Search

Alignment of pretrained LLMs using instruction-based datasets is critical for creating fine-tuned models that reflect human preference. A growing number of alignment-based fine-tuning algorithms and benchmarks emerged recently, fueling the efforts on effective alignments of pre-trained LLMs to ensure helpful, harmless, and honest answers from both open-source and closed-source LLMs. This paper tackles this problem by developing an alignment fusion approach, coined as $H^3$Fusion, with three unique characteristics. First, $H^3$Fusion ensembles multiple individually aligned LLMs to create a final fine-tuned alignment model with enhanced capabilities beyond those of individual models, delivering robust alignment through promoting helpful, harmless, honest fusion. Second, $H^3$Fusion leverages the mixture-of-experts (MoE) methodology in two steps. We first freeze the multi-head attention weights of each individual model while tuning the FFN layer during alignment fusion. Then we merge the aligned model weights with an expert router according to the type of input instruction and dynamically select a subset of experts that are best suited for producing the output response. Finally, we boost the performance of the resulting $H^3$3Fusion model by introducing gating loss and regularization terms. The former penalizes the selection errors of the expert-router, and the latter mediates the expert weights drifting during fine-tuning and dynamically adjusts the fusion behavior of the resulting model by canalizing the activations on the experts. Extensive evaluations on three benchmark datasets show that $H^3$3Fusion is more helpful, less harmful, and more honest from two aspects: it outperforms each individually aligned model by $11.37\%$, and it provides stronger robustness compared to the state-of-the-art LLM ensemble approaches by $13.77\%$. Code is available at github.com/sftekin/h3fusion.

Electrochemical processes play a crucial role in energy storage and conversion systems, yet their computational modeling remains a significant challenge. Accurately incorporating the effects of electric potential has been a central focus in theoretical electrochemistry. Although constant-potential ab initio molecular dynamics (CP-AIMD) has provided valuable insights, it is limited by its substantial computational demands. Here, we introduce the Explicit Electric Potential Machine Learning Force Field (EEP-MLFF) model. Our model integrates the electric potential as an explicit input parameter along with the atom-centered descriptors in the atomic neural network. This approach enables the evaluation of nuclear forces under arbitrary electric potentials, thus facilitating molecular dynamics simulations at a specific potential. By applying the proposed machine learning method to the Cu/1T$^{\prime}$-MoS$_{2}$ system, molecular dynamics simulations reveal that the potential-modulated Cu atom migration and aggregation lead to the formation of small steric Cu clusters (Single Clusters, SCs) at potentials below -0.1 V. The morphological transformations of adsorbed Cu atoms are elucidated through electronic structure analyses, which demonstrates that both Cu-S and Cu-Cu bonding can be effectively tuned by the applied electric potential. Our findings present an opportunity for the convenient manufacture of single metal cluster catalysts through potential modulation. Moreover, this theoretical framework facilitates the exploration of potential-regulated processes and helps investigate the mechanisms of electrochemical reactions., Comment: 17 pages, 6 figures

Artificial intelligence (AI) has significantly advanced computational chemistry research. However, traditional AI methods often rely on task-specific model designs and training, which constrain both the scalability of model size and generalization across different tasks. Here, we introduce ChemFM, a large-scale foundation model specifically developed for chemistry, comprising up to 3 billion parameters. ChemFM is pre-trained on 178 million molecules using self-supervised causal language modeling to extract generalizable molecular representations. This model can be adapted to diverse downstream chemical applications using both full-parameter and parameter-efficient fine-tuning methods. ChemFM consistently outperforms state-of-the-art approaches across multiple chemical tasks. Notably, it achieves up to 67.48% performance improvement across 34 property prediction benchmarks, up to 33.31% reduction in mean average deviation between conditioned and actual properties of generated molecules in conditional molecular generation tasks, and up to 3.7% top-1 accuracy improvement across 4 reaction prediction datasets. Moreover, ChemFM demonstrates superior performance in predicting antibiotic activity and cytotoxicity, highlighting its potential to advance the discovery of novel antibiotics. We anticipate that ChemFM will significantly advance chemistry research by providing a foundation model capable of effectively generalizing across a broad range of tasks with minimal additional training.

The safety alignment ability of Vision-Language Models (VLMs) is prone to be degraded by the integration of the vision module compared to its LLM backbone. We investigate this phenomenon, dubbed as ''safety alignment degradation'' in this paper, and show that the challenge arises from the representation gap that emerges when introducing vision modality to VLMs. In particular, we show that the representations of multi-modal inputs shift away from that of text-only inputs which represent the distribution that the LLM backbone is optimized for. At the same time, the safety alignment capabilities, initially developed within the textual embedding space, do not successfully transfer to this new multi-modal representation space. To reduce safety alignment degradation, we introduce Cross-Modality Representation Manipulation (CMRM), an inference time representation intervention method for recovering the safety alignment ability that is inherent in the LLM backbone of VLMs, while simultaneously preserving the functional capabilities of VLMs. The empirical results show that our framework significantly recovers the alignment ability that is inherited from the LLM backbone with minimal impact on the fluency and linguistic capabilities of pre-trained VLMs even without additional training. Specifically, the unsafe rate of LLaVA-7B on multi-modal input can be reduced from 61.53% to as low as 3.15% with only inference-time intervention. WARNING: This paper contains examples of toxic or harmful language., Comment: Preprint

Combining large language models during training or at inference time has shown substantial performance gain over component LLMs. This paper presents LLM-TOPLA, a diversity-optimized LLM ensemble method with three unique properties: (i) We introduce the focal diversity metric to capture the diversity-performance correlation among component LLMs of an ensemble. (ii) We develop a diversity-optimized ensemble pruning algorithm to select the top-k sub-ensembles from a pool of $N$ base LLMs. Our pruning method recommends top-performing LLM subensembles of size $S$, often much smaller than $N$. (iii) We generate new output for each prompt query by utilizing a learn-to-ensemble approach, which learns to detect and resolve the output inconsistency among all component LLMs of an ensemble. Extensive evaluation on four different benchmarks shows good performance gain over the best LLM ensemble methods: (i) In constrained solution set problems, LLM-TOPLA outperforms the best-performing ensemble (Mixtral) by 2.2\% in accuracy on MMLU and the best-performing LLM ensemble (MoreAgent) on GSM8k by 2.1\%. (ii) In generative tasks, LLM-TOPLA outperforms the top-2 performers (Llama70b/Mixtral) on SearchQA by $3.9\mathrm{x}$ in F1, and on XSum by more than $38$ in ROUGE-1. Our code and dataset, which contains outputs of 8 modern LLMs on 4 benchmarks is available at https://github.com/git-disl/llm-topla

The effectiveness of automatic evaluation of generative models is typically measured by comparing it to human evaluation using correlation metrics. However, metrics like Krippendorff's $\alpha$ and Randolph's $\kappa$, originally designed to measure the reliability of human labeling, make assumptions about human behavior and the labeling process. In this paper, we show how *relying on a single aggregate correlation score* can obscure fundamental differences between human behavior and automatic evaluation methods, including LLM-as-a-Judge. Specifically, we demonstrate that when the proportion of samples with variation or uncertainty in human labels (gathered during human evaluation) is relatively high, machine labels (generated by automatic evaluation methods) may superficially appear to have similar or better correlation with the human majority label compared to human-to-human (HH) correlation. This can create the illusion that automatic evaluation approximates the human majority label. However, as the proportion of samples with consistent human labels increases, the correlation between machine and human labels fall well below HH correlation. Based on these findings, we first propose stratifying results by human label uncertainty to provide a more robust analysis of automatic evaluation performance. Second, recognizing that uncertainty and variation are inherent in perception-based human evaluations, such as those involving attitudes or preferences, we introduce a new metric - *binned Jensen-Shannon Divergence for perception* for such scenarios to better measure the effectiveness of automatic evaluations. Third, we present visualization techniques -- *perception charts*, to compare the strengths and limitations of automatic evaluation and to contextualize correlation measures appropriately

Recent research demonstrates that the nascent fine-tuning-as-a-service business model exposes serious safety concerns -- fine-tuning over a few harmful data uploaded by the users can compromise the safety alignment of the model. The attack, known as harmful fine-tuning attack, has raised a broad research interest among the community. However, as the attack is still new, \textbf{we observe that there are general misunderstandings within the research community.} To clear up concern, this paper provide a comprehensive overview to three aspects of harmful fine-tuning: attacks setting, defense design and evaluation methodology. Specifically, we first present the threat model of the problem, and introduce the harmful fine-tuning attack and its variants. Then we systematically survey the existing literature on attacks/defenses/mechanical analysis of the problem. Finally, we introduce the evaluation methodology and outline future research directions that might contribute to the development of the field. Additionally, we present a list of questions of interest, which might be useful to refer to when reviewers in the peer review process question the realism of the experiment/attack/defense setting. A curated list of relevant papers is maintained and made accessible at: https://github.com/git-disl/awesome_LLM-harmful-fine-tuning-papers.

Data poisoning and leakage risks impede the massive deployment of federated learning in the real world. This chapter reveals the truths and pitfalls of understanding two dominating threats: {\em training data privacy intrusion} and {\em training data poisoning}. We first investigate training data privacy threat and present our observations on when and how training data may be leaked during the course of federated training. One promising defense strategy is to perturb the raw gradient update by adding some controlled randomized noise prior to sharing during each round of federated learning. We discuss the importance of determining the proper amount of randomized noise and the proper location to add such noise for effective mitigation of gradient leakage threats against training data privacy. Then we will review and compare different training data poisoning threats and analyze why and when such data poisoning induced model Trojan attacks may lead to detrimental damage on the performance of the global model. We will categorize and compare representative poisoning attacks and the effectiveness of their mitigation techniques, delivering an in-depth understanding of the negative impact of data poisoning. Finally, we demonstrate the potential of dynamic model perturbation in simultaneously ensuring privacy protection, poisoning resilience, and model performance. The chapter concludes with a discussion on additional risk factors in federated learning, including the negative impact of skewness, data and algorithmic biases, as well as misinformation in training data. Powered by empirical evidence, our analytical study offers some transformative insights into effective privacy protection and security assurance strategies in attack-resilient federated learning., Comment: Chapter of Handbook of Trustworthy Federated Learning

Harmful fine-tuning issue \citep{qi2023fine} poses serious safety concerns for Large language models' fine-tuning-as-a-service. While existing defenses \citep{huang2024vaccine,rosati2024representation} have been proposed to mitigate the issue, their performances are still far away from satisfactory, and the root cause of the problem has not been fully recovered. For the first time in the literature, we in this paper show that \textit{harmful perturbation} over the model weights should be the root cause of alignment-broken of harmful fine-tuning. In order to attenuate the negative impact of harmful perturbation, we propose an alignment-stage solution, dubbed Booster. Technically, along with the original alignment loss, we append a loss regularizer in the alignment stage's optimization. The regularizer ensures that the model's harmful loss reduction before/after simulated harmful perturbation is attenuated, thereby mitigating the subsequent fine-tuning risk. Empirical results show that Booster can effectively reduce the harmful score of the fine-tuned models while maintaining the performance of downstream tasks. Our code is available at \url{https://github.com/git-disl/Booster}.

Safety aligned Large Language Models (LLMs) are vulnerable to harmful fine-tuning attacks \cite{qi2023fine}-- a few harmful data mixed in the fine-tuning dataset can break the LLMs's safety alignment. Existing mitigation strategies include alignment stage solutions \cite{huang2024vaccine, rosati2024representation} and fine-tuning stage solutions \cite{huang2024lazy,mukhoti2023fine}. However, our evaluation shows that both categories of defenses fail \textit{when some specific training hyper-parameters are chosen} -- a large learning rate or a large number of training epochs in the fine-tuning stage can easily invalidate the defense, which however, is necessary to guarantee finetune performance. To this end, we propose Antidote, a post-fine-tuning stage solution, which remains \textbf{\textit{agnostic to the training hyper-parameters in the fine-tuning stage}}. Antidote relies on the philosophy that by removing the harmful parameters, the harmful model can be recovered from the harmful behaviors, regardless of how those harmful parameters are formed in the fine-tuning stage. With this philosophy, we introduce a one-shot pruning stage after harmful fine-tuning to remove the harmful weights that are responsible for the generation of harmful content. Despite its embarrassing simplicity, empirical results show that Antidote can reduce harmful score while maintaining accuracy on downstream tasks.Our project page is at \url{https://huangtiansheng.github.io/Antidote_gh_page/}

Face recognition (FR) can be abused for privacy intrusion. Governments, private companies, or even individual attackers can collect facial images by web scraping to build an FR system identifying human faces without their consent. This paper introduces Chameleon, which learns to generate a user-centric personalized privacy protection mask, coined as P3-Mask, to protect facial images against unauthorized FR with three salient features. First, we use a cross-image optimization to generate one P3-Mask for each user instead of tailoring facial perturbation for each facial image of a user. It enables efficient and instant protection even for users with limited computing resources. Second, we incorporate a perceptibility optimization to preserve the visual quality of the protected facial images. Third, we strengthen the robustness of P3-Mask against unknown FR models by integrating focal diversity-optimized ensemble learning into the mask generation process. Extensive experiments on two benchmark datasets show that Chameleon outperforms three state-of-the-art methods with instant protection and minimal degradation of image quality. Furthermore, Chameleon enables cost-effective FR authorization using the P3-Mask as a personalized de-obfuscation key, and it demonstrates high resilience against adaptive adversaries., Comment: ECCV 2024

Recent studies show that Large Language Models (LLMs) with safety alignment can be jail-broken by fine-tuning on a dataset mixed with harmful data. First time in the literature, we show that the jail-broken effect can be mitigated by separating states in the finetuning stage to optimize the alignment and user datasets. Unfortunately, our subsequent study shows that this simple Bi-State Optimization (BSO) solution experiences convergence instability when steps invested in its alignment state is too small, leading to downgraded alignment performance. By statistical analysis, we show that the \textit{excess drift} towards consensus could be a probable reason for the instability. To remedy this issue, we propose \textbf{L}azy(\textbf{i}) \textbf{s}afety \textbf{a}lignment (\textbf{Lisa}), which introduces a proximal term to constraint the drift of each state. Theoretically, the benefit of the proximal term is supported by the convergence analysis, wherein we show that a sufficient large proximal factor is necessary to guarantee Lisa's convergence. Empirically, our results on four downstream finetuning tasks show that Lisa with a proximal term can significantly increase alignment performance while maintaining the LLM's accuracy on the user tasks. Code is available at \url{https://github.com/git-disl/Lisa}., Comment: Accepted by NeurIPS2024. arXiv admin note: substantial text overlap with arXiv:2402.01109

In this position paper, we argue that human evaluation of generative large language models (LLMs) should be a multidisciplinary undertaking that draws upon insights from disciplines such as user experience research and human behavioral psychology to ensure that the experimental design and results are reliable. The conclusions from these evaluations, thus, must consider factors such as usability, aesthetics, and cognitive biases. We highlight how cognitive biases can conflate fluent information and truthfulness, and how cognitive uncertainty affects the reliability of rating scores such as Likert. Furthermore, the evaluation should differentiate the capabilities and weaknesses of increasingly powerful large language models -- which requires effective test sets. The scalability of human evaluation is also crucial to wider adoption. Hence, to design an effective human evaluation system in the age of generative NLP, we propose the ConSiDERS-The-Human evaluation framework consisting of 6 pillars -- Consistency, Scoring Criteria, Differentiating, User Experience, Responsible, and Scalability., Comment: Accepted in ACL 2024

In this work, we prove that polar lattices, when tailored for lossy compression, are quantization-good in the sense that their normalized second moments approach $\frac{1}{2\pi e}$ as the dimension of lattices increases. It has been predicted by Zamir et al. \cite{ZamirQZ96} that the Entropy Coded Dithered Quantization (ECDQ) system using quantization-good lattices can achieve the rate-distortion bound of i.i.d. Gaussian sources. In our previous work \cite{LingQZ}, we established that polar lattices are indeed capable of attaining the same objective. It is reasonable to conjecture that polar lattices also demonstrate quantization goodness in the context of lossy compression. This study confirms this hypothesis., Comment: 12 pages, 5 figures, submitted to IEEE for possible publication

To address the challenge of constructing short girth-8 quasi-cyclic (QC) low-density parity-check (LDPC) codes, a novel construction framework based on vertical symmetry (VS) is proposed. Basic properties of the VS structure are presented. With the aid of these properties, existing explicit constructions for column weights from three to five which can be transformed into the VS structure are sorted out. Then two novel explicit constructions with the VS structure which guarantee short codes are presented for column weights of three and six. Moreover, an efficient search-based method is also proposed to find short codes with the VS structure. Compared with the state-of-the-art benchmarks, both the explicit constructions and the search-based method presented in this paper can provide shorter codes for most cases. Simulation results show that the new shorter codes can perform almost the same as or better than the longer existing counterparts. Thus, the new shorter codes can fit better with the low-latency requirement for modern communication systems., Comment: 17 pages, 5 figures; This paper has been accepted by IEEE ISIT2024

Graph pattern matching, one of the fundamental graph mining problems, aims to extract structural patterns of interest from an input graph. The state-of-the-art graph matching algorithms and systems are mainly designed for undirected graphs. Directed graph matching is more complex than undirected graph matching because the edge direction must be taken into account before the exploration of each directed edge. Thus, the technologies (e.g. storage, exploiting symmetry for graph matching) for undirected graph matching may not be fully applicable to directed graphs. For example, the redundancy implied in directed graph pattern can not be detected using the symmetry breaking for undirected pattern graph. Here, we present XMiner for efficient directed graph pattern matching whose core idea is 'pattern reduction'. It first analyzes the relationship between constraints implied in a pattern digraph. Then it reduces the pattern graph into a simplified form by finding a minimum constraint cover. Finally, XMiner generates an execution plan and follows it to extract matchings of the pattern graph. So, XMiner works on simplified pattern graph and avoids much data access and redundant computation throughout the matching process. Our experimental results show that XMiner outperforms state-of the-art stand-alone graph matching systems, and scales to complex graph pattern matching tasks on larger graph.

This paper presents FusionShot, a focal diversity optimized few-shot ensemble learning approach for boosting the robustness and generalization performance of pre-trained few-shot models. The paper makes three original contributions. First, we explore the unique characteristics of few-shot learning to ensemble multiple few-shot (FS) models by creating three alternative fusion channels. Second, we introduce the concept of focal error diversity to learn the most efficient ensemble teaming strategy, rather than assuming that an ensemble of a larger number of base models will outperform those sub-ensembles of smaller size. We develop a focal-diversity ensemble pruning method to effectively prune out the candidate ensembles with low ensemble error diversity and recommend top-$K$ FS ensembles with the highest focal error diversity. Finally, we capture the complex non-linear patterns of ensemble few-shot predictions by designing the learn-to-combine algorithm, which can learn the diverse weight assignments for robust ensemble fusion over different member models. Extensive experiments on representative few-shot benchmarks show that the top-K ensembles recommended by FusionShot can outperform the representative SOTA few-shot models on novel tasks (different distributions and unknown at training), and can prevail over existing few-shot learners in both cross-domain settings and adversarial settings. For reproducibility purposes, FusionShot trained models, results, and code are made available at https://github.com/sftekin/fusionshot

The development of game agents holds a critical role in advancing towards Artificial General Intelligence (AGI). The progress of LLMs and their multimodal counterparts (MLLMs) offers an unprecedented opportunity to evolve and empower game agents with human-like decision-making capabilities in complex computer game environments. This paper provides a comprehensive overview of LLM-based game agents from a holistic viewpoint. First, we introduce the conceptual architecture of LLM-based game agents, centered around six essential functional components: perception, memory, thinking, role-playing, action, and learning. Second, we survey existing representative LLM-based game agents documented in the literature with respect to methodologies and adaptation agility across six genres of games, including adventure, communication, competition, cooperation, simulation, and crafting & exploration games. Finally, we present an outlook of future research and development directions in this burgeoning field. A curated list of relevant papers is maintained and made accessible at: https://github.com/git-disl/awesome-LLM-game-agent-papers.

We introduce PokeLLMon, the first LLM-embodied agent that achieves human-parity performance in tactical battle games, as demonstrated in Pokemon battles. The design of PokeLLMon incorporates three key strategies: (i) In-context reinforcement learning that instantly consumes text-based feedback derived from battles to iteratively refine the policy; (ii) Knowledge-augmented generation that retrieves external knowledge to counteract hallucination and enables the agent to act timely and properly; (iii) Consistent action generation to mitigate the panic switching phenomenon when the agent faces a powerful opponent and wants to elude the battle. We show that online battles against human demonstrates PokeLLMon's human-like battle strategies and just-in-time decision making, achieving 49% of win rate in the Ladder competitions and 56% of win rate in the invited battles. Our implementation and playable battle logs are available at: https://github.com/git-disl/PokeLLMon., Comment: 10 pages

The new paradigm of finetuning-as-a-service introduces a new attack surface for Large Language Models (LLMs): a few harmful data uploaded by users can easily trick the finetuning to produce an alignment-broken model. We conduct an empirical analysis and uncover a \textit{harmful embedding drift} phenomenon, showing a probable cause of the alignment-broken effect. Inspired by our findings, we propose Vaccine, a perturbation-aware alignment technique to mitigate the security risk of users finetuning. The core idea of Vaccine is to produce invariant hidden embeddings by progressively adding crafted perturbation to them in the alignment phase. This enables the embeddings to withstand harmful perturbation from un-sanitized user data in the finetuning phase. Our results on open source mainstream LLMs (e.g., Llama2, Opt, Vicuna) demonstrate that Vaccine can boost the robustness of alignment against harmful prompts induced embedding drift while reserving reasoning ability towards benign prompts. Our code is available at \url{https://github.com/git-disl/Vaccine}., Comment: Rejected by ICML2024. Accepted by NeurIPS2024

Emerging Distributed AI systems are revolutionizing big data computing and data processing capabilities with growing economic and societal impact. However, recent studies have identified new attack surfaces and risks caused by security, privacy, and fairness issues in AI systems. In this paper, we review representative techniques, algorithms, and theoretical foundations for trustworthy distributed AI through robustness guarantee, privacy protection, and fairness awareness in distributed learning. We first provide a brief overview of alternative architectures for distributed learning, discuss inherent vulnerabilities for security, privacy, and fairness of AI algorithms in distributed learning, and analyze why these problems are present in distributed learning regardless of specific architectures. Then we provide a unique taxonomy of countermeasures for trustworthy distributed AI, covering (1) robustness to evasion attacks and irregular queries at inference, and robustness to poisoning attacks, Byzantine attacks, and irregular data distribution during training; (2) privacy protection during distributed learning and model inference at deployment; and (3) AI fairness and governance with respect to both data and models. We conclude with a discussion on open challenges and future research directions toward trustworthy distributed AI, such as the need for trustworthy AI policy guidelines, the AI responsibility-utility co-design, and incentives and compliance., Comment: Manuscript accepted to ACM Computing Surveys

In this paper, we first establish the separation theorem between a point and a locally geodesic convex set and then prove the existence of a supporting quasi-hyperplane at any point on the boundary of the closed locally geodesic convex set on a Riemannian manifold. As applications, some optimality conditions are obtained for optimization problems with constraints on Riemannian manifolds.

Humans commonly work with multiple objects in daily life and can intuitively transfer manipulation skills to novel objects by understanding object functional regularities. However, existing technical approaches for analyzing and synthesizing hand-object manipulation are mostly limited to handling a single hand and object due to the lack of data support. To address this, we construct TACO, an extensive bimanual hand-object-interaction dataset spanning a large variety of tool-action-object compositions for daily human activities. TACO contains 2.5K motion sequences paired with third-person and egocentric views, precise hand-object 3D meshes, and action labels. To rapidly expand the data scale, we present a fully automatic data acquisition pipeline combining multi-view sensing with an optical motion capture system. With the vast research fields provided by TACO, we benchmark three generalizable hand-object-interaction tasks: compositional action recognition, generalizable hand-object motion forecasting, and cooperative grasp synthesis. Extensive experiments reveal new insights, challenges, and opportunities for advancing the studies of generalizable hand-object motion analysis and synthesis. Our data and code are available at https://taco2024.github.io.