Victor Zhong

Alex Mallen, Akari Asai, Victor Zhong et al. · uw

Despite their impressive performance on diverse tasks, large language models (LMs) still struggle with tasks requiring rich world knowledge, implying the limitations of relying solely on their parameters to encode a wealth of world knowledge. This paper aims to understand LMs' strengths and limitations in memorizing factual knowledge, by conducting large-scale knowledge probing experiments of 10 models and 4 augmentation methods on PopQA, our new open-domain QA dataset with 14k questions. We find that LMs struggle with less popular factual knowledge, and that scaling fails to appreciably improve memorization of factual knowledge in the long tail. We then show that retrieval-augmented LMs largely outperform orders of magnitude larger LMs, while unassisted LMs remain competitive in questions about high-popularity entities. Based on those findings, we devise a simple, yet effective, method for powerful and efficient retrieval-augmented LMs, which retrieves non-parametric memories only when necessary. Experimental results show that this significantly improves models' performance while reducing the inference costs.

Machel Reid, Victor Zhong, Suchin Gururangan et al. · allen-ai, deepmind

We present M2D2, a fine-grained, massively multi-domain corpus for studying domain adaptation in language models (LMs). M2D2 consists of 8.5B tokens and spans 145 domains extracted from Wikipedia and Semantic Scholar. Using ontologies derived from Wikipedia and ArXiv categories, we organize the domains in each data source into 22 groups. This two-level hierarchy enables the study of relationships between domains and their effects on in- and out-of-domain performance after adaptation. We also present a number of insights into the nature of effective domain adaptation in LMs, as examples of the new types of studies M2D2 enables. To improve in-domain performance, we show the benefits of adapting the LM along a domain hierarchy; adapting to smaller amounts of fine-grained domain-specific data can lead to larger in-domain performance gains than larger amounts of weakly relevant data. We further demonstrate a trade-off between in-domain specialization and out-of-domain generalization within and across ontologies, as well as a strong correlation between out-of-domain performance and lexical overlap between domains.

Victor Zhong, Jesse Mu, Luke Zettlemoyer et al. · stanford, uw

Recent work has shown that augmenting environments with language descriptions improves policy learning. However, for environments with complex language abstractions, learning how to ground language to observations is difficult due to sparse, delayed rewards. We propose Language Dynamics Distillation (LDD), which pretrains a model to predict environment dynamics given demonstrations with language descriptions, and then fine-tunes these language-aware pretrained representations via reinforcement learning (RL). In this way, the model is trained to both maximize expected reward and retain knowledge about how language relates to environment dynamics. On SILG, a benchmark of five tasks with language descriptions that evaluate distinct generalization challenges on unseen environments (NetHack, ALFWorld, RTFM, Messenger, and Touchdown), LDD outperforms tabula-rasa RL, VAE pretraining, and methods that learn from unlabeled demonstrations in inverse RL and reward shaping with pretrained experts. In our analyses, we show that language descriptions in demonstrations improve sample-efficiency and generalization across environments, and that dynamics modelling with expert demonstrations is more effective than with non-experts.

Victor Zhong, Weijia Shi, Wen-tau Yih et al. · uw

We introduce RoMQA, the first benchmark for robust, multi-evidence, multi-answer question answering (QA). RoMQA contains clusters of questions that are derived from related constraints mined from the Wikidata knowledge graph. RoMQA evaluates robustness of QA models to varying constraints by measuring worst-case performance within each question cluster. Compared to prior QA datasets, RoMQA has more human-written questions that require reasoning over more evidence text and have, on average, many more correct answers. In addition, human annotators rate RoMQA questions as more natural or likely to be asked by people. We evaluate state-of-the-art large language models in zero-shot, few-shot, and fine-tuning settings, and find that RoMQA is challenging: zero-shot and few-shot models perform similarly to naive baselines, while supervised retrieval methods perform well below gold evidence upper bounds. Moreover, existing models are not robust to variations in question constraints, but can be made more robust by tuning on clusters of related questions. Our results show that RoMQA is a challenging benchmark for large language models, and provides a quantifiable test to build more robust QA methods.

Tianbao Xie, Siheng Zhao, Chen Henry Wu et al. · cmu

Designing reward functions is a longstanding challenge in reinforcement learning (RL); it requires specialized knowledge or domain data, leading to high costs for development. To address this, we introduce Text2Reward, a data-free framework that automates the generation and shaping of dense reward functions based on large language models (LLMs). Given a goal described in natural language, Text2Reward generates shaped dense reward functions as an executable program grounded in a compact representation of the environment. Unlike inverse RL and recent work that uses LLMs to write sparse reward codes or unshaped dense rewards with a constant function across timesteps, Text2Reward produces interpretable, free-form dense reward codes that cover a wide range of tasks, utilize existing packages, and allow iterative refinement with human feedback. We evaluate Text2Reward on two robotic manipulation benchmarks (ManiSkill2, MetaWorld) and two locomotion environments of MuJoCo. On 13 of the 17 manipulation tasks, policies trained with generated reward codes achieve similar or better task success rates and convergence speed than expert-written reward codes. For locomotion tasks, our method learns six novel locomotion behaviors with a success rate exceeding 94%. Furthermore, we show that the policies trained in the simulator with our method can be deployed in the real world. Finally, Text2Reward further improves the policies by refining their reward functions with human feedback. Video results are available at https://text-to-reward.github.io/ .

Ruisheng Cao, Fangyu Lei, Haoyuan Wu et al. · tsinghua

Data science and engineering workflows often span multiple stages, from warehousing to orchestration, using tools like BigQuery, dbt, and Airbyte. As vision language models (VLMs) advance in multimodal understanding and code generation, VLM-based agents could potentially automate these workflows by generating SQL queries, Python code, and GUI operations. This automation can improve the productivity of experts while democratizing access to large-scale data analysis. In this paper, we introduce Spider2-V, the first multimodal agent benchmark focusing on professional data science and engineering workflows, featuring 494 real-world tasks in authentic computer environments and incorporating 20 enterprise-level professional applications. These tasks, derived from real-world use cases, evaluate the ability of a multimodal agent to perform data-related tasks by writing code and managing the GUI in enterprise data software systems. To balance realistic simulation with evaluation simplicity, we devote significant effort to developing automatic configurations for task setup and carefully crafting evaluation metrics for each task. Furthermore, we supplement multimodal agents with comprehensive documents of these enterprise data software systems. Our empirical evaluation reveals that existing state-of-the-art LLM/VLM-based agents do not reliably automate full data workflows (14.0% success). Even with step-by-step guidance, these agents still underperform in tasks that require fine-grained, knowledge-intensive GUI actions (16.2%) and involve remote cloud-hosted workspaces (10.6%). We hope that Spider2-V paves the way for autonomous multimodal agents to transform the automation of data science and engineering workflow. Our code and data are available at https://spider2-v.github.io.

Tianbao Xie, Danyang Zhang, Jixuan Chen et al.

Autonomous agents that accomplish complex computer tasks with minimal human interventions have the potential to transform human-computer interaction, significantly enhancing accessibility and productivity. However, existing benchmarks either lack an interactive environment or are limited to environments specific to certain applications or domains, failing to reflect the diverse and complex nature of real-world computer use, thereby limiting the scope of tasks and agent scalability. To address this issue, we introduce OSWorld, the first-of-its-kind scalable, real computer environment for multimodal agents, supporting task setup, execution-based evaluation, and interactive learning across various operating systems such as Ubuntu, Windows, and macOS. OSWorld can serve as a unified, integrated computer environment for assessing open-ended computer tasks that involve arbitrary applications. Building upon OSWorld, we create a benchmark of 369 computer tasks involving real web and desktop apps in open domains, OS file I/O, and workflows spanning multiple applications. Each task example is derived from real-world computer use cases and includes a detailed initial state setup configuration and a custom execution-based evaluation script for reliable, reproducible evaluation. Extensive evaluation of state-of-the-art LLM/VLM-based agents on OSWorld reveals significant deficiencies in their ability to serve as computer assistants. While humans can accomplish over 72.36% of the tasks, the best model achieves only 12.24% success, primarily struggling with GUI grounding and operational knowledge. Comprehensive analysis using OSWorld provides valuable insights for developing multimodal generalist agents that were not possible with previous benchmarks. Our code, environment, baseline models, and data are publicly available at https://os-world.github.io.

Fangyu Lei, Jixuan Chen, Yuxiao Ye et al. · tsinghua

Real-world enterprise text-to-SQL workflows often involve complex cloud or local data across various database systems, multiple SQL queries in various dialects, and diverse operations from data transformation to analytics. We introduce Spider 2.0, an evaluation framework comprising 632 real-world text-to-SQL workflow problems derived from enterprise-level database use cases. The databases in Spider 2.0 are sourced from real data applications, often containing over 1,000 columns and stored in local or cloud database systems such as BigQuery and Snowflake. We show that solving problems in Spider 2.0 frequently requires understanding and searching through database metadata, dialect documentation, and even project-level codebases. This challenge calls for models to interact with complex SQL workflow environments, process extremely long contexts, perform intricate reasoning, and generate multiple SQL queries with diverse operations, often exceeding 100 lines, which goes far beyond traditional text-to-SQL challenges. Our evaluations indicate that based on o1-preview, our code agent framework successfully solves only 21.3% of the tasks, compared with 91.2% on Spider 1.0 and 73.0% on BIRD. Our results on Spider 2.0 show that while language models have demonstrated remarkable performance in code generation -- especially in prior text-to-SQL benchmarks -- they require significant improvement in order to achieve adequate performance for real-world enterprise usage. Progress on Spider 2.0 represents crucial steps towards developing intelligent, autonomous, code agents for real-world enterprise settings. Our code, baseline models, and data are available at https://spider2-sql.github.io

Zijian Chen, Xueguang Ma, Shengyao Zhuang et al.

Deep Research agents are rapidly emerging as primary consumers of modern retrieval systems. Unlike human users who issue and refine queries without documenting their intermediate thought processes, Deep Research agents generate explicit natural language reasoning before each search call, revealing rich intent and contextual information that existing retrievers entirely ignore. To exploit this overlooked signal, we introduce: (1) Reasoning-Aware Retrieval, a retrieval paradigm that jointly embeds the agent's reasoning trace alongside its query; and (2) DR-Synth, a data synthesis method that generates Deep Research retriever training data from standard QA datasets. We demonstrate that both components are independently effective, and their combination yields a trained embedding model, AgentIR-4B, with substantial gains. On the challenging BrowseComp-Plus benchmark, AgentIR-4B achieves 68\% accuracy with the open-weight agent Tongyi-DeepResearch, compared to 50\% with conventional embedding models twice its size, and 37\% with BM25. Code and data are available at: https://texttron.github.io/AgentIR/.

Arthur Chen, Zuxin Liu, Jianguo Zhang et al.

Large language model (LLM)-based agents struggle to generalize to novel and complex environments, such as unseen websites or new sets of functions, due to a fundamental mismatch between their pre-training and test-time conditions. This challenge stems from two distinct failure modes: a syntactic misunderstanding of environment-specific components like observation formats, and a semantic misunderstanding of state-transition dynamics, which are only revealed at test time. To address these issues, we propose two distinct and complementary strategies for adapting LLM agents by leveraging environment-specific information available during deployment. First, an online distributional adaptation method parameterizes environmental nuances by learning a lightweight adaptation vector that biases the model's output distribution, enabling rapid alignment with an environment response format. Second, a deployment-time dynamics grounding method employs a persona-driven exploration phase to systematically probe and learn the environment's causal dynamics before task execution, equipping the agent with a nonparametric world model. We evaluate these strategies across diverse agentic benchmarks, including function calling and web navigation. Our empirical results show the effectiveness of both strategies across all benchmarks with minimal computational cost. We find that dynamics grounding is particularly effective in complex environments where unpredictable dynamics pose a major obstacle, demonstrating a robust path toward more generalizable and capable LLM-based agents. For example, on the WebArena multi-site split, this method increases the agent's success rate from 2% to 23%.

Zhengyuan Dong, Victor Zhong, Renée J. Miller

We present ModelTables, a benchmark of tables in Model Lakes that captures the structured semantics of performance and configuration tables often overlooked by text only retrieval. The corpus is built from Hugging Face model cards, GitHub READMEs, and referenced papers, linking each table to its surrounding model and publication context. Compared with open data lake tables, model tables are smaller yet exhibit denser inter table relationships, reflecting tightly coupled model and benchmark evolution. The current release covers over 60K models and 90K tables. To evaluate model and table relatedness, we construct a multi source ground truth using three complementary signals: (1) paper citation links, (2) explicit model card links and inheritance, and (3) shared training datasets. We present one extensive empirical use case for the benchmark which is table search. We compare canonical Data Lake search operators (unionable, joinable, keyword) and Information Retrieval baselines (dense, sparse, hybrid retrieval) on this benchmark. Union based semantic table retrieval attains 54.8 % P@1 overall (54.6 % on citation, 31.3 % on inheritance, 30.6 % on shared dataset signals); table based dense retrieval reaches 66.5 % P@1, and metadata hybrid retrieval achieves 54.1 %. This evaluation indicates clear room for developing better table search methods. By releasing ModelTables and its creation protocol, we provide the first large scale benchmark of structured data describing AI model. Our use case of table discovery in Model Lakes, provides intuition and evidence for developing more accurate semantic retrieval, structured comparison, and principled organization of structured model knowledge. Source code, data, and other artifacts have been made available at https://github.com/RJMillerLab/ModelTables.

Xinyuan Wang, Bowen Wang, Dunjie Lu et al. · cmu

Vision-language models have demonstrated impressive capabilities as computer-use agents (CUAs) capable of automating diverse computer tasks. As their commercial potential grows, critical details of the most capable CUA systems remain closed. As these agents will increasingly mediate digital interactions and execute consequential decisions on our behalf, the research community needs access to open CUA frameworks to study their capabilities, limitations, and risks. To bridge this gap, we propose OpenCUA, a comprehensive open-source framework for scaling CUA data and foundation models. Our framework consists of: (1) an annotation infrastructure that seamlessly captures human computer-use demonstrations; (2) AgentNet, the first large-scale computer-use task dataset spanning 3 operating systems and 200+ applications and websites; (3) a scalable pipeline that transforms demonstrations into state-action pairs with reflective long Chain-of-Thought reasoning that sustain robust performance gains as data scales. Our end-to-end agent models demonstrate strong performance across CUA benchmarks. In particular, OpenCUA-72B achieves an average success rate of 45.0% on OSWorld-Verified, establishing a new state-of-the-art (SOTA) among open-source models. Further analysis confirms that our approach generalizes well across domains and benefits significantly from increased test-time computation. We release our annotation tool, datasets, code, and models to build open foundations for further CUA research.

Sitao Cheng, Xunjian Yin, Ruiwen Zhou et al.

The mechanism by which RL contributes to reasoning capabilities-whether it incentivizes the synthesis of new skills or merely amplifies existing behaviors-remains a subject of intense debate. In this work, we investigate this question through the lens of Complementary Reasoning, a complex task that requires integrating internal parametric knowledge with external contextual information. Using a controlled synthetic dataset of human biographies, we strictly decouple this ability into two atomic skills: Parametric Reasoning (relying on internal knowledge) and Contextual Reasoning (depending on external information). To rigorously assess capability boundaries, we evaluate generalization across three distinct levels of difficulty: I.I.D., Composition, and Zero-shot settings. We find that while SFT is sufficient for in-distribution performance, it struggles with O.O.D. generalization, particularly in Zero-shot settings where relational combinations are novel. Crucially, we identify the SFT Generalization Paradox: Models supervised solely on the composite task achieve near-perfect in-distribution accuracy but collapse on out-of-distribution generalization, indicating their reliance on rote memorization of path shortcuts. In contrast, we find that RL acts as a reasoning synthesizer rather than a probability amplifier. However, we uncover a strict atomic prerequisite: RL can only synthesize these complex strategies if the base model has first mastered the independent atomic skills (Parametric and Contextual) via SFT. These findings challenge the view of RL as a mere amplifier, suggesting that given sufficient atomic foundations, RL can actively synthesize complex reasoning strategies from learned primitives without explicit supervision on such complex strategies. This indicates that decoupled atomic training followed by RL offers a scalable path to generalization for complex reasoning tasks.

Benjamin Schneider, Xavier Schneider, Victor Zhong et al.

Long-horizon embodied tasks remain a fundamental challenge in AI, as current methods rely on hand-engineered rewards or action-labeled demonstrations, neither of which scales. We introduce ASH, an agentic system that learns an embodied policy from unlabeled, noisy internet video, without reward shaping or expert annotation. ASH follows a self-improvement loop; when it gets stuck, ASH learns an Inverse Dynamics Model (IDM) from its own trajectories, and uses its IDM to extract supervision from relevant internet video. ASH uses unsupervised learning to identify key moments from large-scale internet video and retains them as long-term memory -- allowing it to tackle long-horizon problems. We evaluate ASH on two complementary environments demanding multi-hour planning: Pokemon Emerald, a turn-based RPG, and The Legend of Zelda: The Minish Cap, a real-time action-adventure game. In both games, behavioral cloning, retrieval-augmented and zero-shot foundation-model baselines plateau, while ASH sustains progression across our 8-hour evaluation. ASH reaches an average of $11.2/12$ milestones in Pokemon Emerald and $9.9/12$ in Legend of Zelda, while the strongest baseline gets stuck in both environments at an average of $6.5/12$ and $6.0/12$ milestones, respectively. We demonstrate that self-improving agents are a scalable recipe for long-horizon embodied learning.

Tianbao Xie, Chen Henry Wu, Peng Shi et al.

Structured knowledge grounding (SKG) leverages structured knowledge to complete user requests, such as semantic parsing over databases and question answering over knowledge bases. Since the inputs and outputs of SKG tasks are heterogeneous, they have been studied separately by different communities, which limits systematic and compatible research on SKG. In this paper, we overcome this limitation by proposing the UnifiedSKG framework, which unifies 21 SKG tasks into a text-to-text format, aiming to promote systematic SKG research, instead of being exclusive to a single task, domain, or dataset. We use UnifiedSKG to benchmark T5 with different sizes and show that T5, with simple modifications when necessary, achieves state-of-the-art performance on almost all of the 21 tasks. We further demonstrate that multi-task prefix-tuning improves the performance on most tasks, largely improving the overall performance. UnifiedSKG also facilitates the investigation of zero-shot and few-shot learning, and we show that T0, GPT-3, and Codex struggle in zero-shot and few-shot learning for SKG. We also use UnifiedSKG to conduct a series of controlled experiments on structured knowledge encoding variants across SKG tasks. UnifiedSKG is easily extensible to more tasks, and it is open-sourced at https://github.com/hkunlp/unifiedskg.

Victor Zhong, Luke Zettlemoyer

Conversational machine reading systems help users answer high-level questions (e.g. determine if they qualify for particular government benefits) when they do not know the exact rules by which the determination is made(e.g. whether they need certain income levels or veteran status). The key challenge is that these rules are only provided in the form of a procedural text (e.g. guidelines from government website) which the system must read to figure out what to ask the user. We present a new conversational machine reading model that jointly extracts a set of decision rules from the procedural text while reasoning about which are entailed by the conversational history and which still need to be edited to create questions for the user. On the recently introduced ShARC conversational machine reading dataset, our Entailment-driven Extract and Edit network (E3) achieves a new state-of-the-art, outperforming existing systems as well as a new BERT-based baseline. In addition, by explicitly highlighting which information still needs to be gathered, E3 provides a more explainable alternative to prior work. We release source code for our models and experiments at https://github.com/vzhong/e3.

Victor Zhong, Dipendra Misra, Xingdi Yuan et al. · microsoft-research

We introduce Language Feedback Models (LFMs) that identify desirable behaviour - actions that help achieve tasks specified in the instruction - for imitation learning in instruction following. To train LFMs, we obtain feedback from Large Language Models (LLMs) on visual trajectories verbalized to language descriptions. First, by using LFMs to identify desirable behaviour to imitate, we improve in task-completion rate over strong behavioural cloning baselines on three distinct language grounding environments (Touchdown, ScienceWorld, and ALFWorld). Second, LFMs outperform using LLMs as experts to directly predict actions, when controlling for the number of LLM output tokens. Third, LFMs generalize to unseen environments, improving task-completion rate by 3.5-12.0% through one round of adaptation. Finally, LFM can be modified to provide human-interpretable feedback without performance loss, allowing human verification of desirable behaviour for imitation learning.

Arthur Chen, Victor Zhong

We introduce and formalize the Synthetic Dataset Quality Estimation (SynQuE) problem: ranking synthetic datasets by their expected real-world task performance using only limited unannotated real data. This addresses a critical and open challenge where data is scarce due to collection costs or privacy constraints. We establish the first comprehensive benchmarks for this problem by introducing and evaluating proxy metrics that choose synthetic data for training to maximize task performance on real data. We introduce the first proxy metrics for SynQuE by adapting distribution and diversity-based distance measures to our context via embedding models. To address the shortcomings of these metrics on complex planning tasks, we propose LENS, a novel proxy that leverages large language model reasoning. Our results show that SynQuE proxies correlate with real task performance across diverse tasks, including sentiment analysis, Text2SQL, web navigation, and image classification, with LENS consistently outperforming others on complex tasks by capturing nuanced characteristics. For instance, on text-to-SQL parsing, training on the top-3 synthetic datasets selected via SynQuE proxies can raise accuracy from 30.4% to 38.4 (+8.1)% on average compared to selecting data indiscriminately. This work establishes SynQuE as a practical framework for synthetic data selection under real-data scarcity and motivates future research on foundation model-based data characterization and fine-grained data selection.

Yuxuan Li, Victor Zhong

Learning to plan in grounded environments typically requires carefully designed reward functions or high-quality annotated demonstrations. Recent works show that pretrained foundation models, such as large language models (LLMs) and vision language models (VLMs), capture background knowledge helpful for planning, which reduces the amount of reward design and demonstrations needed for policy learning. We evaluate how well LLMs and VLMs provide feedback across symbolic, language, and continuous control environments. We consider prominent types of feedback for planning including binary feedback, preference feedback, action advising, goal advising, and delta action feedback. We also consider inference methods that impact feedback performance, including in-context learning, chain-of-thought, and access to environment dynamics. We find that foundation models can provide diverse high-quality feedback across domains. Moreover, larger and reasoning models consistently provide more accurate feedback, exhibit less bias, and benefit more from enhanced inference methods. Finally, feedback quality degrades for environments with complex dynamics or continuous state spaces and action spaces.

Jesse Mu, Victor Zhong, Roberta Raileanu et al.

Reinforcement learning (RL) agents are particularly hard to train when rewards are sparse. One common solution is to use intrinsic rewards to encourage agents to explore their environment. However, recent intrinsic exploration methods often use state-based novelty measures which reward low-level exploration and may not scale to domains requiring more abstract skills. Instead, we explore natural language as a general medium for highlighting relevant abstractions in an environment. Unlike previous work, we evaluate whether language can improve over existing exploration methods by directly extending (and comparing to) competitive intrinsic exploration baselines: AMIGo (Campero et al., 2021) and NovelD (Zhang et al., 2021). These language-based variants outperform their non-linguistic forms by 47-85% across 13 challenging tasks from the MiniGrid and MiniHack environment suites.

Victor Zhong, Austin W. Hanjie, Sida I. Wang et al.

Existing work in language grounding typically study single environments. How do we build unified models that apply across multiple environments? We propose the multi-environment Symbolic Interactive Language Grounding benchmark (SILG), which unifies a collection of diverse grounded language learning environments under a common interface. SILG consists of grid-world environments that require generalization to new dynamics, entities, and partially observed worlds (RTFM, Messenger, NetHack), as well as symbolic counterparts of visual worlds that require interpreting rich natural language with respect to complex scenes (ALFWorld, Touchdown). Together, these environments provide diverse grounding challenges in richness of observation space, action space, language specification, and plan complexity. In addition, we propose the first shared model architecture for RL on these environments, and evaluate recent advances such as egocentric local convolution, recurrent state-tracking, entity-centric attention, and pretrained LM using SILG. Our shared architecture achieves comparable performance to environment-specific architectures. Moreover, we find that many recent modelling advances do not result in significant gains on environments other than the one they were designed for. This highlights the need for a multi-environment benchmark. Finally, the best models significantly underperform humans on SILG, which suggests ample room for future work. We hope SILG enables the community to quickly identify new methodologies for language grounding that generalize to a diverse set of environments and their associated challenges.

Machel Reid, Victor Zhong

Many types of text style transfer can be achieved with only small, precise edits (e.g. sentiment transfer from I had a terrible time... to I had a great time...). We propose a coarse-to-fine editor for style transfer that transforms text using Levenshtein edit operations (e.g. insert, replace, delete). Unlike prior single-span edit methods, our method concurrently edits multiple spans in the source text. To train without parallel style text pairs (e.g. pairs of +/- sentiment statements), we propose an unsupervised data synthesis procedure. We first convert text to style-agnostic templates using style classifier attention (e.g. I had a SLOT time...), then fill in slots in these templates using fine-tuned pretrained language models. Our method outperforms existing generation and editing style transfer methods on sentiment (Yelp, Amazon) and politeness (Polite) transfer. In particular, multi-span editing achieves higher performance and more diverse output than single-span editing. Moreover, compared to previous methods on unsupervised data synthesis, our method results in higher quality parallel style pairs and improves model performance.

Austin W. Hanjie, Victor Zhong, Karthik Narasimhan

We investigate the use of natural language to drive the generalization of control policies and introduce the new multi-task environment Messenger with free-form text manuals describing the environment dynamics. Unlike previous work, Messenger does not assume prior knowledge connecting text and state observations $-$ the control policy must simultaneously ground the game manual to entity symbols and dynamics in the environment. We develop a new model, EMMA (Entity Mapper with Multi-modal Attention) which uses an entity-conditioned attention module that allows for selective focus over relevant descriptions in the manual for each entity in the environment. EMMA is end-to-end differentiable and learns a latent grounding of entities and dynamics from text to observations using only environment rewards. EMMA achieves successful zero-shot generalization to unseen games with new dynamics, obtaining a 40% higher win rate compared to multiple baselines. However, win rate on the hardest stage of Messenger remains low (10%), demonstrating the need for additional work in this direction.

Victor Zhong, Mike Lewis, Sida I. Wang et al.

We propose Grounded Adaptation for Zero-shot Executable Semantic Parsing (GAZP) to adapt an existing semantic parser to new environments (e.g. new database schemas). GAZP combines a forward semantic parser with a backward utterance generator to synthesize data (e.g. utterances and SQL queries) in the new environment, then selects cycle-consistent examples to adapt the parser. Unlike data-augmentation, which typically synthesizes unverified examples in the training environment, GAZP synthesizes examples in the new environment whose input-output consistency are verified. On the Spider, Sparc, and CoSQL zero-shot semantic parsing tasks, GAZP improves logical form and execution accuracy of the baseline parser. Our analyses show that GAZP outperforms data-augmentation in the training environment, performance increases with the amount of GAZP-synthesized data, and cycle-consistency is central to successful adaptation.

Victor Zhong, Tim Rocktäschel, Edward Grefenstette

Obtaining policies that can generalise to new environments in reinforcement learning is challenging. In this work, we demonstrate that language understanding via a reading policy learner is a promising vehicle for generalisation to new environments. We propose a grounded policy learning problem, Read to Fight Monsters (RTFM), in which the agent must jointly reason over a language goal, relevant dynamics described in a document, and environment observations. We procedurally generate environment dynamics and corresponding language descriptions of the dynamics, such that agents must read to understand new environment dynamics instead of memorising any particular information. In addition, we propose txt2$π$, a model that captures three-way interactions between the goal, document, and observations. On RTFM, txt2$π$ generalises to new environments with dynamics not seen during training via reading. Furthermore, our model outperforms baselines such as FiLM and language-conditioned CNNs on RTFM. Through curriculum learning, txt2$π$ produces policies that excel on complex RTFM tasks requiring several reasoning and coreference steps.

Sewon Min, Victor Zhong, Luke Zettlemoyer et al.

Multi-hop Reading Comprehension (RC) requires reasoning and aggregation across several paragraphs. We propose a system for multi-hop RC that decomposes a compositional question into simpler sub-questions that can be answered by off-the-shelf single-hop RC models. Since annotations for such decomposition are expensive, we recast sub-question generation as a span prediction problem and show that our method, trained using only 400 labeled examples, generates sub-questions that are as effective as human-authored sub-questions. We also introduce a new global rescoring approach that considers each decomposition (i.e. the sub-questions and their answers) to select the best final answer, greatly improving overall performance. Our experiments on HotpotQA show that this approach achieves the state-of-the-art results, while providing explainable evidence for its decision making in the form of sub-questions.

Victor Zhong, Caiming Xiong, Nitish Shirish Keskar et al.

End-to-end neural models have made significant progress in question answering, however recent studies show that these models implicitly assume that the answer and evidence appear close together in a single document. In this work, we propose the Coarse-grain Fine-grain Coattention Network (CFC), a new question answering model that combines information from evidence across multiple documents. The CFC consists of a coarse-grain module that interprets documents with respect to the query then finds a relevant answer, and a fine-grain module which scores each candidate answer by comparing its occurrences across all of the documents with the query. We design these modules using hierarchies of coattention and self-attention, which learn to emphasize different parts of the input. On the Qangaroo WikiHop multi-evidence question answering task, the CFC obtains a new state-of-the-art result of 70.6% on the blind test set, outperforming the previous best by 3% accuracy despite not using pretrained contextual encoders.

Sewon Min, Victor Zhong, Richard Socher et al.

Neural models for question answering (QA) over documents have achieved significant performance improvements. Although effective, these models do not scale to large corpora due to their complex modeling of interactions between the document and the question. Moreover, recent work has shown that such models are sensitive to adversarial inputs. In this paper, we study the minimal context required to answer the question, and find that most questions in existing datasets can be answered with a small set of sentences. Inspired by this observation, we propose a simple sentence selector to select the minimal set of sentences to feed into the QA model. Our overall system achieves significant reductions in training (up to 15 times) and inference times (up to 13 times), with accuracy comparable to or better than the state-of-the-art on SQuAD, NewsQA, TriviaQA and SQuAD-Open. Furthermore, our experimental results and analyses show that our approach is more robust to adversarial inputs.

Victor Zhong, Caiming Xiong, Richard Socher

Dialogue state tracking, which estimates user goals and requests given the dialogue context, is an essential part of task-oriented dialogue systems. In this paper, we propose the Global-Locally Self-Attentive Dialogue State Tracker (GLAD), which learns representations of the user utterance and previous system actions with global-local modules. Our model uses global modules to share parameters between estimators for different types (called slots) of dialogue states, and uses local modules to learn slot-specific features. We show that this significantly improves tracking of rare states and achieves state-of-the-art performance on the WoZ and DSTC2 state tracking tasks. GLAD obtains 88.1% joint goal accuracy and 97.1% request accuracy on WoZ, outperforming prior work by 3.7% and 5.5%. On DSTC2, our model obtains 74.5% joint goal accuracy and 97.5% request accuracy, outperforming prior work by 1.1% and 1.0%.

Caiming Xiong, Victor Zhong, Richard Socher

Traditional models for question answering optimize using cross entropy loss, which encourages exact answers at the cost of penalizing nearby or overlapping answers that are sometimes equally accurate. We propose a mixed objective that combines cross entropy loss with self-critical policy learning. The objective uses rewards derived from word overlap to solve the misalignment between evaluation metric and optimization objective. In addition to the mixed objective, we improve dynamic coattention networks (DCN) with a deep residual coattention encoder that is inspired by recent work in deep self-attention and residual networks. Our proposals improve model performance across question types and input lengths, especially for long questions that requires the ability to capture long-term dependencies. On the Stanford Question Answering Dataset, our model achieves state-of-the-art results with 75.1% exact match accuracy and 83.1% F1, while the ensemble obtains 78.9% exact match accuracy and 86.0% F1.

Victor Zhong, Caiming Xiong, Richard Socher

A significant amount of the world's knowledge is stored in relational databases. However, the ability for users to retrieve facts from a database is limited due to a lack of understanding of query languages such as SQL. We propose Seq2SQL, a deep neural network for translating natural language questions to corresponding SQL queries. Our model leverages the structure of SQL queries to significantly reduce the output space of generated queries. Moreover, we use rewards from in-the-loop query execution over the database to learn a policy to generate unordered parts of the query, which we show are less suitable for optimization via cross entropy loss. In addition, we will publish WikiSQL, a dataset of 80654 hand-annotated examples of questions and SQL queries distributed across 24241 tables from Wikipedia. This dataset is required to train our model and is an order of magnitude larger than comparable datasets. By applying policy-based reinforcement learning with a query execution environment to WikiSQL, our model Seq2SQL outperforms attentional sequence to sequence models, improving execution accuracy from 35.9% to 59.4% and logical form accuracy from 23.4% to 48.3%.

Caiming Xiong, Victor Zhong, Richard Socher

Several deep learning models have been proposed for question answering. However, due to their single-pass nature, they have no way to recover from local maxima corresponding to incorrect answers. To address this problem, we introduce the Dynamic Coattention Network (DCN) for question answering. The DCN first fuses co-dependent representations of the question and the document in order to focus on relevant parts of both. Then a dynamic pointing decoder iterates over potential answer spans. This iterative procedure enables the model to recover from initial local maxima corresponding to incorrect answers. On the Stanford question answering dataset, a single DCN model improves the previous state of the art from 71.0% F1 to 75.9%, while a DCN ensemble obtains 80.4% F1.

Ankit Kumar, Ozan Irsoy, Peter Ondruska et al.

Most tasks in natural language processing can be cast into question answering (QA) problems over language input. We introduce the dynamic memory network (DMN), a neural network architecture which processes input sequences and questions, forms episodic memories, and generates relevant answers. Questions trigger an iterative attention process which allows the model to condition its attention on the inputs and the result of previous iterations. These results are then reasoned over in a hierarchical recurrent sequence model to generate answers. The DMN can be trained end-to-end and obtains state-of-the-art results on several types of tasks and datasets: question answering (Facebook's bAbI dataset), text classification for sentiment analysis (Stanford Sentiment Treebank) and sequence modeling for part-of-speech tagging (WSJ-PTB). The training for these different tasks relies exclusively on trained word vector representations and input-question-answer triplets.