Existing benchmarks do not test Large Multimodal Models (LMMs) on their interactive intelligence with human users, which is vital for developing generalpurpose AI assistants. We design InterFeedback, an interactive framework, which can be applied to any LMM and dataset to assess this ability autonomously. On top of this, we introduce InterFeedback-Bench that evaluates interactive intelligence using two representative datasets, MMMU-Pro and MathVerse, to test 10 different open-source LMMs. Additionally, we present InterFeedback-Human, a newly collected dataset of 120 cases designed for manually testing interactive performance in leading models such as OpenAI-o1 and Claude-3.5-Sonnet. Our evaluation results indicate that even the state-of-the-art LMM, OpenAI-o1, struggles to refine its responses based on human feedback, achieving an average score of less than 50%. Our findings point to the need for methods that can enhance LMMs' capabilities to interpret and benefit from feedback. In this paper, we are curious about the question "Can Large Multimodal Models evolve through Interactive Human Feedback?" It is central to developing general-purpose AI assistants with Large Multimodal Models (LMMs). While these models show exceptional performance on tackling multimodal tasks directly, their ability to interact with humans remains largely unknown. We argue that an LMM functioning as the general assistant should possess two capabilities: 1) exceptional problem-solving ability and 2) the ability to improve itself through feedback (e.g., human feedback, execution results).

Deep Reinforcement Learning (RL) has been explored and verified to be effective in solving decision-making tasks in various domains, such as robotics, transportation, recommender systems, etc. It learns from the interaction with environments and updates the policy using the collected experience. However, due to the limited real-world data and unbearable consequences of taking detrimental actions, the learning of RL policy is mainly restricted within the simulators. This practice guarantees safety in learning but introduces an inevitable sim-to-real gap in terms of deployment, thus causing degraded performance and risks in execution. There are attempts to solve the sim-to-real problems from different domains with various techniques, especially in the era with emerging techniques such as large foundations or language models that have cast light on the sim-to-real. This survey paper, to the best of our knowledge, is the first taxonomy that formally frames the sim-to-real techniques from key elements of the Markov Decision Process (State, Action, Transition, and Reward). Based on the framework, we cover comprehensive literature from the classic to the most advanced methods including the sim-to-real techniques empowered by foundation models, and we also discuss the specialties that are worth attention in different domains of sim-to-real problems. Then we summarize the formal evaluation process of sim-to-real performance with accessible code or benchmarks. The challenges and opportunities are also presented to encourage future exploration of this direction. We are actively maintaining a repository to include the most up-to-date sim-to-real research work to help domain researchers.

We perform Bayesian optimization using a Gaussian process perspective on Bayesian Additive Regression Trees (BART). Our BART Kernel (BARK) uses tree agreement to define a posterior over piecewise-constant functions, and we explore the space of tree kernels using a Markov chain Monte Carlo approach. Where BART only samples functions, the resulting BARK model obtains samples of Gaussian processes defining distributions over functions, which allow us to build acquisition functions for Bayesian optimization. Our tree-based approach enables global optimization over the surrogate, even for mixed-feature spaces. Moreover, where many previous tree-based kernels provide uncertainty quantification over function values, our sampling scheme captures uncertainty over the tree structure itself. Our experiments show the strong performance of BARK on both synthetic and applied benchmarks, due to the combination of our fully Bayesian surrogate and the optimization procedure.

Large Language Model~(LLM) based agents have been increasingly popular in solving complex and dynamic tasks, which requires proper evaluation systems to assess their capabilities. Nevertheless, existing benchmarks usually either focus on single-objective tasks or use overly broad assessing metrics, failing to provide a comprehensive inspection of the actual capabilities of LLM-based agents in complicated decision-making tasks. To address these issues, we introduce DSGBench, a more rigorous evaluation platform for strategic decision-making. Firstly, it incorporates six complex strategic games which serve as ideal testbeds due to their long-term and multi-dimensional decision-making demands and flexibility in customizing tasks of various difficulty levels or multiple targets. Secondly, DSGBench employs a fine-grained evaluation scoring system which examines the decision-making capabilities by looking into the performance in five specific dimensions and offering a comprehensive assessment in a well-designed way. Furthermore, DSGBench also incorporates an automated decision-tracking mechanism which enables in-depth analysis of agent behaviour patterns and the changes in their strategies. We demonstrate the advances of DSGBench by applying it to multiple popular LLM-based agents and our results suggest that DSGBench provides valuable insights in choosing LLM-based agents as well as improving their future development. DSGBench is available at https://github.com/DeciBrain-Group/DSGBench.

Deception is a common strategy adapted by autonomous systems in adversarial settings. Existing deception methods primarily focus on increasing opacity or misdirecting agents away from their goal or itinerary. In this work, we propose a deception problem aiming to mislead the robot towards a decoy goal through altering sensor events under a constrained budget of alteration. The environment along with the robot's interaction with it is modeled as a Partially Observable Markov Decision Process (POMDP), and the robot's action selection is governed by a Finite State Controller (FSC). Given a constrained budget for sensor event modifications, the objective is to compute a sensor alteration that maximizes the probability of the robot reaching a decoy goal. We establish the computational hardness of the problem by a reduction from the $0/1$ Knapsack problem and propose a Mixed Integer Linear Programming (MILP) formulation to compute optimal deception strategies. We show the efficacy of our MILP formulation via a sequence of experiments.

We characterize the class of quantum measurements that matches the applications of quantum theory to cognition (and decision making) - quantum-like modeling. Projective measurements describe the canonical measurements of the basic observables of quantum physics. However, the combinations of the basic cognitive effects, such as the question order and response replicability effects, cannot be described by projective measurements. We motivate the use of the special class of quantum measurements, namely {\it sharp repeatable non-projective measurements} - ${\cal SR\bar{P}}. $ This class is practically unused in quantum physics. Thus, physics and cognition explore different parts of quantum measurement theory. Quantum-like modeling isn't automatic borrowing of the quantum formalism. Exploring the class ${\cal SR\bar{P}}$ highlights the role of {\it noncommutativity of the state update maps generated by measurement back action.} Thus, ``non-classicality'' in quantum physics as well as quantum-like modeling for cognition is based on two different types of noncommutativity, of operators (observables) and instruments (state update maps): {\it observable-noncommutativity} vs. {\it state update-noncommutativity}. We speculate that distinguishing quantum-like properties of the cognitive effects are the expressions of the latter, or possibly both.

Latent variable models are powerful tools for learning low-dimensional manifolds from high-dimensional data. However, when dealing with constrained data such as unit-norm vectors or symmetric positive-definite matrices, existing approaches ignore the underlying geometric constraints or fail to provide meaningful metrics in the latent space. To address these limitations, we propose to learn Riemannian latent representations of such geometric data. To do so, we estimate the pullback metric induced by a Wrapped Gaussian Process Latent Variable Model, which explicitly accounts for the data geometry. This enables us to define geometry-aware notions of distance and shortest paths in the latent space, while ensuring that our model only assigns probability mass to the data manifold. This generalizes previous work and allows us to handle complex tasks in various domains, including robot motion synthesis and analysis of brain connectomes.

Despite their outstanding performance in the majority of scenarios, contemporary language models still occasionally generate undesirable outputs, for example, hallucinated text. While such behaviors have previously been linked to uncertainty, there is a notable lack of methods that actively consider uncertainty during text generation. In this work, we show how Minimum Bayes Risk (MBR) decoding, which selects model generations according to an expected risk, can be generalized into a principled uncertainty-aware decoding method. In short, we account for model uncertainty during decoding by incorporating a posterior over model parameters into MBR's computation of expected risk. We show that this modified expected risk is useful for both choosing outputs and deciding when to abstain from generation and can provide improvements without incurring overhead. We benchmark different methods for learning posteriors and show that performance improves with prediction diversity. We release our code publicly.

Evaluating AI safety requires statistically rigorous methods and risk metrics for understanding how the use of AI affects aggregated risk. However, much AI safety literature focuses upon risks arising from AI models in isolation, lacking consideration of how modular use of AI affects risk distribution of workflow components or overall risk metrics. There is also a lack of statistical grounding enabling sensitisation of risk models in the presence of absence of AI to estimate causal contributions of AI. This is in part due to the dearth of AI impact data upon which to fit distributions. In this work, we address these gaps in two ways. First, we demonstrate how scenario modelling (grounded in established statistical techniques such as Markov chains, copulas and Monte Carlo simulation) can be used to model AI risk holistically. Second, we show how lookalike distributions from phenomena analogous to AI can be used to estimate AI impacts in the absence of directly observable data. We demonstrate the utility of our methods for benchmarking cumulative AI risk via risk analysis of a logistic scenario simulations.

Active learning aims to efficiently build a labeled training set by strategically selecting samples to query labels from annotators. In this sequential process, each sample acquisition influences subsequent selections, causing dependencies among samples in the labeled set. However, these dependencies are overlooked during the model parameter estimation stage when updating the model using Maximum Likelihood Estimation (MLE), a conventional method that assumes independent and identically distributed (i.i.d.) data. We propose Dependency-aware MLE (DMLE), which corrects MLE within the active learning framework by addressing sample dependencies typically neglected due to the i.i.d. assumption, ensuring consistency with active learning principles in the model parameter estimation process. This improved method achieves superior performance across multiple benchmark datasets, reaching higher performance in earlier cycles compared to conventional MLE. Specifically, we observe average accuracy improvements of 6\%, 8.6\%, and 10.5\% for $k=1$, $k=5$, and $k=10$ respectively, after collecting the first 100 samples, where entropy is the acquisition function and $k$ is the query batch size acquired at every active learning cycle.