work are not only limited to language models but also extend to reinforcement learning in general.

Beyond Text: Improving LLM's Decision Making for Robot Navigation via Vocal Cues

This work highlights a critical shortcoming in text-based Large Language Models (LLMs) used for human-robot interaction, demonstrating that text alone as a conversation modality falls short in such applications. While LLMs excel in processing text in these human conversations, they struggle with the nuances of verbal instructions in scenarios like social navigation, where ambiguity and uncertainty can erode trust in robotic and other AI systems. We can address this shortcoming by moving beyond text and additionally focusing on the paralinguistic features of these audio responses. These features are the aspects of spoken communication that do not involve the literal wording (lexical content) but convey meaning and nuance through how something is said. We present "Beyond Text"; an approach that improves LLM decision-making by integrating audio transcription along with a subsection of these features, which focus on the affect and more relevant in human-robot conversations. This approach not only achieves a 70.26% winning rate, outperforming existing LLMs by 48.30%, but also enhances robustness against token manipulation adversarial attacks, highlighted by a 22.44% less decrease ratio than the text-only language model in winning rate. "Beyond Text" marks an advancement in social robot navigation and broader Human-Robot interactions, seamlessly integrating text-based guidance with human-audio-informed language models.

REBEL: A Regularization-Based Solution for Reward Overoptimization in Reinforcement Learning from Human Feedback

In this work, we propose REBEL, an algorithm for sample efficient reward regularization based robotic reinforcement learning from human feedback (RRLHF). Reinforcement learning (RL) performance for continuous control robotics tasks is sensitive to the underlying reward function. In practice, the reward function often ends up misaligned with human intent, values, social norms, etc., leading to catastrophic failures in the real world. We leverage human preferences to learn regularized reward functions and eventually align the agents with the true intended behavior. We introduce a novel notion of reward regularization to the existing RRLHF framework, which is termed as agent preferences. So, we not only consider human feedback in terms of preferences, we also propose to take into account the preference of the underlying RL agent while learning the reward function. We show that this helps to improve the over-optimization associated with the design of reward functions in RL. We experimentally show that REBEL exhibits up to 70% improvement in sample efficiency to achieve a similar level of episodic reward returns as compared to the state-of-the-art methods such as PEBBLE and PEBBLE+SURF.

Towards Possibilities & Impossibilities of AI-generated Text Detection: A Survey

Large Language Models (LLMs) have revolutionized the domain of natural language processing (NLP) with remarkable capabilities of generating human-like text responses. However, despite these advancements, several works in the existing literature have raised serious concerns about the potential misuse of LLMs such as spreading misinformation, generating fake news, plagiarism in academia, and contaminating the web. To address these concerns, a consensus among the research community is to develop algorithmic solutions to detect AI-generated text. The basic idea is that whenever we can tell if the given text is either written by a human or an AI, we can utilize this information to address the above-mentioned concerns. To that end, a plethora of detection frameworks have been proposed, highlighting the possibilities of AI-generated text detection. But in parallel to the development of detection frameworks, researchers have also concentrated on designing strategies to elude detection, i.e., focusing on the impossibilities of AI-generated text detection. This is a crucial step in order to make sure the detection frameworks are robust enough and it is not too easy to fool a detector. Despite the huge interest and the flurry of research in this domain, the community currently lacks a comprehensive analysis of recent developments. In this survey, we aim to provide a concise categorization and overview of current work encompassing both the prospects and the limitations of AI-generated text detection. To enrich the collective knowledge, we engage in an exhaustive discussion on critical and challenging open questions related to ongoing research on AI-generated text detection.