Estevam Hruschka

Bosung Kim, Hayate Iso, Nikita Bhutani et al. · nvidia

The task of triplet extraction aims to extract pairs of entities and their corresponding relations from unstructured text. Most existing methods train an extraction model on training data involving specific target relations, and are incapable of extracting new relations that were not observed at training time. Generalizing the model to unseen relations typically requires fine-tuning on synthetic training data which is often noisy and unreliable. We show that by reducing triplet extraction to a template infilling task over a pre-trained language model (LM), we can equip the extraction model with zero-shot learning capabilities and eliminate the need for additional training data. We propose a novel framework, ZETT (ZEro-shot Triplet extraction by Template infilling), that aligns the task objective to the pre-training objective of generative transformers to generalize to unseen relations. Experiments on FewRel and Wiki-ZSL datasets demonstrate that ZETT shows consistent and stable performance, outperforming previous state-of-the-art methods, even when using automatically generated templates. https://github.com/megagonlabs/zett/

Vishwas Mruthyunjaya, Pouya Pezeshkpour, Estevam Hruschka et al. · cmu

Symbolic knowledge graphs (KGs) play a pivotal role in knowledge-centric applications such as search, question answering and recommendation. As contemporary language models (LMs) trained on extensive textual data have gained prominence, researchers have extensively explored whether the parametric knowledge within these models can match up to that present in knowledge graphs. Various methodologies have indicated that enhancing the size of the model or the volume of training data enhances its capacity to retrieve symbolic knowledge, often with minimal or no human supervision. Despite these advancements, there is a void in comprehensively evaluating whether LMs can encompass the intricate topological and semantic attributes of KGs, attributes crucial for reasoning processes. In this work, we provide an exhaustive evaluation of language models of varying sizes and capabilities. We construct nine qualitative benchmarks that encompass a spectrum of attributes including symmetry, asymmetry, hierarchy, bidirectionality, compositionality, paths, entity-centricity, bias and ambiguity. Additionally, we propose novel evaluation metrics tailored for each of these attributes. Our extensive evaluation of various LMs shows that while these models exhibit considerable potential in recalling factual information, their ability to capture intricate topological and semantic traits of KGs remains significantly constrained. We note that our proposed evaluation metrics are more reliable in evaluating these abilities than the existing metrics. Lastly, some of our benchmarks challenge the common notion that larger LMs (e.g., GPT-4) universally outshine their smaller counterparts (e.g., BERT).

Yunshu Wu, Hayate Iso, Pouya Pezeshkpour et al. · nvidia

Large Language Models (LLMs) have shown promising performance in summary evaluation tasks, yet they face challenges such as high computational costs and the Lost-in-the-Middle problem where important information in the middle of long documents is often overlooked. To address these issues, this paper introduces a novel approach, Extract-then-Evaluate, which involves extracting key sentences from a long source document and then evaluating the summary by prompting LLMs. The results reveal that the proposed method not only significantly reduces evaluation costs but also exhibits a higher correlation with human evaluations. Furthermore, we provide practical recommendations for optimal document length and sentence extraction methods, contributing to the development of cost-effective yet more accurate methods for LLM-based text generation evaluation.

Pouya Pezeshkpour, Estevam Hruschka

Large Language Models (LLMs) have demonstrated remarkable capabilities in various NLP tasks. However, previous works have shown these models are sensitive towards prompt wording, and few-shot demonstrations and their order, posing challenges to fair assessment of these models. As these models become more powerful, it becomes imperative to understand and address these limitations. In this paper, we focus on LLMs robustness on the task of multiple-choice questions -- commonly adopted task to study reasoning and fact-retrieving capability of LLMs. Investigating the sensitivity of LLMs towards the order of options in multiple-choice questions, we demonstrate a considerable performance gap of approximately 13% to 75% in LLMs on different benchmarks, when answer options are reordered, even when using demonstrations in a few-shot setting. Through a detailed analysis, we conjecture that this sensitivity arises when LLMs are uncertain about the prediction between the top-2/3 choices, and specific options placements may favor certain prediction between those top choices depending on the question caused by positional bias. We also identify patterns in top-2 choices that amplify or mitigate the model's bias toward option placement. We found that for amplifying bias, the optimal strategy involves positioning the top two choices as the first and last options. Conversely, to mitigate bias, we recommend placing these choices among the adjacent options. To validate our conjecture, we conduct various experiments and adopt two approaches to calibrate LLMs' predictions, leading to up to 8 percentage points improvement across different models and benchmarks.

Zeyu He, Hannah Kim, Dan Zhang et al.

In orchestrated multi-agent systems, humans often struggle to manage plans due to their complexity and limited transparency. Existing approaches rely on outcome-level supervision, where users verify only final outputs without visibility into intermediate reasoning. We formalize a design space for human-LLM co-planning interactions along three axes: mode (semantic vs. structural), scope (global vs. targeted), and level (low vs. high-level edits). We realize it in AMBIPOM, a prototype supporting process-level supervision through both semantic and structural interactions. Through a user study, we characterize how users navigate this space, revealing hybrid workflows and effort-control-risk trade-offs; through a controlled benchmark, we analyze how LLMs revise plans under varying scope and revision strategies. Our findings yield design insights for more transparent, controllable, and effective human-AI co-planning. We release code and data at https://github.com/megagonlabs/ambipom.

Farima Fatahi Bayat, Pouya Pezeshkpour, Estevam Hruschka

Tool-augmented Language Models (TaLMs) can invoke external tools to solve problems beyond their parametric capacity. However, it remains unclear whether these tool-enabled gains reflect trustworthy reasoning. Focusing on the Code Interpreter tool, we show that even when tools are selected and executed correctly, TaLMs treat tool outputs as substitutes for reasoning, producing solutions that appear correct but lack coherent justification. We term this failure mode Tool-Induced Myopia (TIM), and study it using PYMATH, a benchmark of 1,679 competition-level mathematical problems for which Python code is helpful but not sufficient. We further develop a multi-dimensional evaluation suite to quantify reasoning degradation in TaLMs relative to their non-tool counterparts. Our findings reveal that while TaLMs achieve up to a 19.3 percentage point gain in final-answer accuracy, their reasoning behavior consistently deteriorates (e.g., non-tool LLMs win up to 41.5% more often in pairwise comparisons of the reasoning process). This degradation intensifies with tool use; the more frequently a model invokes tools, the less coherent its reasoning becomes. Moreover, tool use shifts errors from arithmetic mistakes toward global reasoning failures (logic, assumption, creativity); with TIM present in ~55% of high-risk cases. Finally, we propose a preference-optimization-based framework that realigns TaLMs to use tools as assistive evidence, improving both final-answer accuracy and reasoning depth under tool use. Codes and data are available at: https://github.com/megagonlabs/TIM.

Pouya Pezeshkpour, Hayate Iso, Thom Lake et al. · nvidia

Numerous HR applications are centered around resumes and job descriptions. While they can benefit from advancements in NLP, particularly large language models, their real-world adoption faces challenges due to absence of comprehensive benchmarks for various HR tasks, and lack of smaller models with competitive capabilities. In this paper, we aim to bridge this gap by introducing the Resume-Job Description Benchmark (RJDB). We meticulously craft this benchmark to cater to a wide array of HR tasks, including matching and explaining resumes to job descriptions, extracting skills and experiences from resumes, and editing resumes. To create this benchmark, we propose to distill domain-specific knowledge from a large language model (LLM). We rely on a curated skill-occupation graph to ensure diversity and provide context for LLMs generation. Our benchmark includes over 50 thousand triples of job descriptions, matched resumes and unmatched resumes. Using RJDB, we train multiple smaller student models. Our experiments reveal that the student models achieve near/better performance than the teacher model (GPT-4), affirming the effectiveness of the benchmark. Additionally, we explore the utility of RJDB on out-of-distribution data for skill extraction and resume-job description matching, in zero-shot and weak supervision manner. We release our datasets and code to foster further research and industry applications.

Moin Aminnaseri, Farima Fatahi Bayat, Nikita Bhutani et al.

NL2SQL systems aim to address the growing need for natural language interaction with data. However, real-world information rarely maps to a single SQL query because (1) users express queries iteratively (2) questions often span multiple data sources beyond the closed-world assumption of a single database, and (3) queries frequently rely on commonsense or external knowledge. Consequently, satisfying realistic data needs require integrating heterogeneous sources, modalities, and contextual data. In this paper, we present Blue's Data Intelligence Layer (DIL) designed to support multi-source, multi-modal, and data-centric applications. Blue is a compound AI system that orchestrates agents and data for enterprise settings. DIL serves as the data intelligence layer for agentic data processing, to bridge the semantic gap between user intent and available information by unifying structured enterprise data, world knowledge accessible through LLMs, and personal context obtained through interaction. At the core of DIL is a data registry that stores metadata for diverse data sources and modalities to enable both native and natural language queries. DIL treats LLMs, the Web, and the User as source 'databases', each with their own query interface, elevating them to first-class data sources. DIL relies on data planners to transform user queries into executable query plans. These plans are declarative abstractions that unify relational operators with other operators spanning multiple modalities. DIL planners support decomposition of complex requests into subqueries, retrieval from diverse sources, and finally reasoning and integration to produce final results. We demonstrate DIL through two interactive scenarios in which user queries dynamically trigger multi-source retrieval, cross-modal reasoning, and result synthesis, illustrating how compound AI systems can move beyond single database NL2SQL.

Dan Zhang, Hannah Kim, Rafael Li Chen et al.

We present MEGAnno, a novel exploratory annotation framework designed for NLP researchers and practitioners. Unlike existing labeling tools that focus on data labeling only, our framework aims to support a broader, iterative ML workflow including data exploration and model development. With MEGAnno's API, users can programmatically explore the data through sophisticated search and automated suggestion functions and incrementally update task schema as their project evolve. Combined with our widget, the users can interactively sort, filter, and assign labels to multiple items simultaneously in the same notebook where the rest of the NLP project resides. We demonstrate MEGAnno's flexible, exploratory, efficient, and seamless labeling experience through a sentiment analysis use case.

Aditi Mishra, Sajjadur Rahman, Hannah Kim et al.

Large language models (LLMs) are proficient at generating fluent text with minimal task-specific supervision. Yet, their ability to provide well-grounded rationalizations for knowledge-intensive tasks remains under-explored. Such tasks, like commonsense multiple-choice questions, require rationales based on world knowledge to support predictions and refute alternate options. We consider the task of generating knowledge-guided rationalization in natural language by using expert-written examples in a few-shot manner. Surprisingly, crowd-workers preferred knowledge-grounded rationales over crowdsourced rationalizations, citing their factuality, sufficiency, and comprehensive refutations. Although LLMs-generated rationales were preferable, further improvements in conciseness and novelty are required. In another study, we show how rationalization of incorrect model predictions erodes humans' trust in LLM-generated rationales. Motivated by these observations, we create a two-stage pipeline to review task predictions and eliminate potential incorrect decisions before rationalization, enabling trustworthy rationale generation.

Sajjadur Rahman, Hannah Kim, Dan Zhang et al.

Human-centered AI workflows involve stakeholders with multiple roles interacting with each other and automated agents to accomplish diverse tasks. In this paper, we call for a holistic view when designing support mechanisms, such as interaction paradigms, interfaces, and systems, for these multifaceted workflows.

Farima Fatahi Bayat, Moin Aminnaseri, Pouya Pezeshkpour et al.

Large language models (LLMs) have become increasingly capable of following instructions and complex reasoning, making prompting a flexible interface for adapting models without parameter updates. Yet prompt design remains labor-intensive and highly sensitive to formatting, phrasing, and instruction order, motivating automated prompt optimization methods that reduce manual effort while preserving inference-time flexibility. However, existing methods often search over prompt candidates or use fixed critique-refine pipelines driven by individual examples or small batches, limiting their ability to capture systematic error patterns and make targeted edits grounded in failure history. We propose Reflective Prompt Tuning (RPT), a framework that uses LLM function calling to simulate the iterative workflow of human prompt engineers. An LLM optimizer calls a diagnostic function that evaluates the target model over an entire optimization set, summarizes recurring failure modes, and returns a structured diagnostic report. The optimizer uses this report, together with an accumulated memory of prior reports, to revise the prompt for the next iteration. RPT further supports confidence-aware optimization by using calibration signals in diagnostic feedback and final prompt selection. Across three reasoning tasks, RPT improves over initial prompts by up to 12.9 points, remains competitive with state of the art, and improves confidence calibration. Our analyses show that RPT is especially effective on multi-hop and mathematical reasoning, producing targeted prompt revisions that align with diagnosed failure patterns and lead to gains in task performance and calibration.

Seiji Maekawa, Moin Aminnaseri, Pouya Pezeshkpour et al.

Dense retrieval systems increasingly need to handle complex queries. In many realistic settings, users express intent through long instructions or task-specific descriptions, while target documents remain relatively simple and static. This asymmetry creates a retrieval mismatch: understanding queries may require strong reasoning and instruction-following, whereas efficient document indexing favors lightweight encoders. Existing retrieval systems often address this mismatch by directly improving the embedding model, but fine-tuning large embedding models to better follow such instructions is computationally expensive, memory-intensive, and operationally burdensome. To address this challenge, we propose Efficient Retrieval Adapter (ERA), a label-efficient framework that trains retrieval adapters in two stages: self-supervised alignment and supervised adaptation. Inspired by the pre-training and supervised fine-tuning stages of LLMs, ERA first aligns the embedding spaces of a large query embedder and a lightweight document embedder, and then uses limited labeled data to adapt the query-side representation, bridging both the representation gap between embedding models and the semantic gap between complex queries and simple documents without re-indexing the corpus. Experiments on the MAIR benchmark, spanning 126 retrieval tasks across 6 domains, show that ERA improves retrieval in low-label settings, outperforms methods that rely on larger amounts of labeled data, and effectively combines stronger query embedders with weaker document embedders across domains.

Catarina G. Belem, Pouya Pezeshkpour, Hayate Iso et al. · nvidia

Although many studies have investigated and reduced hallucinations in large language models (LLMs) for single-document tasks, research on hallucination in multi-document summarization (MDS) tasks remains largely unexplored. Specifically, it is unclear how the challenges arising from handling multiple documents (e.g., repetition and diversity of information) affect models outputs. In this work, we investigate how hallucinations manifest in LLMs when summarizing topic-specific information from multiple documents. Since no benchmarks exist for investigating hallucinations in MDS, we use existing news and conversation datasets, annotated with topic-specific insights, to create two novel multi-document benchmarks. When evaluating 5 LLMs on our benchmarks, we observe that on average, up to 75% of the content in LLM-generated summary is hallucinated, with hallucinations more likely to occur towards the end of the summaries. Moreover, when summarizing non-existent topic-related information, gpt-3.5-turbo and GPT-4o still generate summaries about 79.35% and 44% of the time, raising concerns about their tendency to fabricate content. To understand the characteristics of these hallucinations, we manually evaluate 700+ insights and find that most errors stem from either failing to follow instructions or producing overly generic insights. Motivated by these observations, we investigate the efficacy of simple post-hoc baselines in mitigating hallucinations but find them only moderately effective. Our results underscore the need for more effective approaches to systematically mitigate hallucinations in MDS. We release our dataset and code at github.com/megagonlabs/Hallucination_MDS.

Nima Shahbazi, Seiji Maekawa, Nikita Bhutani et al.

While recent advances in large language models have significantly improved Text-to-SQL and table question answering systems, most existing approaches assume that all query-relevant information is explicitly represented in structured schemas. In practice, many enterprise databases contain hybrid schemas where structured attributes coexist with free-form textual fields, requiring systems to reason over both types of information. To address this challenge, we introduce OmniTQA, a cost-aware hybrid query processing framework that operates over both structured and semi-structured data. OmniTQA treats semantic reasoning as a first-class query operator, seamlessly integrating LLM-based semantic operations with classical relational operators into an executable directed acyclic graph. To manage the high latency and cost of LLM inference, it extends classical query optimization with data-aware planning, combining atomic query decomposition and operator reordering to minimize semantic workload. The framework also features a dual-engine execution architecture that dynamically routes tasks between a relational database and an LLM module, using operator-aware batching to scale efficiently. Extensive experiments across a diverse suite of structured and semi-structured table question answering benchmarks demonstrate that OmniTQA consistently outperforms existing symbolic, semantic, and hybrid baselines in both accuracy and cost efficiency. These gains are particularly pronounced for complex queries, large tables and multi-relation schemas.

Nedelina Teneva, Estevam Hruschka

Despite the recent popularity of knowledge graph (KG) related tasks and benchmarks such as KG embeddings, link prediction, entity alignment and evaluation of the reasoning abilities of pretrained language models as KGs, the structure and properties of real KGs are not well studied. In this paper, we perform a large scale comparative study of 29 real KG datasets from diverse domains such as the natural sciences, medicine, and NLP to analyze their properties and structural patterns. Based on our findings, we make several recommendations regarding KG-based model development and evaluation. We believe that the rich structural information contained in KGs can benefit the development of better KG models across fields and we hope this study will contribute to breaking the existing data silos between different areas of research (e.g., ML, NLP, AI for sciences).

Jackson Hassell, Dan Zhang, Hannah Kim et al.

We investigate how agents built on pretrained large language models can learn target classification functions from labeled examples without parameter updates. While conventional approaches like fine-tuning are often costly, inflexible, and opaque, we propose a memory-augmented framework that leverages both labeled data and LLM-generated critiques. Our framework uses episodic memory to store instance-level critiques-capturing specific past experiences-and semantic memory to distill these into reusable, task-level guidance. Across a diverse set of tasks, incorporating critiques yields up to a 24.8 percent accuracy improvement over retrieval-based (RAG-style) baselines that rely only on labels. Through extensive empirical evaluation, we uncover distinct behavioral differences between OpenAI and opensource models, particularly in how they handle fact-oriented versus preference-based data. To interpret how models respond to different representations of supervision encoded in memory, we introduce a novel metric, suggestibility. This helps explain observed behaviors and illuminates how model characteristics and memory strategies jointly shape learning dynamics. Our findings highlight the promise of memory-driven, reflective learning for building more adaptive and interpretable LLM agents.

Naoki Otani, Nikita Bhutani, Hannah Kim et al.

Explicit planning is a critical capability for LLM-based agents solving complex data-centric tasks, which require precise tool calling over external data sources. Existing strategies fall into two paradigms based on planning horizon: (1) full-horizon (FH), which generates a complete plan before execution, and (2) single-step horizon (SH), which interleaves each action (tool call) with incremental reasoning and observation. While step-by-step execution is a common default under the assumption that eager execution monitoring is necessary for adaptability, we revisit this assumption for well-defined data-centric tasks. Our controlled empirical study isolates planning horizon as the key architectural feature and systematically analyzes the effects of topological complexity and tool robustness on both paradigms. Our experiments across Knowledge Base Question Answering and Multi-hop QA show that FH planning with lazy replanning achieves accuracy parity with SH across varying depths, breadths, and robustness levels, while using 2-3x fewer tokens. These findings suggest that for well-defined data-centric tasks, eager step-wise monitoring is often unnecessary, and full-horizon planning with on-demand replanning can offer a more efficient default.

Pouya Pezeshkpour, Eser Kandogan, Nikita Bhutani et al.

Remarkable performance of large language models (LLMs) in a variety of tasks brings forth many opportunities as well as challenges of utilizing them in production settings. Towards practical adoption of LLMs, multi-agent systems hold great promise to augment, integrate, and orchestrate LLMs in the larger context of enterprise platforms that use existing proprietary data and models to tackle complex real-world tasks. Despite the tremendous success of these systems, current approaches rely on narrow, single-focus objectives for optimization and evaluation, often overlooking potential constraints in real-world scenarios, including restricted budgets, resources and time. Furthermore, interpreting, analyzing, and debugging these systems requires different components to be evaluated in relation to one another. This demand is currently not feasible with existing methodologies. In this postion paper, we introduce the concept of reasoning capacity as a unifying criterion to enable integration of constraints during optimization and establish connections among different components within the system, which also enable a more holistic and comprehensive approach to evaluation. We present a formal definition of reasoning capacity and illustrate its utility in identifying limitations within each component of the system. We then argue how these limitations can be addressed with a self-reflective process wherein human-feedback is used to alleviate shortcomings in reasoning and enhance overall consistency of the system.

Pouya Pezeshkpour, Estevam Hruschka

Verification is becoming central to both reinforcement-learning-based training and inference-time control of large language models (LLMs). Yet current verifiers face a fundamental trade-off: LLM-based verifiers are expressive but hard to control and prone to error, while deterministic executable verifiers are reliable and interpretable but often limited in capability. We study the following question: given a development set of LLM outputs and labels for a target objective, such as correctness, can we automatically induce a minimal set of Python verifiers whose joint satisfaction closely matches that objective? We propose AutoPyVerifier, a framework that uses an LLM to synthesize candidate verifier functions and then refines them through search over a directed acyclic graph (DAG). By navigating the DAG, AutoPyVerifier systematically explores the space of deterministic executable verifiers and selects a compact verifier set whose joint satisfaction best approximates the target objective. Across mathematical reasoning, coding, function calling, and instruction-following benchmarks for several state-of-the-art LLMs, AutoPyVerifier improves target-objective prediction by up to 55.0 F1 points over the initial LLM-generated verifier sets. Additional analyses show that the most useful verification targets vary by benchmark and model, and that the DAG-based search shifts the learned verifier sets toward more structural and semantically grounded checks. We further show that exposing the discovered verifier set to an LLM as an external tool improves downstream accuracy by up to 17.0 points. We release our code

Eser Kandogan, Nikita Bhutani, Dan Zhang et al.

Large language models (LLMs) have gained significant interest in industry due to their impressive capabilities across a wide range of tasks. However, the widespread adoption of LLMs presents several challenges, such as integration into existing applications and infrastructure, utilization of company proprietary data, models, and APIs, and meeting cost, quality, responsiveness, and other requirements. To address these challenges, there is a notable shift from monolithic models to compound AI systems, with the premise of more powerful, versatile, and reliable applications. However, progress thus far has been piecemeal, with proposals for agentic workflows, programming models, and extended LLM capabilities, without a clear vision of an overall architecture. In this paper, we propose a 'blueprint architecture' for compound AI systems for orchestrating agents and data for enterprise applications. In our proposed architecture the key orchestration concept is 'streams' to coordinate the flow of data and instructions among agents. Existing proprietary models and APIs in the enterprise are mapped to 'agents', defined in an 'agent registry' that serves agent metadata and learned representations for search and planning. Agents can utilize proprietary data through a 'data registry' that similarly registers enterprise data of various modalities. Tying it all together, data and task 'planners' break down, map, and optimize tasks and queries for given quality of service (QoS) requirements such as cost, accuracy, and latency. We illustrate an implementation of the architecture for a use-case in the HR domain and discuss opportunities and challenges for 'agentic AI' in the enterprise.

Naoki Otani, Nikita Bhutani, Estevam Hruschka

Advances in Natural Language Processing (NLP) have the potential to transform HR processes, from recruitment to employee management. While recent breakthroughs in NLP have generated significant interest in its industrial applications, a comprehensive overview of how NLP can be applied across HR activities is still lacking. This paper discovers opportunities for researchers and practitioners to harness NLP's transformative potential in this domain. We analyze key fundamental tasks such as information extraction and text classification, and their roles in downstream applications like recommendation and language generation, while also discussing ethical concerns. Additionally, we identify gaps in current research and encourage future work to explore holistic approaches for achieving broader objectives in this field.

Hayate Iso, Pouya Pezeshkpour, Nikita Bhutani et al. · nvidia

Large Language Models (LLMs) offer the potential to automate hiring by matching job descriptions with candidate resumes, streamlining recruitment processes, and reducing operational costs. However, biases inherent in these models may lead to unfair hiring practices, reinforcing societal prejudices and undermining workplace diversity. This study examines the performance and fairness of LLMs in job-resume matching tasks within the English language and U.S. context. It evaluates how factors such as gender, race, and educational background influence model decisions, providing critical insights into the fairness and reliability of LLMs in HR applications. Our findings indicate that while recent models have reduced biases related to explicit attributes like gender and race, implicit biases concerning educational background remain significant. These results highlight the need for ongoing evaluation and the development of advanced bias mitigation strategies to ensure equitable hiring practices when using LLMs in industry settings.

Kushan Mitra, Dan Zhang, Sajjadur Rahman et al.

Large Language Models (LLMs) have shown impressive capability in language generation and understanding, but their tendency to hallucinate and produce factually incorrect information remains a key limitation. To verify LLM-generated contents and claims from other sources, traditional verification approaches often rely on holistic models that assign a single factuality label to complex claims, potentially obscuring nuanced errors. In this paper, we advocate for a shift towards fine-grained verification, where complex claims are broken down into smaller sub-claims for individual verification, allowing for more precise identification of inaccuracies, improved transparency, and reduced ambiguity in evidence retrieval. However, generating sub-claims poses challenges, such as maintaining context and ensuring semantic equivalence with respect to the original claim. We introduce FactLens, a benchmark for evaluating fine-grained fact verification, with metrics and automated evaluators of sub-claim quality. The benchmark data is manually curated to ensure high-quality ground truth. Our results show alignment between automated FactLens evaluators and human judgments, and we discuss the impact of sub-claim characteristics on the overall verification performance.

Seiji Maekawa, Jackson Hassell, Pouya Pezeshkpour et al.

As language models gain access to external tools via structured function calls, they become increasingly more capable of solving complex, multi-step tasks. However, existing benchmarks for tool-augmented language models (TaLMs) provide insufficient control over factors such as the number of functions accessible, task complexity, and input size, and remain vulnerable to data contamination. We present FuncBenchGen, a unified, contamination-free framework that evaluates TaLMs by generating synthetic multi-step tool-use tasks. The key idea is to cast tool use as traversal over a hidden function-dependency DAG where nodes are function calls and an edge between nodes represents one function consuming the output of another. Given a set of external function schemas, initial variable values, and a target variable, models must compose the correct call sequence to compute the target variable. FuncBenchGen allows users to precisely control task difficulty (e.g., graph size, dependency depth, and distractor functions) while avoiding data leakage. We apply our FuncBenchGen framework to evaluate seven LLMs on tool use tasks of varying difficulty. Reasoning-optimized models consistently outperform general-purpose models with GPT-5 significantly outperforming other models. Performance declines sharply as dependency depth increases. Furthermore, connected irrelevant functions prove especially difficult to handle. We find that strong models often make syntactically valid function calls but propagate incorrect or stale argument values across steps, revealing brittle state tracking by LLMs in multi-turn tool use. Motivated by this observation, we introduce a simple mitigation strategy that explicitly restates prior variable values to the agent at each step. Surprisingly, this lightweight change yields substantial gains across models. e.g., yielding a success rate improvement from 62.5% to 81.3% for GPT-5.

Kushan Mitra, Dan Zhang, Hannah Kim et al.

Understanding user intent is essential for effective planning in conversational assistants, particularly those powered by large language models (LLMs) coordinating multiple agents. However, real-world dialogues are often ambiguous, underspecified, or dynamic, making intent detection a persistent challenge. Traditional classification-based approaches struggle to generalize in open-ended settings, leading to brittle interpretations and poor downstream planning. We propose RECAP (REwriting Conversations for Agent Planning), a new benchmark designed to evaluate and advance intent rewriting, reframing user-agent dialogues into concise representations of user goals. RECAP captures diverse challenges such as ambiguity, intent drift, vagueness, and mixed-goal conversations. Alongside the dataset, we introduce an LLM-based evaluator that assesses planning utility given the rewritten intent. Using RECAP, we develop a prompt-based rewriting approach that outperforms baselines. We further demonstrate that fine-tuning two DPO-based rewriters yields additional utility gains. Our results highlight intent rewriting as a critical and tractable component for improving agent planning in open-domain dialogue systems.

Pouya Pezeshkpour, Moin Aminnaseri, Estevam Hruschka

Vision-language models (VLMs) have demonstrated impressive performance by effectively integrating visual and textual information to solve complex tasks. However, it is not clear how these models reason over the visual and textual data together, nor how the flow of information between modalities is structured. In this paper, we examine how VLMs reason by analyzing their biases when confronted with scenarios that present conflicting image and text cues, a common occurrence in real-world applications. To uncover the extent and nature of these biases, we build upon existing benchmarks to create five datasets containing mismatched image-text pairs, covering topics in mathematics, science, and visual descriptions. Our analysis shows that VLMs favor text in simpler queries but shift toward images as query complexity increases. This bias correlates with model scale, with the difference between the percentage of image- and text-preferred responses ranging from +56.8% (image favored) to -74.4% (text favored), depending on the task and model. In addition, we explore three mitigation strategies: simple prompt modifications, modifications that explicitly instruct models on how to handle conflicting information (akin to chain-of-thought prompting), and a task decomposition strategy that analyzes each modality separately before combining their results. Our findings indicate that the effectiveness of these strategies in identifying and mitigating bias varies significantly and is closely linked to the model's overall performance on the task and the specific modality in question.

Pouya Pezeshkpour, Estevam Hruschka

Retrieval Augmented Generation (RAG) frameworks have shown significant promise in leveraging external knowledge to enhance the performance of large language models (LLMs). However, conventional RAG methods often retrieve documents based solely on surface-level relevance, leading to many issues: they may overlook deeply buried information within individual documents, miss relevant insights spanning multiple sources, and are not well-suited for tasks beyond traditional question answering. In this paper, we propose Insight-RAG, a novel framework designed to address these issues. In the initial stage of Insight-RAG, instead of using traditional retrieval methods, we employ an LLM to analyze the input query and task, extracting the underlying informational requirements. In the subsequent stage, a specialized LLM -- trained on the document database -- is queried to mine content that directly addresses these identified insights. Finally, by integrating the original query with the retrieved insights, similar to conventional RAG approaches, we employ a final LLM to generate a contextually enriched and accurate response. Using two scientific paper datasets, we created evaluation benchmarks targeting each of the mentioned issues and assessed Insight-RAG against traditional RAG pipeline. Our results demonstrate that the Insight-RAG pipeline successfully addresses these challenges, outperforming existing methods by a significant margin in most cases. These findings suggest that integrating insight-driven retrieval within the RAG framework not only enhances performance but also broadens the applicability of RAG to tasks beyond conventional question answering.

Pouya Pezeshkpour, Estevam Hruschka

Large Language Models (LLMs) have demonstrated remarkable performance on various tasks, yet their ability to extract and internalize deeper insights from domain-specific datasets remains underexplored. In this study, we investigate how continual pre-training can enhance LLMs' capacity for insight learning across three distinct forms: declarative, statistical, and probabilistic insights. Focusing on two critical domains: medicine and finance, we employ LoRA to train LLMs on two existing datasets. To evaluate each insight type, we create benchmarks to measure how well continual pre-training helps models go beyond surface-level knowledge. We also assess the impact of document modification on capturing insights. The results show that, while continual pre-training on original documents has a marginal effect, modifying documents to retain only essential information significantly enhances the insight-learning capabilities of LLMs.

Pouya Pezeshkpour, Estevam Hruschka

Utilizing large language models (LLMs) to rank a set of items has become a common approach in recommendation and retrieval systems. Typically, these systems focus on ordering a substantial number of documents in a monotonic order based on a given query. However, real-world scenarios often present a different challenge: ranking a comparatively smaller set of items, but according to a variety of diverse and occasionally conflicting conditions. In this paper, we define and explore the task of multi-conditional ranking by introducing MCRank, a benchmark tailored for assessing multi-conditional ranking across various item types and conditions. Our analysis of LLMs using MCRank indicates a significant decrease in performance as the number and complexity of items and conditions grow. To overcome this limitation, we propose a novel decomposed reasoning method, consisting of EXtracting and Sorting the conditions, and then Iteratively Ranking the items (EXSIR). Our extensive experiments show that this decomposed reasoning method enhances LLMs' performance significantly, achieving up to a 14.4% improvement over existing LLMs. We also provide a detailed analysis of LLMs performance across various condition categories, and examine the effectiveness of decomposition step. Furthermore, we compare our method with existing approaches such as Chain-of-Thought and existing ranking models, demonstrating the superiority of our approach and complexity of MCR task. We released our dataset and code.

Pouya Pezeshkpour, Estevam Hruschka

Large language models are increasingly deployed as specialized agents that plan, call tools, and take actions over extended horizons. Yet many existing evaluations assume a "clean interface" where dynamics are specified and stable, tools and sensors are reliable, and success is captured by a single explicit objective-often overestimating real-world readiness. In practice, agents face underspecified rules, unreliable signals, shifting environments, and implicit, multi-stakeholder goals. The challenge is therefore not just solving tasks, but adapting while solving: deciding what to trust, what is wanted, when to verify, and when to fall back or escalate. We stress-test deployment-relevant robustness under four operational circumstances: partial observability, dynamic environments, noisy signals, and dynamic agent state. We benchmark agentic LLMs in a grid-based game with a simple goal but long-horizon execution. Episodes violate clean-interface assumptions yet remain solvable, forcing agents to infer rules, pay for information, adapt to environmental and internal shifts, and act cautiously under noise. Across five state-of-the-art LLM agents, we find large gaps between nominal task-solving and deployment-like robustness. Performance generally degrades as grid size and horizon increase, but rankings are unstable: weaker models can beat stronger ones when strategy matches the uncertainty regime. Despite no explicit instruction, agents trade off completion, efficiency, and penalty avoidance, suggesting partial objective inference. Ablations and feature analyses reveal model-specific sensitivities and failure drivers, motivating work on verification, safe action selection, and objective inference under partial observability, noise, and non-stationarity.

Tianyang Xu, Dan Zhang, Kushan Mitra et al.

Large language model (LLM) agents are increasingly deployed to tackle complex tasks, often necessitating collaboration among multiple specialized agents. However, multi-agent collaboration introduces new challenges in planning, coordination, and verification. Execution failures frequently arise not from flawed reasoning alone, but from subtle misalignments in task interpretation, output format, or inter-agent handoffs. To address these challenges, we present VeriMAP, a framework for multi-agent collaboration with verification-aware planning. The VeriMAP planner decomposes tasks, models subtask dependencies, and encodes planner-defined passing criteria as subtask verification functions (VFs) in Python and natural language. We evaluate VeriMAP on diverse datasets, demonstrating that it outperforms both single- and multi-agent baselines while enhancing system robustness and interpretability. Our analysis highlights how verification-aware planning enables reliable coordination and iterative refinement in multi-agent systems, without relying on external labels or annotations.

Chen Shen, Sajjadur Rahman, Estevam Hruschka

Current approaches for question answering (QA) over tabular data, such as NL2SQL systems, perform well for factual questions where answers are directly retrieved from tables. However, they fall short on probabilistic questions requiring reasoning under uncertainty. In this paper, we introduce a new benchmark LUCARIO and a framework for probabilistic QA over large tabular data. Our method induces Bayesian Networks from tables, translates natural language queries into probabilistic queries, and uses large language models (LLMs) to generate final answers. Empirical results demonstrate significant improvements over baselines, highlighting the benefits of hybrid symbolic-neural reasoning.

Eser Kandogan, Sajjadur Rahman, Nikita Bhutani et al.

Large Language Models (LLMs) have showcased remarkable capabilities surpassing conventional NLP challenges, creating opportunities for use in production use cases. Towards this goal, there is a notable shift to building compound AI systems, wherein LLMs are integrated into an expansive software infrastructure with many components like models, retrievers, databases and tools. In this paper, we introduce a blueprint architecture for compound AI systems to operate in enterprise settings cost-effectively and feasibly. Our proposed architecture aims for seamless integration with existing compute and data infrastructure, with ``stream'' serving as the key orchestration concept to coordinate data and instructions among agents and other components. Task and data planners, respectively, break down, map, and optimize tasks and data to available agents and data sources defined in respective registries, given production constraints such as accuracy and latency.

Dongjin Choi, Sara Evensen, Çağatay Demiralp et al.

Despite rapid developments in the field of machine learning research, collecting high-quality labels for supervised learning remains a bottleneck for many applications. This difficulty is exacerbated by the fact that state-of-the-art models for NLP tasks are becoming deeper and more complex, often increasing the amount of training data required even for fine-tuning. Weak supervision methods, including data programming, address this problem and reduce the cost of label collection by using noisy label sources for supervision. However, until recently, data programming was only accessible to users who knew how to program. To bridge this gap, the Data Programming by Demonstration framework was proposed to facilitate the automatic creation of labeling functions based on a few examples labeled by a domain expert. This framework has proven successful for generating high-accuracy labeling models for document classification. In this work, we extend the DPBD framework to span-level annotation tasks, arguably one of the most time-consuming NLP labeling tasks. We built a novel tool, TagRuler, that makes it easy for annotators to build span-level labeling functions without programming and encourages them to explore trade-offs between different labeling models and active learning strategies. We empirically demonstrated that an annotator could achieve a higher F1 score using the proposed tool compared to manual labeling for different span-level annotation tasks.