In psycholinguistics, the creation of controlled materials is crucial to ensure that research outcomes are solely attributed to the intended manipulations and not influenced by extraneous factors. To achieve this, psycholinguists typically pretest linguistic materials, where a common pretest is to solicit plausibility judgments from human evaluators on specific sentences. In this work, we investigate whether Language Models (LMs) can be used to generate these plausibility judgements. We investigate a wide range of LMs across multiple linguistic structures and evaluate whether their plausibility judgements correlate with human judgements. We find that GPT-4 plausibility judgements highly correlate with human judgements across the structures we examine, whereas other LMs correlate well with humans on commonly used syntactic structures. We then test whether this correlation implies that LMs can be used instead of humans for pretesting. We find that when coarse-grained plausibility judgements are needed, this works well, but when fine-grained judgements are necessary, even GPT-4 does not provide satisfactory discriminative power.

Large Language Models (LLMs) show significant potential in economic and strategic interactions, where communication via natural language is often prevalent. This raises key questions: Do LLMs behave rationally? How do they perform compared to humans? Do they tend to reach an efficient and fair outcome? What is the role of natural language in strategic interaction? How do characteristics of the economic environment influence these dynamics? These questions become crucial concerning the economic and societal implications of integrating LLM-based agents into real-world data-driven systems, such as online retail platforms and recommender systems. To answer these questions, we introduce a benchmark for standardizing research on two-player, sequential, language-based games. Inspired by the economic literature, we define three base families of games with consistent parameterization, degrees of freedom and economic measures to evaluate agents' performance (self-gain), as well as the game outcome (efficiency and fairness). We develop an open-source framework for interaction simulation and analysis, and utilize it to collect a dataset of LLM vs. LLM interactions across numerous game configurations and an additional dataset of human vs. LLM interactions. Through extensive experimentation, we demonstrate how our framework and dataset can be used to: (i) compare the behavior of LLM-based agents in various economic contexts; (ii) evaluate agents in both individual and collective performance measures; and (iii) quantify the effect of the economic characteristics of the environments on the behavior of agents. Our results suggest that the market parameters, as well as the choice of the LLMs, tend to have complex and interdependent effects on the economic outcome, which calls for careful design and analysis of the language-based economic ecosystem.

Language agents, built on top of language models (LMs), are systems that can interact with complex environments, such as the open web. In this work, we examine whether such agents can perform realistic and time-consuming tasks on the web, e.g., monitoring real-estate markets or locating relevant nearby businesses. We introduce AssistantBench, a challenging new benchmark consisting of 214 realistic tasks that can be automatically evaluated, covering different scenarios and domains. We find that AssistantBench exposes the limitations of current systems, including language models and retrieval-augmented language models, as no model reaches an accuracy of more than 25 points. While closed-book LMs perform well in terms of accuracy, they exhibit low precision and tend to hallucinate facts. State-of-the-art web agents reach a score of near zero. Additionally, we introduce SeePlanAct (SPA), a new web agent that significantly outperforms previous agents, and an ensemble of SPA and closed-book models reaches the best overall performance. Moreover, we analyze failures of current systems and highlight that open web navigation remains a major challenge.

Retrieval-augmented language models (LMs) have received much attention recently. However, typically the retriever is not trained jointly as a native component of the LM, but added to an already-pretrained LM, which limits the ability of the LM and the retriever to adapt to one another. In this work, we propose the Retrieval-Pretrained Transformer (RPT), an architecture and training procedure for jointly training a retrieval-augmented LM from scratch for the task of modeling long texts. Given a recently generated text chunk in a long document, the LM computes query representations, which are then used to retrieve earlier chunks in the document, located potentially tens of thousands of tokens before. Information from retrieved chunks is fused into the LM representations to predict the next target chunk. We train the retriever component with a semantic objective, where the goal is to retrieve chunks that increase the probability of the next chunk, according to a reference LM. We evaluate RPT on four long-range language modeling tasks, spanning books, code, and mathematical writing, and demonstrate that RPT improves retrieval quality and subsequently perplexity across the board compared to strong baselines.

There is increasing interest in using LLMs as decision-making "agents." Doing so includes many degrees of freedom: which model should be used; how should it be prompted; should it be asked to introspect, conduct chain-of-thought reasoning, etc? Settling these questions -- and more broadly, determining whether an LLM agent is reliable enough to be trusted -- requires a methodology for assessing such an agent's economic rationality. In this paper, we provide one. We begin by surveying the economic literature on rational decision making, taxonomizing a large set of fine-grained "elements" that an agent should exhibit, along with dependencies between them. We then propose a benchmark distribution that quantitatively scores an LLMs performance on these elements and, combined with a user-provided rubric, produces a "rationality report card." Finally, we describe the results of a large-scale empirical experiment with 14 different LLMs, characterizing the both current state of the art and the impact of different model sizes on models' ability to exhibit rational behavior.

Existing benchmarks for open-domain question answering (ODQA) typically focus on questions whose answers can be extracted from a single paragraph. By contrast, many natural questions, such as "What players were drafted by the Brooklyn Nets?" have a list of answers. Answering such questions requires retrieving and reading from many passages, in a large corpus. We introduce QAMPARI, an ODQA benchmark, where question answers are lists of entities, spread across many paragraphs. We created QAMPARI by (a) generating questions with multiple answers from Wikipedia's knowledge graph and tables, (b) automatically pairing answers with supporting evidence in Wikipedia paragraphs, and (c) manually paraphrasing questions and validating each answer. We train ODQA models from the retrieve-and-read family and find that QAMPARI is challenging in terms of both passage retrieval and answer generation, reaching an F1 score of 32.8 at best. Our results highlight the need for developing ODQA models that handle a broad range of question types, including single and multi-answer questions.

In-context learning is a recent paradigm in natural language understanding, where a large pre-trained language model (LM) observes a test instance and a few training examples as its input, and directly decodes the output without any update to its parameters. However, performance has been shown to strongly depend on the selected training examples (termed prompt). In this work, we propose an efficient method for retrieving prompts for in-context learning using annotated data and a LM. Given an input-output pair, we estimate the probability of the output given the input and a candidate training example as the prompt, and label training examples as positive or negative based on this probability. We then train an efficient dense retriever from this data, which is used to retrieve training examples as prompts at test time. We evaluate our approach on three sequence-to-sequence tasks where language utterances are mapped to meaning representations, and find that it substantially outperforms prior work and multiple baselines across the board.

The de-facto standard decoding method for semantic parsing in recent years has been to autoregressively decode the abstract syntax tree of the target program using a top-down depth-first traversal. In this work, we propose an alternative approach: a Semi-autoregressive Bottom-up Parser (SmBoP) that constructs at decoding step t the top-K sub-trees of height ≤ t. Our parser enjoys several benefits compared to top-down autoregressive parsing. From an efficiency perspective, bottom-up parsing allows to decode all sub-trees of a certain height in parallel, leading to logarithmic runtime complexity rather than linear. From a modeling perspective, a bottom-up parser learns representations for meaningful semantic sub-programs at each step, rather than for semantically-vacuous partial trees. We apply SmBoP on Spider, a challenging zero-shot semantic parsing benchmark, and show that SmBoP leads to a 2.2x speed-up in decoding time and a ~5x speed-up in training time, compared to a semantic parser that uses autoregressive decoding. SmBoP obtains 71.1 denotation accuracy on Spider, establishing a new state-of-the-art, and 69.5 exact match, comparable to the 69.6 exact match of the autoregressive RAT-SQL+GraPPa.

During this project we set to take on the foundations laid down in Geva et al. (2018), which introduced a novel framework to navigate document structure for the reading comprehension task using deep reinforcement learning and improve upon it. The goal of the Reading Comprehension (RC) task is to enable machines to understand documents, and answer questions about them. Previous work on the RC task for long documents relied on scanning with a RNN entire documents tokenby-token which was time consuming even with a basic model. The Information Retrieval (IR) approach is to retrieve excerpts from the document and over these excerpts run an RNN, this approach is problematic since as document length increases, the efficiency of one-shot IR methods decreases and thousands of tokens are retrieved. In contrast, the work done in Geva et al. (2018) introduces a new approach, in the article documents were represented as trees, and a reinforcement learning agent learns to combine tokens from the document tree with feedback from a more expressive answer extraction model to navigate within the tree. The authors showed that it is possible to train an RL agent to successfully navigate a document while consuming a fraction of the tokens that a standard RC model uses. We follow up on this approach and further study the methods proposed. We accomplish this by redesigning of the code base used to allow use of any general actor-critic reinforcement learning framework for the TriviaQA-NoP dataset. We showcase our results using the IMPALA framework and the V-trace algorithm introduced in Espeholt et al. (2018), and denote our model IMPALA Question Answer Navigator (IQAN)2. In order to highlight the impact the new architecture has on training, we modify the original environment to use a sparse reward function, remove the replay buffer and do not employ tree sampling,