Haibo Ding

Zhengbao Jiang, Luyu Gao, Jun Araki et al. · cmu

Systems for knowledge-intensive tasks such as open-domain question answering (QA) usually consist of two stages: efficient retrieval of relevant documents from a large corpus and detailed reading of the selected documents to generate answers. Retrievers and readers are usually modeled separately, which necessitates a cumbersome implementation and is hard to train and adapt in an end-to-end fashion. In this paper, we revisit this design and eschew the separate architecture and training in favor of a single Transformer that performs Retrieval as Attention (ReAtt), and end-to-end training solely based on supervision from the end QA task. We demonstrate for the first time that a single model trained end-to-end can achieve both competitive retrieval and QA performance, matching or slightly outperforming state-of-the-art separately trained retrievers and readers. Moreover, end-to-end adaptation significantly boosts its performance on out-of-domain datasets in both supervised and unsupervised settings, making our model a simple and adaptable solution for knowledge-intensive tasks. Code and models are available at https://github.com/jzbjyb/ReAtt.

Zhengbao Jiang, Jun Araki, Haibo Ding et al.

Generative question answering (QA) models generate answers to questions either solely based on the parameters of the model (the closed-book setting) or additionally retrieving relevant evidence (the open-book setting). Generative QA models can answer some relatively complex questions, but the mechanism through which they do so is still poorly understood. We perform several studies aimed at better understanding the multi-hop reasoning capabilities of generative QA models. First, we decompose multi-hop questions into multiple corresponding single-hop questions, and find marked inconsistency in QA models' answers on these pairs of ostensibly identical question chains. Second, we find that models lack zero-shot multi-hop reasoning ability: when trained only on single-hop questions, models generalize poorly to multi-hop questions. Finally, we demonstrate that it is possible to improve models' zero-shot multi-hop reasoning capacity through two methods that approximate real multi-hop natural language (NL) questions by training on either concatenation of single-hop questions or logical forms (SPARQL). In sum, these results demonstrate that multi-hop reasoning does not emerge naturally in generative QA models, but can be encouraged by advances in training or modeling techniques.

Kang Zhou, Sangmin Woo, Haibo Ding et al.

Agent evaluation requires assessing complex multi-step behaviors involving tool use and intermediate reasoning, making it costly and expertise-intensive. A natural question arises: can frontier coding assistants reliably automate this evaluation process? Our study shows that simply prompting coding assistants is insufficient for this task. Without domain-specific evaluation knowledge, frontier coding assistants achieve only a 30% execution success rate and produce over-engineered evaluations averaging 12+ metrics per agent, indicating that strong coding ability does not automatically translate to reliable agent evaluation. We introduce EvalAgent, an AI assistant that automates the end-to-end agent evaluation pipeline. EvalAgent encodes evaluation domain expertise as evaluation skills (procedural instructions, reusable code and templates, and dynamically retrieved API documentation) that compose into a trace-based pipeline producing complete evaluation artifacts including metrics, executable code, and reports. To systematically assess generated evaluations, we introduce a meta-evaluation framework alongside AgentEvalBench, a benchmark comprising 20 agents, each paired with evaluation requirements and test scenarios. We further propose the Eval@1 metric to measure whether generated evaluation code both executes and yields meaningful results on the first run. Our experiments show that EvalAgent produces focused evaluations, improving Eval@1 from 17.5% to 65%, and achieving 79.5% human expert preference over baseline approaches. Further ablation studies show that evaluation skills are critical for handling complex evaluation: removing them causes Eval@1 to drop significantly from 65% to 30%.

Sangmin Woo, Kang Zhou, Yun Zhou et al.

Large Vision Language Models (LVLMs) often suffer from object hallucination, which undermines their reliability. Surprisingly, we find that simple object-based visual prompting -- overlaying visual cues (e.g., bounding box, circle) on images -- can significantly mitigate such hallucination; however, different visual prompts (VPs) vary in effectiveness. To address this, we propose Black-Box Visual Prompt Engineering (BBVPE), a framework to identify optimal VPs that enhance LVLM responses without needing access to model internals. Our approach employs a pool of candidate VPs and trains a router model to dynamically select the most effective VP for a given input image. This black-box approach is model-agnostic, making it applicable to both open-source and proprietary LVLMs. Evaluations on benchmarks such as POPE and CHAIR demonstrate that BBVPE effectively reduces object hallucination.

Zhengbao Jiang, Jun Araki, Haibo Ding et al.

Recent works have shown that language models (LM) capture different types of knowledge regarding facts or common sense. However, because no model is perfect, they still fail to provide appropriate answers in many cases. In this paper, we ask the question "how can we know when language models know, with confidence, the answer to a particular query?" We examine this question from the point of view of calibration, the property of a probabilistic model's predicted probabilities actually being well correlated with the probabilities of correctness. We examine three strong generative models -- T5, BART, and GPT-2 -- and study whether their probabilities on QA tasks are well calibrated, finding the answer is a relatively emphatic no. We then examine methods to calibrate such models to make their confidence scores correlate better with the likelihood of correctness through fine-tuning, post-hoc probability modification, or adjustment of the predicted outputs or inputs. Experiments on a diverse range of datasets demonstrate the effectiveness of our methods. We also perform analysis to study the strengths and limitations of these methods, shedding light on further improvements that may be made in methods for calibrating LMs. We have released the code at https://github.com/jzbjyb/lm-calibration.

Kiran Ramnath, Kang Zhou, Sheng Guan et al.

Since the advent of large language models (LLMs), prompt engineering has been a crucial step for eliciting desired responses for various Natural Language Processing (NLP) tasks. However, prompt engineering remains an impediment for end users due to rapid advances in models, tasks, and associated best practices. To mitigate this, Automatic Prompt Optimization (APO) techniques have recently emerged that use various automated techniques to help improve the performance of LLMs on various tasks. In this paper, we present a comprehensive survey summarizing the current progress and remaining challenges in this field. We provide a formal definition of APO, a 5-part unifying framework, and then proceed to rigorously categorize all relevant works based on their salient features therein. We hope to spur further research guided by our framework.

Yiyue Qian, Shinan Zhang, Yun Zhou et al.

Large Language Models (LLMs) have revolutionized AI-generated content evaluation, with the LLM-as-a-Judge paradigm becoming increasingly popular. However, current single-LLM evaluation approaches face significant challenges, including inconsistent judgments and inherent biases from pre-training data. To address these limitations, we propose CollabEval, a novel multi-agent evaluation framework that implements a three-phase Collaborative Evaluation process: initial evaluation, multi-round discussion, and final judgment. Unlike existing approaches that rely on competitive debate or single-model evaluation, CollabEval emphasizes collaboration among multiple agents with strategic consensus checking for efficiency. Our extensive experiments demonstrate that CollabEval consistently outperforms single-LLM approaches across multiple dimensions while maintaining robust performance even when individual models struggle. The framework provides comprehensive support for various evaluation criteria while ensuring efficiency through its collaborative design.

Darren Yow-Bang Wang, Zhengyuan Shen, Soumya Smruti Mishra et al.

Structured outputs are essential for large language models (LLMs) in critical applications like agents and information extraction. Despite their capabilities, LLMs often generate outputs that deviate from predefined schemas, significantly hampering reliable application development. We present SLOT (Structured LLM Output Transformer), a model-agnostic approach that transforms unstructured LLM outputs into precise structured formats. While existing solutions predominantly rely on constrained decoding techniques or are tightly coupled with specific models, SLOT employs a fine-tuned lightweight language model as a post-processing layer, achieving flexibility across various LLMs and schema specifications. We introduce a systematic pipeline for data curation and synthesis alongside a formal evaluation methodology that quantifies both schema accuracy and content fidelity. Our results demonstrate that fine-tuned Mistral-7B model with constrained decoding achieves near perfect schema accuracy (99.5%) and content similarity (94.0%), outperforming Claude-3.5-Sonnet by substantial margins (+25 and +20 percentage points, respectively). Notably, even compact models like Llama-3.2-1B can match or exceed the structured output capabilities of much larger proprietary models when equipped with SLOT, enabling reliable structured generation in resource-constrained environments.

Zhichao Xu, Aosong Feng, Yijun Tian et al.

In recent years, dense retrieval has been the focus of information retrieval (IR) research. While effective, dense retrieval produces uninterpretable dense vectors, and suffers from the drawback of large index size. Learned sparse retrieval (LSR) has emerged as promising alternative, achieving competitive retrieval performance while also being able to leverage the classical inverted index data structure for efficient retrieval. However, limited works have explored scaling LSR beyond BERT scale. In this work, we identify two challenges in training large language models (LLM) for LSR: (1) training instability during the early stage of contrastive training; (2) suboptimal performance due to pre-trained LLM's unidirectional attention. To address these challenges, we propose two corresponding techniques: (1) a lightweight adaptation training phase to eliminate training instability; (2) two model variants to enable bidirectional information. With these techniques, we are able to train LSR models with 8B scale LLM, and achieve competitive retrieval performance with reduced index size. Furthermore, we are among the first to analyze the performance-efficiency tradeoff of LLM-based LSR model through the lens of model quantization. Our findings provide insights into adapting LLMs for efficient retrieval modeling.

Sullam Jeoung, Yueyan Chen, Yi Zhang et al.

Prompts are the interface for eliciting the capabilities of large language models (LLMs). Understanding their structure and components is critical for analyzing LLM behavior and optimizing performance. However, the field lacks a comprehensive framework for systematic prompt analysis and understanding. We introduce PromptPrism, a linguistically-inspired taxonomy that enables prompt analysis across three hierarchical levels: functional structure, semantic component, and syntactic pattern. We show the practical utility of PromptPrism by applying it to three applications: (1) a taxonomy-guided prompt refinement approach that automatically improves prompt quality and enhances model performance across a range of tasks; (2) a multi-dimensional dataset profiling method that extracts and aggregates structural, semantic, and syntactic characteristics from prompt datasets, enabling comprehensive analysis of prompt distributions and patterns; (3) a controlled experimental framework for prompt sensitivity analysis by quantifying the impact of semantic reordering and delimiter modifications on LLM performance. Our experimental results validate the effectiveness of our taxonomy across these applications, demonstrating that PromptPrism provides a foundation for refining, profiling, and analyzing prompts.

Zheng Huang, Kiran Ramnath, Yueyan Chen et al.

Diffusion language models (DLMs) have recently emerged as a compelling alternative to autoregressive generation, offering parallel generation and improved global coherence. During inference, DLMs generate text by iteratively denoising masked sequences in parallel; however, determining which positions to unmask and which tokens to commit forms a large combinatorial search problem. Existing inference methods approximate this search using heuristics, which often yield suboptimal decoding paths; other approaches instead rely on additional training to guide token selection. To introduce a principled search mechanism for DLMs inference, we introduce MEDAL, an inference-time scaling framework that integrates Monte Carlo Tree SEarch initialization for Diffusion LAnguage Model inference. We employ Monte Carlo Tree Search at the initialization stage to explore promising unmasking trajectories, providing a robust starting point for subsequent refinement. This design enables efficient inference-time scaling, allowing generation quality to improve as the search budget increases, without additional training. Across multiple benchmarks, MEDAL achieves up to 22.0% improvement over existing inference strategies, establishing a new paradigm for search-based inference in DLMs.

Zhichao Xu, Zongyu Wu, Yun Zhou et al.

Inspired by the success of reinforcement learning (RL) in Large Language Model (LLM) training for domains like math and code, recent works have begun exploring how to train LLMs to use search engines more effectively as tools for retrieval-augmented generation. Although these methods achieve performance improvement across QA benchmarks, many prioritize final answer correctness while overlooking the quality of intermediate reasoning steps, which may lead to chain-of-thought unfaithfulness. In this paper, we first introduce a comprehensive evaluation framework for evaluating RL-based search agents, covering three distinct faithfulness metrics: information-think faithfulness, think-answer faithfulness, and think-search faithfulness. Our evaluations reveal that a prototypical RL-based search agent, Search-R1, has significant room for improvement in this regard. To foster faithful reasoning, we introduce VERITAS (Verifying Entailed Reasoning through Intermediate Traceability in Agentic Search), a novel framework that integrates fine-grained faithfulness rewards into the reinforcement learning process. Our experiments show that models trained with VERITAS not only significantly improve reasoning faithfulness, but also achieve comparable task performance across seven QA benchmarks.

Aosong Feng, Balasubramaniam Srinivasan, Yun Zhou et al.

Routing incoming queries to the most cost-effective LLM while maintaining response quality poses a fundamental challenge in optimizing performance-cost trade-offs for large-scale commercial systems. We present IPR\, -- \,a quality-constrained \textbf{I}ntelligent \textbf{P}rompt \textbf{R}outing framework that dynamically selects optimal models based on predicted response quality and user-specified tolerance levels. IPR introduces three key innovations: (1) a modular architecture with lightweight quality estimators trained on 1.5M prompts annotated with calibrated quality scores, enabling fine-grained quality prediction across model families; (2) a user-controlled routing mechanism with tolerance parameter $τ\in [0,1]$ that provides explicit control over quality-cost trade-offs; and (3) an extensible design using frozen encoders with model-specific adapters, reducing new model integration from days to hours. To rigorously train and evaluate IPR, we curate an industrial-level dataset IPRBench\footnote{IPRBench will be released upon legal approval.}, a comprehensive benchmark containing 1.5 million examples with response quality annotations across 11 LLM candidates. Deployed on a major cloud platform, IPR achieves 43.9\% cost reduction while maintaining quality parity with the strongest model in the Claude family and processes requests with sub-150ms latency. The deployed system and additional product details are publicly available at https://aws.amazon.com/bedrock/intelligent-prompt-routing/

Jiacheng Li, Haibo Ding, Jingbo Shang et al.

We study the problem of building entity tagging systems by using a few rules as weak supervision. Previous methods mostly focus on disambiguation entity types based on contexts and expert-provided rules, while assuming entity spans are given. In this work, we propose a novel method TALLOR that bootstraps high-quality logical rules to train a neural tagger in a fully automated manner. Specifically, we introduce compound rules that are composed from simple rules to increase the precision of boundary detection and generate more diverse pseudo labels. We further design a dynamic label selection strategy to ensure pseudo label quality and therefore avoid overfitting the neural tagger. Experiments on three datasets demonstrate that our method outperforms other weakly supervised methods and even rivals a state-of-the-art distantly supervised tagger with a lexicon of over 2,000 terms when starting from only 20 simple rules. Our method can serve as a tool for rapidly building taggers in emerging domains and tasks. Case studies show that learned rules can potentially explain the predicted entities.

Xinyan Zhao, Haibo Ding, Zhe Feng

Instead of using expensive manual annotations, researchers have proposed to train named entity recognition (NER) systems using heuristic labeling rules. However, devising labeling rules is challenging because it often requires a considerable amount of manual effort and domain expertise. To alleviate this problem, we propose \textsc{GLaRA}, a graph-based labeling rule augmentation framework, to learn new labeling rules from unlabeled data. We first create a graph with nodes representing candidate rules extracted from unlabeled data. Then, we design a new graph neural network to augment labeling rules by exploring the semantic relations between rules. We finally apply the augmented rules on unlabeled data to generate weak labels and train a NER model using the weakly labeled data. We evaluate our method on three NER datasets and find that we can achieve an average improvement of +20\% F1 score over the best baseline when given a small set of seed rules.

Zhengbao Jiang, Antonios Anastasopoulos, Jun Araki et al.

Language models (LMs) have proven surprisingly successful at capturing factual knowledge by completing cloze-style fill-in-the-blank questions such as "Punta Cana is located in _." However, while knowledge is both written and queried in many languages, studies on LMs' factual representation ability have almost invariably been performed on English. To assess factual knowledge retrieval in LMs in different languages, we create a multilingual benchmark of cloze-style probes for 23 typologically diverse languages. To properly handle language variations, we expand probing methods from single- to multi-word entities, and develop several decoding algorithms to generate multi-token predictions. Extensive experimental results provide insights about how well (or poorly) current state-of-the-art LMs perform at this task in languages with more or fewer available resources. We further propose a code-switching-based method to improve the ability of multilingual LMs to access knowledge, and verify its effectiveness on several benchmark languages. Benchmark data and code have been released at https://x-factr.github.io.