We present AutoBencher, a declarative framework for automatic benchmark construction, and use it to scalably discover novel insights and vulnerabilities of existing language models. Concretely, given a few desiderata of benchmarks (e.g., question difficulty, topic salience), we operationalize each desideratum and cast benchmark creation as an optimization problem. Specifically, we experiment with two settings with different optimization objectives: (i) for capability evaluation, we declare the goal of finding a salient, difficult dataset that induces novel performance patterns; (ii) for safety evaluation, we declare the goal of finding a dataset of unsafe prompts that existing LMs fail to decline. To tackle this optimization problem, we use a language model to iteratively propose and refine dataset descriptions, which are then used to generate topic-specific questions and answers. These descriptions are optimized to improve the declared desiderata. We use AutoBencher (powered by GPT-4) to create datasets for math, multilinguality, knowledge, and safety. The scalability of AutoBencher allows it to test fine-grained categories and tail knowledge, creating datasets that elicit 22% more model errors (i.e., difficulty) than existing benchmarks. On the novelty ends, AutoBencher also helps identify specific gaps not captured by existing benchmarks: e.g., Gemini-Pro has knowledge gaps on Permian Extinction and Fordism while GPT-4o fails to decline harmful requests about cryptocurrency scams.

Standard exploration methods typically rely on random coverage of the state space or coverage-promoting exploration bonuses. However, in partially observed settings, the biggest exploration challenge is often posed by the need to discover information-gathering strategies—e.g., an agent that has to navigate to a location in traffic might learn to first check traffic conditions and then choose a route. In this work, we design a POMDP agent that gathers information about the hidden state, using ideas from the meta-exploration literature. Our approach provides an exploration bonus that rewards the agent for gathering information about the state that is relevant for completing the task. While this requires the agent to know what this information is during training, it can obtained in several ways: in the most general case, off-policy algorithms can leverage knowledge about the entire trajectory to determine such information in hindsight, but the user can also provide prior knowledge (e.g., privileged information) to help inform the training process. Through experiments in several partially-observed environments, we find that our approach is competitive with prior methods when minimal exploration is needed, but substantially outperforms them when more complex strategies are required. Our algorithm also shows the ability to learn without any privileged information, by reasoning about the entire trajectory in hindsight and and effectively using any information it reveals about the hidden state.

In this paper, we detail the successful deployment of a machine learning autograder that significantly decreases the grading labor required in the Breakout computer science assignment. This assignment - which tasks students with programming a game consisting of a controllable paddle and a ball that bounces off the paddle to break bricks - is popular for engaging students with introductory computer science concepts, but creates a large grading burden. Due to the game's interactive nature, grading defies traditional unit tests and instead typically requires 8+ minutes of manually playing each student's game to search for bugs. This amounts to 45+ hours of grading in a standard course offering and prevents further widespread adoption of the assignment. Our autograder alleviates this burden by playing each student's game with a reinforcement learning agent and providing videos of discovered bugs to instructors. In an A/B test with manual grading, we find that our human-in-the-loop AI autograder reduces grading time by 44%, while slightly improving grading accuracy by 6%, ultimately saving roughly 30 hours over our deployment in two offerings of the assignment. Our results further suggest the practicality of grading other interactive assignments (e.g., other games or building websites) via similar machine learning techniques. Live demo at https://ezliu.github.io/breakoutgrader.

Whereas machine learning models typically learn language by directly training on language tasks (e.g., next-word prediction), language emerges in human children as a byproduct of solving non-language tasks (e.g., acquiring food). Motivated by this observation, we ask: can embodied reinforcement learning (RL) agents also indirectly learn language from non-language tasks? Learning to associate language with its meaning requires a dynamic environment with varied language. Therefore, we investigate this question in a multi-task environment with language that varies across the different tasks. Specifically, we design an office navigation environment, where the agent’s goal is to find a particular office, and office locations differ in different buildings (i.e., tasks). Each building includes a floor plan with a simple language description of the goal office’s location, which can be visually read as an RGB image when visited. We find RL agents indeed are able to indirectly learn language. Agents trained with current meta-RL algorithms successfully generalize to reading floor plans with held-out layouts and language phrases, and quickly navigate to the correct office, despite receiving no direct language supervision.

Identifying statistical regularities in solutions to some tasks in multi-task reinforcement learning can accelerate the learning of new tasks.Skill learning offers one way of identifying these regularities by decomposing pre-collected experiences into a sequence of skills.A popular approach to skill learning is maximizing the likelihood of the pre-collected experience with latent variable models,where the latent variables represent the skills. However, there are often many solutions that maximize the likelihood equally well, including degenerate solutions. To address this underspecification, we propose a new objective that combines the maximum likelihood objective with a penalty on the description length of the skills. This penalty incentivizes the skills to maximally extract common structures from the experiences. Empirically, our objective learns skills that solve downstream tasks in fewer samples compared to skills learned from only maximizing likelihood. Further, while most prior works in the offline multi-task setting focus on tasks with low-dimensional observations, our objective can scale to challenging tasks with high-dimensional image observations.

Developing interactive software, such as websites or games, is a particularly engaging way to learn computer science. However, teaching and giving feedback on such software is time-consuming — standard approaches require instructors to manually grade student-implemented interactive programs. As a result, online platforms that serve millions, like Code.org, are unable to provide any feedback on assignments for implementing interactive programs, which critically hinders students’ ability to learn. One approach toward automatic grading is to learn an agent that interacts with a student’s program and explores states indicative of errors via reinforcement learning. However, existing work on this approach only provides binary feedback of whether a program is correct or not, while students require finer-grained feedback on the specific errors in their programs to understand their mistakes. In this work, we show that exploring to discover errors can be cast as a meta-exploration problem. This enables us to construct a principled objective for discovering errors and an algorithm for optimizing this objective, which provides fine-grained feedback. We evaluate our approach on a set of over 700K real anonymized student programs from a Code.org interactive assignment. Our approach provides feedback with 94.3% accuracy, improving over existing approaches by 17.7% and coming within 1.5% of human-level accuracy. Project web page: https://ezliu.github.io/dreamgrader.

Standard training via empirical risk minimization (ERM) can produce models that achieve low error on average but high error on minority groups, especially in the presence of spurious correlations between the input and label. Prior approaches to this problem, like group distributionally robust optimization (group DRO), generally require group annotations for every training point. On the other hand, approaches that do not use group annotations generally do not improve minority performance. For example, we find that joint DRO, which dynamically upweights examples with high training loss, tends to optimize for examples that are irrelevant to the specific groups we seek to do well on. In this paper, we propose a simple two-stage approach, JTT, that achieves comparable performance to group DRO while only requiring group annotations on a significantly smaller validation set. JTT first attempts to identify informative training examples, which are often minority examples, by training an initial ERM classifier and selecting the examples with high training loss. Then, it trains a final classifier by upsampling the selected examples. Crucially, unlike joint DRO, JTT does not iteratively upsample examples that have high loss under the final classifier. On four image classification and natural language processing tasks with spurious correlations, we show that JTT closes 85% of the gap in accuracy on the worst group between ERM and group DRO.

The goal of meta-reinforcement learning (meta-RL) is to build agents that can quickly learn new tasks by leveraging prior experience on related tasks. Learning a new task often requires both exploring to gather task-relevant information and exploiting this information to solve the task. In principle, optimal exploration and exploitation can be learned end-to-end by simply maximizing task performance. However, such meta-RL approaches struggle with local optima due to a chicken-and-egg problem: learning to explore requires good exploitation to gauge the exploration’s utility, but learning to exploit requires information gathered via exploration. Optimizing separate objectives for exploration and exploitation can avoid this problem, but prior meta-RL exploration objectives yield suboptimal policies that gather information irrelevant to the task. We alleviate both concerns by constructing an exploitation objective that automatically identifies task-relevant information and an exploration objective to recover only this information. This avoids local optima in end-to-end training, without sacrificing optimal exploration. Empirically, DREAM substantially outperforms existing approaches on complex meta-RL problems, such as sparse-reward 3D visual navigation. Videos of DREAM: https://ezliu.github.io/dream/

Program execution speed critically depends on increasing cache hits, as cache hits are orders of magnitude faster than misses. To increase cache hits, we focus on the problem of cache replacement: choosing which cache line to evict upon inserting a new line. This is challenging because it requires planning far ahead and currently there is no known practical solution. As a result, current replacement policies typically resort to heuristics designed for specific common access patterns, which fail on more diverse and complex access patterns. In contrast, we propose an imitation learning approach to automatically learn cache access patterns by leveraging Belady’s, an oracle policy that computes the optimal eviction decision given the future cache accesses. While directly applying Belady’s is infeasible since the future is unknown, we train a policy conditioned only on past accesses that accurately approximates Belady’s even on diverse and complex access patterns, and call this approach Parrot. When evaluated on 13 of the most memory-intensive SPEC applications, Parrot increases cache miss rates by 20% over the current state of the art. In addition, on a large-scale web search benchmark, Parrot increases cache hit rates by 61% over a conventional LRU policy. We release a Gym environment to facilitate research in this area, as data is plentiful, and further advancements can have significant real-world impact.

The web provides a rich, open-domain environment with textual, structural, and spatial properties.We propose a new task for grounding language in this environment: given a natural language command (e.g., "click on the second article"), choose the correct element on the web page (e.g., a hyperlink or text box).We collected a dataset of over 50,000 commands that capture various phenomena such as functional references (e.g."find who made this site"), relational reasoning (e.g."article by john"), and visual reasoning (e.g."topmost article").We also implemented and analyzed three baseline models that capture different phenomena present in the dataset.

Our goal is to learn a semantic parser that maps natural language utterances into executable programs when only indirect supervision is available: examples are labeled with the correct execution result, but not the program itself.Consequently, we must search the space of programs for those that output the correct result, while not being misled by spurious programs: incorrect programs that coincidentally output the correct result.We connect two common learning paradigms, reinforcement learning (RL) and maximum marginal likelihood (MML), and then present a new learning algorithm that combines the strengths of both.The new algorithm guards against spurious programs by combining the systematic search traditionally employed in MML with the randomized exploration of RL, and by updating parameters such that probability is spread more evenly across consistent programs.We apply our learning algorithm to a new neural semantic parser and show significant gains over existing state-of-theart results on a recent context-dependent semantic parsing task.Spurious: move(hasShirt(red), 1) Correct: move(hasHat(yellow), leftOf(hasShirt(blue)))Spurious: move(hasShirt(red), 1) Correct: move(hasHat(yellow), leftOf(hasShirt(blue)))