Computing olympiads contain some of the most challenging problems for humans, requiring complex algorithmic reasoning, puzzle solving, in addition to generating efficient code. However, it has been understudied as a domain to evaluate language models (LMs). In this paper, we introduce the USACO benchmark with 307 problems from the USA Computing Olympiad, along with high-quality unit tests, reference code, and official analyses for each problem. These resources enable us to construct and test a range of LM inference methods for competitive programming for the first time. We find GPT-4 only achieves a 8.7% pass@1 accuracy with zero-shot chain-of-thought prompting, and our best inference method improves it to 20.2% using a combination of self-reflection and retrieval over episodic knowledge. However, this is far from solving the benchmark. To better understand the remaining challenges, we design a novel human-in-the-loop study and surprisingly find that a small number of targeted hints enable GPT-4 to solve 13 out of 15 problems previously unsolvable by any model and method. Our benchmark, baseline methods, quantitative results, and qualitative analysis serve as an initial step toward LMs with grounded, creative, and algorithmic reasoning.
State-of-the-art large language models (LLMs) have become indispensable tools for various tasks. However, training LLMs to serve as effective assistants for humans requires careful consideration. A promising approach is reinforcement learning from human feedback (RLHF), which leverages human feedback to update the model in accordance with human preferences and mitigate issues like toxicity and hallucinations. Yet, an understanding of RLHF for LLMs is largely entangled with initial design choices that popularized the method and current research focuses on augmenting those choices rather than fundamentally improving the framework. In this paper, we analyze RLHF through the lens of reinforcement learning principles to develop an understanding of its fundamentals, dedicating substantial focus to the core component of RLHF -- the reward model. Our study investigates modeling choices, caveats of function approximation, and their implications on RLHF training algorithms, highlighting the underlying assumptions made about the expressivity of reward. Our analysis improves the understanding of the role of reward models and methods for their training, concurrently revealing limitations of the current methodology. We characterize these limitations, including incorrect generalization, model misspecification, and the sparsity of feedback, along with their impact on the performance of a language model. The discussion and analysis are substantiated by a categorical review of current literature, serving as a reference for researchers and practitioners to understand the challenges of RLHF and build upon existing efforts.
Installing probabilistic world models into artificial agents opens an efficient channel for humans to communicate with and control these agents. In addition to updating agent policies, humans can modify their internal world models in order to influence their decisions. The challenge, however, is that currently existing world models are difficult for humans to adapt because they lack a natural communication interface. Aimed at addressing this shortcoming, we develop Language-Guided World Models (LWMs), which can capture environment dynamics by reading language descriptions. These models enhance agent communication efficiency, allowing humans to simultaneously alter their behavior on multiple tasks with concise language feedback. They also enable agents to self-learn from texts originally written to instruct humans. To facilitate the development of LWMs, we design a challenging benchmark based on the game of MESSENGER (Hanjie et al., 2021), requiring compositional generalization to new language descriptions and environment dynamics. Our experiments reveal that the current state-of-the-art Transformer architecture performs poorly on this benchmark, motivating us to design a more robust architecture. To showcase the practicality of our proposed LWMs, we simulate a scenario where these models augment the interpretability and safety of an agent by enabling it to generate and discuss plans with a human before execution. By effectively incorporating language feedback on the plan, the models boost the agent performance in the real environment by up to three times without collecting any interactive experiences in this environment.
Quantitative evaluation metrics have traditionally been pivotal in gauging the advancements of artificial intelligence systems, including large language models (LLMs). However, these metrics have inherent limitations. Given the intricate nature of real-world tasks, a single scalar to quantify and compare is insufficient to capture the fine-grained nuances of model behavior. Metrics serve only as a way to compare and benchmark models, and do not yield actionable diagnostics, thus making the model improvement process challenging. Model developers find themselves amid extensive manual efforts involving sifting through vast datasets and attempting hit-or-miss adjustments to training data or setups. In this work, we address the shortcomings of quantitative metrics by proposing QualEval, which augments quantitative scalar metrics with automated qualitative evaluation as a vehicle for model improvement. QualEval uses a powerful LLM reasoner and our novel flexible linear programming solver to generate human-readable insights that when applied, accelerate model improvement. The insights are backed by a comprehensive dashboard with fine-grained visualizations and human-interpretable analyses. We corroborate the faithfulness of QualEval by demonstrating that leveraging its insights, for example, improves the absolute performance of the Llama 2 model by up to 15% points relative on a challenging dialogue task (DialogSum) when compared to baselines. QualEval successfully increases the pace of model development, thus in essence serving as a data-scientist-in-a-box. Given the focus on critiquing and improving current evaluation metrics, our method serves as a refreshingly new technique for both model evaluation and improvement.
Assistant AI agents should be capable of rapidly acquiring novel skills and adapting to new user preferences. Traditional frameworks like imitation learning and reinforcement learning do not facilitate this capability because they support only low-level, inefficient forms of communication. In contrast, humans communicate with progressive efficiency by defining and sharing abstract intentions. Reproducing similar capability in AI agents, we develop a novel learning framework named Communication-Efficient Interactive Learning (CEIL). By equipping a learning agent with an abstract, dynamic language and an intrinsic motivation to learn with minimal communication effort, CEIL leads to emergence of a human-like pattern where the learner and the teacher communicate progressively efficiently by exchanging increasingly more abstract intentions. CEIL demonstrates impressive performance and communication efficiency on a 2D MineCraft domain featuring long-horizon decision-making tasks. Agents trained with CEIL quickly master new tasks, outperforming non-hierarchical and hierarchical imitation learning by up to 50% and 20% in absolute success rate, respectively, given the same number of interactions with the teacher. Especially, the framework performs robustly with teachers modeled after human pragmatic communication behavior.
Language models have outpaced our ability to evaluate them effectively, but for their future development it is essential to study the frontier of their capabilities. We consider real-world software engineering to be a rich, sustainable, and challenging testbed for evaluating the next generation of language models. We therefore introduce SWE-bench, an evaluation framework including $2,294$ software engineering problems drawn from real GitHub issues and corresponding pull requests across $12$ popular Python repositories. Given a codebase along with a description of an issue to be resolved, a language model is tasked with editing the codebase to address the issue. Resolving issues in SWE-bench frequently requires understanding and coordinating changes across multiple functions, classes, and even files simultaneously, calling for models to interact with execution environments, process extremely long contexts and perform complex reasoning that goes far beyond traditional code generation. Our evaluations show that both state-of-the-art proprietary models and our fine-tuned model SWE-Llama can resolve only the simplest issues. Claude 2 and GPT-4 solve a mere $4.8$% and $1.7$% of instances respectively, even when provided with an oracle retriever. Advances on SWE-bench represent steps towards LMs that are more practical, intelligent, and autonomous.
Recent efforts have augmented language models (LMs) with external tools or environments, leading to the development of language agents that can reason and act. However, most of these agents rely on few-shot prompting techniques with off-the-shelf LMs. In this paper, we investigate and argue for the overlooked direction of fine-tuning LMs to obtain language agents. Using a setup of question answering (QA) with a Google search API, we explore a variety of base LMs, prompting methods, fine-tuning data, and QA tasks, and find language agents are consistently improved after fine-tuning their backbone LMs. For example, fine-tuning Llama2-7B with 500 agent trajectories generated by GPT-4 leads to a 77% HotpotQA performance increase. Furthermore, we propose FireAct, a novel approach to fine-tuning LMs with trajectories from multiple tasks and prompting methods, and show having more diverse fine-tuning data can further improve agents. Along with other findings regarding scaling effects, robustness, generalization, efficiency and cost, our work establishes comprehensive benefits of fine-tuning LMs for agents, and provides an initial set of experimental designs, insights, as well as open questions toward language agent fine-tuning.
Recent efforts have incorporated large language models (LLMs) with external resources (e.g., the Internet) or internal control flows (e.g., prompt chaining) for tasks requiring grounding or reasoning. However, these efforts have largely been piecemeal, lacking a systematic framework for constructing a fully-fledged language agent. To address this challenge, we draw on the rich history of agent design in symbolic artificial intelligence to develop a blueprint for a new wave of cognitive language agents. We first show that LLMs have many of the same properties as production systems, and recent efforts to improve their grounding or reasoning mirror the development of cognitive architectures built around production systems. We then propose Cognitive Architectures for Language Agents (CoALA), a conceptual framework to systematize diverse methods for LLM-based reasoning, grounding, learning, and decision making as instantiations of language agents in the framework. Finally, we use the CoALA framework to highlight gaps and propose actionable directions toward more capable language agents in the future.
Imitation Learning (IL) is one of the most widely used methods in machine learning. Yet, while powerful, many works find it is often not able to fully recover the underlying expert behavior. However, none of these works deeply investigate the role of scaling up the model and data size. Inspired by recent work in Natural Language Processing (NLP) where "scaling up" has resulted in increasingly more capable LLMs, we investigate whether carefully scaling up model and data size can bring similar improvements in the imitation learning setting. To demonstrate our findings, we focus on the game of NetHack, a challenging environment featuring procedural generation, stochasticity, long-term dependencies, and partial observability. We find IL loss and mean return scale smoothly with the compute budget and are strongly correlated, resulting in power laws for training compute-optimal IL agents with respect to model size and number of samples. We forecast and train several NetHack agents with IL and find they outperform prior state-of-the-art by at least 2x in all settings. Our work both demonstrates the scaling behavior of imitation learning in a challenging domain, as well as the viability of scaling up current approaches for increasingly capable agents in NetHack, a game that remains elusively hard for current AI systems.