ScienceAgentBench: Toward Rigorous Assessment of Language Agents for Data-Driven Scientific Discovery

The advancements of language language models (LLMs) have piqued growing interest in developing LLM-based language agents to automate scientific discovery end-to-end, which has sparked both excitement and skepticism about the true capabilities of such agents. In this work, we argue that for an agent to fully automate scientific discovery, it must be able to complete all essential tasks in the workflow. Thus, we call for rigorous assessment of agents on individual tasks in a scientific workflow before making bold claims on end-to-end automation. To this end, we present ScienceAgentBench, a new benchmark for evaluating language agents for data-driven scientific discovery. To ensure the scientific authenticity and real-world relevance of our benchmark, we extract 102 tasks from 44 peer-reviewed publications in four disciplines and engage nine subject matter experts to validate them. We unify the target output for every task to a self-contained Python program file and employ an array of evaluation metrics to examine the generated programs, execution results, and costs. Each task goes through multiple rounds of manual validation by annotators and subject matter experts to ensure its annotation quality and scientific plausibility. We also propose two effective strategies to mitigate data contamination concerns. Using our benchmark, we evaluate five open-weight and proprietary LLMs, each with three frameworks: direct prompting, OpenHands, and self-debug. Given three attempts for each task, the best-performing agent can only solve 32.4% of the tasks independently and 34.3% with expert-provided knowledge. These results underscore the limited capacities of current language agents in generating code for data-driven discovery, let alone end-to-end automation for scientific research.

Grokked Transformers are Implicit Reasoners: A Mechanistic Journey to the Edge of Generalization

We study whether transformers can learn to implicitly reason over parametric knowledge, a skill that even the most capable language models struggle with. Focusing on two representative reasoning types, composition and comparison, we consistently find that transformers can learn implicit reasoning, but only through grokking, i.e., extended training far beyond overfitting. The levels of generalization also vary across reasoning types: when faced with out-of-distribution examples, transformers fail to systematically generalize for composition but succeed for comparison. We delve into the model's internals throughout training, conducting analytical experiments that reveal: 1) the mechanism behind grokking, such as the formation of the generalizing circuit and its relation to the relative efficiency of generalizing and memorizing circuits, and 2) the connection between systematicity and the configuration of the generalizing circuit. Our findings guide data and training setup to better induce implicit reasoning and suggest potential improvements to the transformer architecture, such as encouraging cross-layer knowledge sharing. Furthermore, we demonstrate that for a challenging reasoning task with a large search space, GPT-4-Turbo and Gemini-1.5-Pro based on non-parametric memory fail badly regardless of prompting styles or retrieval augmentation, while a fully grokked transformer can achieve near-perfect accuracy, showcasing the power of parametric memory for complex reasoning.

Beamforming Design for Double-Active-RIS-aided Communication Systems with Inter-Excitation

In this paper, we investigate a double-active-reconfigurable intelligent surface (RIS)-aided downlink wireless communication system, where a multi-antenna base station (BS) serves multiple single-antenna users with both double reflection and single reflection links. Due to the signal amplification capability of active RISs, the mutual influence between active RISs, which is termed as the "inter-excitation" effect, cannot be ignored. Then, we develop a feedback-type model to characterize the signal containing the inter-excitation effect. Based on the signal model, we formulate a weighted sum rate (WSR) maximization problem by jointly optimizing the beamforming matrix at the BS and the reflecting coefficient matrices at the two active RISs, subject to power constraints at the BS and active RISs, as well as the maximum amplification gain constraints of the active RISs. To solve this non-convex problem, we first transform the problem into a more tractable form using the fractional programming (FP) method. Then, by introducing auxiliary variables, the problem can be converted into an equivalent form that can be solved by using a low-complexity penalty dual decomposition (PDD) algorithm. Finally, simulation results indicate that it is crucial to consider the inter-excitation effect between active RISs in beamforming design for double-active-RIS-aided communication systems. Additionally, it prevails over other benchmark schemes with single active RIS and double passive RISs in terms of achievable rate.

LLMs in the Imaginarium: Tool Learning through Simulated Trial and Error

Tools are essential for large language models (LLMs) to acquire up-to-date information and take consequential actions in external environments. Existing work on tool-augmented LLMs primarily focuses on the broad coverage of tools and the flexibility of adding new tools. However, a critical aspect that has surprisingly been understudied is simply how accurately an LLM uses tools for which it has been trained. We find that existing LLMs, including GPT-4 and open-source LLMs specifically fine-tuned for tool use, only reach a correctness rate in the range of 30% to 60%, far from reliable use in practice. We propose a biologically inspired method for tool-augmented LLMs, simulated trial and error (STE), that orchestrates three key mechanisms for successful tool use behaviors in the biological system: trial and error, imagination, and memory. Specifically, STE leverages an LLM's 'imagination' to simulate plausible scenarios for using a tool, after which the LLM interacts with the tool to learn from its execution feedback. Both short-term and long-term memory are employed to improve the depth and breadth of the exploration, respectively. Comprehensive experiments on ToolBench show that STE substantially improves tool learning for LLMs under both in-context learning and fine-tuning settings, bringing a boost of 46.7% to Mistral-Instruct-7B and enabling it to outperform GPT-4. We also show effective continual learning of tools via a simple experience replay strategy.

Reconfigurable Intelligent Surface-Aided Dual-Function Radar and Communication Systems With MU-MIMO Communication

In this paper, we investigate an reconfigurable intelligent surface (RIS)-aided integrated sensing and communication (ISAC) system. Our objective is to maximize the achievable sum rate of the multi-antenna communication users through the joint active and passive beamforming. {Specifically}, the weighted minimum mean-square error (WMMSE) method is { first} used to reformulate the original problem into an equivalent one. Then, we utilize an alternating optimization (AO) { algorithm} to decouple the optimization variables and decompose this challenging problem into two subproblems. Given reflecting coefficients, a penalty-based algorithm is utilized to deal with the non-convex radar signal-to-noise ratio (SNR) constraints. For the given beamforming matrix of the BS, we apply majorization-minimization (MM) to transform the problem into a quadratic constraint quadratic programming (QCQP) problem, which is ultimately solved using a semidefinite relaxation (SDR)-based algorithm. Simulation results illustrate the advantage of deploying RIS in the considered multi-user MIMO (MU-MIMO) ISAC systems.

Joint Beamforming Design for Double Active RIS-assisted Radar-Communication Coexistence Systems

Integrated sensing and communication (ISAC) technology has been considered as one of the key candidate technologies in the next-generation wireless communication systems. However, when radar and communication equipment coexist in the same system, i.e. radar-communication coexistence (RCC), the interference from communication systems to radar can be large and cannot be ignored. Recently, reconfigurable intelligent surface (RIS) has been introduced into RCC systems to reduce the interference. However, the "multiplicative fading" effect introduced by passive RIS limits its performance. To tackle this issue, we consider a double active RIS-assisted RCC system, which focuses on the design of the radar's beamforming vector and the active RISs' reflecting coefficient matrices, to maximize the achievable data rate of the communication system. The considered system needs to meet the radar detection constraint and the power budgets at the radar and the RISs. Since the problem is non-convex, we propose an algorithm based on the penalty dual decomposition (PDD) framework. Specifically, we initially introduce auxiliary variables to reformulate the coupled variables into equation constraints and incorporate these constraints into the objective function through the PDD framework. Then, we decouple the equivalent problem into several subproblems by invoking the block coordinate descent (BCD) method. Furthermore, we employ the Lagrange dual method to alternately optimize these subproblems. Simulation results verify the effectiveness of the proposed algorithm. Furthermore, the results also show that under the same power budget, deploying double active RISs in RCC systems can achieve higher data rate than those with single active RIS and double passive RISs.

Secure Wireless Communication in Active RIS-Assisted DFRC System

This work considers a dual-functional radar and communication (DFRC) system with an active reconfigurable intelligent surface (RIS) and a potential eavesdropper. Our purpose is to maximize the secrecy rate (SR) of the system by jointly designing the beamforming matrix at the DFRC base station (BS) and the reflecting coefficients at the active RIS, subject to the signal-to-interference-plus-noise-ratio (SINR) constraint of the radar echo and the power consumption constraints at the DFRC-BS and active RIS. An alternating optimization (AO) algorithm based on semi-definite relaxation (SDR) and majorizationminimization (MM) is applied to solve the SR-maximization problem by alternately optimizing the beamforming matrix and the reflecting coefficients. Specifically, we first apply the SDR and successive convex approximation (SCA) methods to transform the two subproblems into more tractable forms, then the MM method is applied to derive a concave surrogate function and iteratively solve the subproblems. Finally, simulation results indicate that the active RIS can better confront the impact of "multiplicative fading" and outperforms traditional passive RIS in terms of both secure data rate and radar sensing performance.

How Trustworthy are Open-Source LLMs? An Assessment under Malicious Demonstrations Shows their Vulnerabilities

The rapid progress in open-source Large Language Models (LLMs) is significantly driving AI development forward. However, there is still a limited understanding of their trustworthiness. Deploying these models at scale without sufficient trustworthiness can pose significant risks, highlighting the need to uncover these issues promptly. In this work, we conduct an assessment of open-source LLMs on trustworthiness, scrutinizing them across eight different aspects including toxicity, stereotypes, ethics, hallucination, fairness, sycophancy, privacy, and robustness against adversarial demonstrations. We propose an enhanced Chain of Utterances-based (CoU) prompting strategy by incorporating meticulously crafted malicious demonstrations for trustworthiness attack. Our extensive experiments encompass recent and representative series of open-source LLMs, including Vicuna, MPT, Falcon, Mistral, and Llama 2. The empirical outcomes underscore the efficacy of our attack strategy across diverse aspects. More interestingly, our result analysis reveals that models with superior performance in general NLP tasks do not always have greater trustworthiness; in fact, larger models can be more vulnerable to attacks. Additionally, models that have undergone instruction tuning, focusing on instruction following, tend to be more susceptible, although fine-tuning LLMs for safety alignment proves effective in mitigating adversarial trustworthiness attacks.

Mind2Web: Towards a Generalist Agent for the Web

We introduce Mind2Web, the first dataset for developing and evaluating generalist agents for the web that can follow language instructions to complete complex tasks on any website. Existing datasets for web agents either use simulated websites or only cover a limited set of websites and tasks, thus not suitable for generalist web agents. With over 2,000 open-ended tasks collected from 137 websites spanning 31 domains and crowdsourced action sequences for the tasks, Mind2Web provides three necessary ingredients for building generalist web agents: 1) diverse domains, websites, and tasks, 2) use of real-world websites instead of simulated and simplified ones, and 3) a broad spectrum of user interaction patterns. Based on Mind2Web, we conduct an initial exploration of using large language models (LLMs) for building generalist web agents. While the raw HTML of real-world websites are often too large to be fed to LLMs, we show that first filtering it with a small LM significantly improves the effectiveness and efficiency of LLMs. Our solution demonstrates a decent level of performance, even on websites or entire domains the model has never seen before, but there is still a substantial room to improve towards truly generalizable agents. We open-source our dataset, model implementation, and trained models (https://osu-nlp-group.github.io/Mind2Web) to facilitate further research on building a generalist agent for the web.

Can ChatGPT Defend the Truth? Automatic Dialectical Evaluation Elicits LLMs' Deficiencies in Reasoning

We explore testing the reasoning ability of large language models (LLMs), such as ChatGPT, by engaging with them in a debate-like conversation that probes deeper into their understanding of the subject. Specifically, we formulate a new task where given a question, the LLM can generate a correct solution while the user believes in a wrong solution in the beginning, and they need to discuss to make the correct decision through dialogue. Such a setting requires the LLM to not only achieve the correct answer on its own (which could be done by shallow memorization), but also be able to defend the truth instead of blindly believing or getting misled by the user's (invalid) arguments and critiques, thus testing in greater depth whether the LLM grasps the essence of the reasoning required to solve the problem. To automate this evaluation framework and save human labor, we simulate the user using another LLM conditioned on a synthesized wrong solution. Across a range of complex reasoning benchmarks spanning math, commonsense, logic and tasks from BIG-Bench, we find that despite being able to generate correct step-by-step solutions in the beginning, ChatGPT cannot maintain its belief in truth for a significant portion of examples when challenged by often-time absurdly invalid arguments. Our work reveals LLMs' weaknesses not captured by conventional benchmarking, and also points to danger zones of aligning models with human feedback.