论文检索

Mohsen Nayebi Kerdabadi, Arya Hadizadeh Moghaddam, Chen Chen, Dongjie Wang, Zijun Yao

In electronic health record (EHR) mining, learning high-quality representations of medical concepts (e.g., standardized diagnosis, medication, and procedure codes) is fundamental for downstream clinical prediction. However, robust concept representation learning is hindered by two key challenges: (i) clinically important cross-type dependencies (e.g., diagnosis-medication and medication-procedure relations) are often missing or incomplete in existing ontology resources, limiting the ability to model complex EHR patterns; and (ii) rich clinical semantics are often missing from structured resources, and even when available as text, are difficult to integrate with KG structure for representation learning. To address these challenges, we present MedCo, an LLM-empowered graph learning framework for medical concept representation. MedCo first builds a global knowledge graph (KG) over medical codes by combining statistically reliable associations mined from EHRs with type-constrained LLM prompting to infer semantic relations. It then utilizes LLMs to enrich the KG into a text-attributed graph by generating node descriptions and edge rationales, providing semantic signals for both concepts and their relationships. Finally, MedCo jointly trains a LoRA-tuned LLaMA text encoder with a heterogeneous GNN, fusing text semantics and graph structure into unified concept embeddings. Extensive experiments on MIMIC-III and MIMIC-IV show that MedCo consistently improves prediction performance and serves as an effective plug-in concept encoder for standard EHR pipelines.

Artur Kulmizev, Erika Lombart, Patrick Watrin, Marie-Catherine de Marneffe

Large language models (LLMs) have shown considerable promise for annotation purposes, yet questions remain about their ability to capture human label variation (HLV) — genuine disagreement between annotators often observed across NLP tasks. Here, we investigate how label and explanation variation manifests within and across LLMs with respect to the Natural Language Inference (NLI) task. Using zero-shot prompting with exact human annotation instructions, we treat individual model generations as participants and examine three response sampling strategies: varying generation parameters, leveraging within-family model size differences, and pooling responses from distinct LLMs. We show that, while model ensembles can generate label distributions similar to humans, they likewise exhibit distinct, idiosyncratic judgments and disagreement patterns. We further analyze explanation variation, observing that, although models generate longer explanations than humans, they demonstrate substantially less stylistic diversity. Our findings suggest that, while LLMs may serve as useful tools for generating diverse annotations, they should not be viewed as drop-in replacements for human annotators — particularly in applications requiring authentic representation of diversity in human judgments, such as NLI.

Ruiyao Xu, Mihir Parmar, Tiankai Yang, Zhengyu Hu, Yue Zhao, Kaize Ding

Learning from preference-based feedback has become an effective approach for aligning LLMs across diverse tasks. However, high-quality human-annotated preference data remains expensive and scarce. Existing methods address this challenge through either self-rewarding, which scales by using purely AI-generated labels but risks unreliability, or active learning, which ensures quality through oracle annotation but cannot fully leverage unlabeled data. In this paper, we present CoAct, a novel framework that synergistically combines self-rewarding and active learning through strategic human-AI collaboration. CoAct leverages self-consistency to identify both reliable self-labeled data and samples requiring oracle verification. Additionally, oracle feedback guides the model to generate new instructions within its solvable capability. Evaluated on three reasoning benchmarks across two model families, CoAct achieves average improvements of +13.25% on GSM8K, +8.19% on MATH, and +13.16% on WebInstruct, consistently outperforming all baselines.

Harshavardhana T Gowda, Daniel C Comstock, Lee M. Miller

We present a neuromuscular speech interface that translates electromyographic (EMG) signals recorded from orofacial muscles during speech articulation directly into audio. We find that self-supervised speech (S3) representations are strongly linearly related to the electrical power of muscle activity: a simple linear mapping predicts EMG power from S3 representations with a correlation of *r* = 0.85. In addition, EMG power vectors associated with distinct articulatory gestures form structured, separable clusters. Together, these observations suggest that S3 models implicitly encode articulatory mechanisms, as reflected in EMG activity. Leveraging this structure, we map EMG signals into the S3 representation space and synthesize speech, enabling end-to-end EMG-to-speech generation without explicit articulatory modeling or vocoder training. We demonstrate this system with a participant with amyotrophic lateral sclerosis (ALS), converting orofacial EMG recorded while she *silently* articulated speech into audio.

Pratyay Banerjee, Masud Moshtaghi, Shivashankar Subramanian, Amita Misra, Ankit Chadha

Large language models still struggle with reliable long-term conversational memory: simply enlarging context windows or applying naïve retrieval often introduces noise and destabilizes responses. We present APEX-MEM, a conversational memory system that combines three key innovations: (1) a property graph which use domain-agnostic ontology to structure conversations as temporally grounded events in an entity-centric framework, (2) append-only storage that preserves the full temporal evolution of information, and (3) a multi-tool retrieval agent that understands and resolves conflicting or evolving information at query time, producing a compact and contextually relevant memory summary. This retrieval-time resolution preserves the full interaction history while suppressing irrelevant details. APEX-MEM achieves 88.88% accuracy on LOCOMO and 86.2% on LongMemEval, outperforming state-of-the-art session-aware approaches and demonstrating that structured property graphs enable more temporally coherent long-term conversational reasoning.

Minqian Liu, Ioana Baldini, David Rabinowitz, David S Rosenberg, Sebastian Gehrmann, Mark Dredze

AI guardrail systems support usage policies by determining whether a user query or a generated response is allowed or forbidden under the policy. Fine-tuned guardrails – such as LlamaGuard and ShieldGemma – include policy definitions in prompts during training that can be updated during inference to aid generalization. However, our analysis reveals that these models still overfit the training policies, which prevents adaptation to new domains. We propose Augmented Policy Training (APT), a training recipe that enhances guardrail adaptability to unseen policies by using a suite of policy perturbation strategies during training to reduce overfitting and increase generalization. Notably, a small 1B model trained in this manner achieves comparable or better performance than existing 8B guardrails on unseen policies. Our work reveals critical limitations of existing AI guardrails, offers a promising solution, and provides actionable insights for adapting systems to new domains and policies.

Nicolas Bougie, Gian Maria Marconi, Xiaotong Ye, Narimawa Watanabe

Evaluating recommender systems remains challenging due to the gap between offline metrics and real user behavior, as well as the scarcity of interaction data. Recent work explores large language model (LLM) agents as synthetic users, yet they typically rely on few-shot prompting, which yields a shallow understanding of the environment and limits their ability to faithfully reproduce user actions. We introduce AlignUSER, a framework that learns world-model-driven agents from human interactions. Given rollout sequences of actions and states, we formalize world modeling as a next state prediction task that helps the agent internalize the environment. To align actions with human personas, we generate counterfactual trajectories around demonstrations and prompt the LLM to compare its decisions with human choices, identify suboptimal actions, and extract lessons. The learned policy is then used to drive agent interactions with the recommender system. We evaluate AlignUSER across multiple datasets and demonstrate closer alignment with genuine humans than prior work, both at the micro and macro levels.

Calvin Isch, Grace Jennings

Large language models (LLMs) are increasingly used to summarize academic work, yet model summaries can subtly exaggerate or mischaracterize findings. We examine how Narrative License (NL), rhetorical shifts that amplify claims beyond the underlying evidence, emerges in LLM summaries of scholarly articles. Using diverse prompting strategies across six leading models, we assess three dimensions of NL: causal overreach, rhetorical confidence, and sentiment (N = 100 peer-reviewed articles). Under basic summarization prompts, models frequently increase NL relative to academic abstracts; however, guardrail prompts can reduce these distortions. We further test how model "sycophancy" shapes NL, finding that stated stances and user personas produce predictable shifts in each element. These findings suggest that users and the benchmarks used to evaluate summarization should explicitly consider subtle rhetorical distortions and user alignment to ensure faithful scientific communication.

Nicholas Edwards, Yukyung Lee, Yujun Audrey Mao, Yulu Qin, Sebastian Schuster, Najoung Kim

Agents based on Large Language Models (LLMs) have shown promise for performing sophisticated software engineering tasks autonomously. In addition, there has been progress towards developing agents that can perform parts of the research pipeline in machine learning and the natural sciences. We argue that research extension and its implementation is a critical capability for such systems, and introduce RExBench to support the evaluation of this capability. RExBench is a benchmark consisting of realistic extensions of 12 research papers that aim to investigate novel research hypotheses. Each task is set up as an extension to an existing research paper and codebase, accompanied by domain expert-written instructions. RExBench is robust to data contamination, and supports an automatic evaluation infrastructure that executes agent outputs to determine whether the success criteria are met. We use this benchmark to evaluate 12 LLM agents implemented using two different frameworks: aider and OpenHands. We find that all agents fail to autonomously implement the majority of the extensions, with the best agent at around 33% success rate. Although the success rate improves with additional human-written hints, the best performance under this setting remains below 44%. This indicates that current agents are still short of being able to handle realistic research extension tasks without substantial human guidance.

Tu Anh Dinh, Jan Niehues

Language model (LM) probability is not a reliable quality estimator, as natural language is ambiguous. When multiple output options are valid, the model’s probability distribution is spread across them, which can misleadingly indicate low output quality. This issue is caused by two reasons: (1) LMs’ final output activation is softmax, which does not allow multiple correct options to receive high probabilities simultaneuously and (2) LMs’ training data is single, one-hot encoded references, indicating that there is only one correct option at each output step. We propose training a module for Quality Estimation on top of pre-trained LMs to address these limitations. The module, called Sigmoid Head, is an extra unembedding head with sigmoid activation to tackle the first limitation. To tackle the second limitation, during the negative sampling process to train the Sigmoid Head, we use a heuristic to avoid selecting potentially alternative correct tokens. Our Sigmoid Head is computationally efficient during training and inference. The probability from Sigmoid Head is notably better quality signal compared to the original softmax head. As the Sigmoid Head does not rely on human-annotated quality data, it is more robust to out-of-domain settings compared to supervised QE.

Lingxiao Guan, Yuanhao Huang, Jie Liu

In Question Answering (QA), Retrieval Augmented Generation (RAG) has revolutionized performance in various domains. However, how to effectively capture multi-document relationships remains an open question. This is particularly critical for biomedical tasks due to their reliance on information spread across multiple documents. In this work, we propose a novel method CLAIMS, which utilizes propositional claims to construct a local knowledge graph from retrieved documents. Summaries are then derived via layerwise summarization from the knowledge graph to contextualize a small language model to perform QA. The structured summaries effectively capture explicit and implicit relationships between entities in the documents, thus having a more comprehensive context to provide to LLMs. CLAIMS achieved comparable or superior performance over RAG baselines on several biomedical QA benchmarks. We also evaluated its generalizability and each individual step of our approach with a targeted set of metrics, demonstrating its effectiveness.

Tianyang Xu, Marcelo Sandoval-Castañeda, Karen Livescu, Greg Shakhnarovich, Kanishka Misra

What is the interplay between semantic representations learned by language models (LM) from surface form alone to those learned from more grounded evidence? We study this question for a scenario where part of the input comes from a different modality—in our case, in a vision-language model (VLM), where a pretrained LM is aligned with a pretrained image encoder. As a case study we focus on the task of predicting hypernyms of objects represented in images. We do so in a VLM setup where the image encoder and LM are kept frozen, and only the intermediate mappings are learned. We progressively deprive the VLM of explicit evidence for hypernyms, and test whether the knowledge is recoverable from the LM. We find that the LMs we study can recover this knowledge and generalize even in the most extreme version of this experiment (when the model receives no evidence of a hypernym during training). Additional experiments suggest that this cross-modal taxonomic generalization persists under counterfactual image–label mappings only when the counterfactual data have high visual similarity within each category. Taken together, these findings suggest that cross-modal generalization in LMs arises from an interaction between linguistic structure and the information present in the input.

Haiqing Li, Wenliang Zhong, Yinhao Wu, Hehuan Ma, Yuzhi Guo, Thao M. Dang, Junzhou Huang

Guideline-following is increasingly important in compliance, customer support, and other regulated workflows, where correctness is defined by explicit rule systems rather than heuristics. Learning to follow guidelines is challenging because guidelines are interdependent: rules can trigger, suppress, or conflict with one another, while locally plausible responses may violate global constraints. Most existing methods treat guidelines as static text and rely on implicit reasoning or deeper decoding, making rule interactions and satisfaction status hard to observe and control. A more feasible approach is to model guideline execution with an explicit state that tracks evolving rule evidence across steps. However, conventional world models are a poor fit: they typically assume privileged feedback or well-defined transition dynamics, assumptions that do not hold when reasoning occurs purely in language space under ambiguous, text-defined constraints. As a solution, we propose RGCWM, a Rule-Grounded Causal World Model that builds an explicit state space from the guideline text itself. RGCWM represents rule applicability and satisfaction as a continuously updated evidence state, externalizes inter-rule dependencies as a causal structure, and plans at inference time by counterfactually evaluating candidate responses under model-estimated state transitions. Experiments show that this shift from implicit text reasoning to state-based reasoning enables stable, controllable execution of complex interacting rules across diverse domains.

Ilias Triantafyllopoulos, Renyi Qu, Salvatore Giorgi, Brenda Curtis, Lyle Ungar, João Sedoc

Retrieval-Augmented Generation (RAG) systems are increasingly deployed in high-stakes domains, where safety depends not only on how a system answers, but also on whether a query should be answered given a knowledge base (KB). Out-of-domain (OOD) queries can cause dense retrieval to surface weakly related context and lead the generator to produce fluent but unjustified responses. We study lightweight, KB-aligned OOD detection as an always-on gate for RAG systems. Our approach applies PCA to KB embeddings and scores queries in a compact subspace selected either by explained-variance retention (EVR) or by a separability-driven -test ranking. We evaluate geometric semantic-search rules and lightweight classifiers across 16 domains, including high-stakes COVID-19 and Substance Use KBs, and stress-test robustness using both LLM-generated attacks and an in-the-wild 4chan attack. We find that low-dimensional detectors achieve competitive OOD performance while being faster, cheaper, and more interpretable than prompted LLM-based judges. Finally, human and LLM-based evaluations show that OOD queries primarily degrade the relevance of RAG outputs, highlighting the need for efficient external OOD detection to maintain safe, in-scope behavior.

Amirhosein Ghasemabadi, Keith G. Mills, Baochun Li, Di Niu

Test-Time Scaling (TTS) methods for enhancing Large Language Model (LLM) reasoning often incur substantial inference costs, due to reliance on long chain-of-thought (CoT) generation, self-consistency sampling methods, or searching under Process Reward Models (PRMs). This paper introduces Guided by Gut (GG), an efficient self-guided TTS framework that enables LLMs to perform step-by-step reasoning at a low cost, without any reward models or verifiers. GG performs a lightweight tree search guided solely by intrinsic confidence signals of the LLM at each reasoning step and improves the reliability of such internal confidence signals by reinforcement learning. Empirical evaluations on challenging mathematical reasoning benchmarks demonstrate that GG enables smaller models (e.g., 1.5B-7B parameters) to achieve accuracy matching or surpassing significantly larger models (e.g., 32B–70B parameters), while reducing GPU memory usage by up to 10×. Compared to TTS with PRMs, GG achieves comparable accuracy with 8× faster inference speeds and 4–5× lower memory usage. Additionally, GG reduces KV cache memory usage by approximately 50% compared to Best-of-N sampling, facilitating more efficient and practical deployment of TTS techniques.

Changdae Oh, Seongheon Park, To Eun Kim, Jiatong Li, Wendi Li, Samuel Yeh, Sean Du, Hamed Hassani, Paul Bogdan, Dawn Song 等

Uncertainty quantification (UQ) for large language models (LLMs) is a key building block for safety guardrails of daily LLM applications. Yet, even as LLM agents are increasingly deployed in highly complex tasks, most UQ research still centers on single-turn question-answering. We argue that UQ research must shift to realistic settings with interactive agents, and that a new principled framework for agent UQ is needed. This paper presents three pillars to build a solid ground for future agent UQ research: (1. Foundations) We present the first general formulation of agent UQ that subsumes broad classes of existing UQ setups; (2. Challenges) We identify four technical challenges specifically tied to agentic setups—selection of uncertainty estimator, uncertainty of heterogeneous entities, modeling uncertainty dynamics in interactive systems, and lack of fine-grained benchmarks—with numerical analysis on a real-world agent benchmark, \tau^2-bench; (3. Future Directions) We conclude with noting on the practical implications of agent UQ and remaining open problems as forward-looking discussion for future explorations.

Tiejin Chen, Huaiyuan Yao, Jia Chen, Evangelos E. Papalexakis, Hua Wei

While Large Language Model-based Multi-Agent Systems (MAS) consistently outperform single-agent systems on complex tasks, their intricate interactions introduce critical reliability challenges arising from communication dynamics and role dependencies. Existing Uncertainty Quantification methods, typically designed for single-turn outputs, fail to address the unique complexities of the MAS. Specifically, these methods struggle with three distinct challenges: the cascading uncertainty in multi-step reasoning, the variability of inter-agent communication paths, and the diversity of communication topologies. To bridge this gap, we introduce MATU, a novel framework that quantifies uncertainty through tensor decomposition. MATU moves beyond analyzing final text outputs by representing entire reasoning trajectories as embedding matrices and organizing multiple execution runs into a higher-order tensor. By applying tensor decomposition, we disentangle and quantify distinct sources of uncertainty, offering a comprehensive reliability measure that is generalizable across different agent structures. We provide comprehensive experiments to show that MATU effectively estimates holistic and robust uncertainty across diverse tasks and communication topologies.

Myra Cheng, Robert D. Hawkins, Dan Jurafsky

Recent evaluations show that large language models (LLMs) frequently fail to challenge users’ harmful beliefs in domains ranging from medical advice to social reasoning. We present a unifying analysis through the lens of pragmatics: these safety failures can be understood and addressed as LLMs exhibiting excessive accommodation and insufficient epistemic vigilance. We show that the pragmatic factors affecting accommodation and epistemic vigilance in humans (at-issueness, linguistic encoding, and source reliability) influence LLM behaviors in similar ways. We demonstrate how these factors explain performance differences across three safety benchmarks that test models’ ability to challenge harmful beliefs, spanning misinformation (Cancer-Myth, SAGE-Eval) and sycophancy (ELEPHANT). This pragmatic lens further motivates prompting interventions, such as adding the phrase "wait a minute", that drastically improve performance on these difficult benchmarks by shifting pragmatic cues. Our results have practical implications for benchmark design and underscore the importance of pragmatics for understanding model behavior and improving performance.

Eshgin Hasanov, Md Mahadi Hassan Sibat, Santu Karmaker, Aashish Yadavally

Large language models (LLMs) have shown remarkable capabilities across diverse coding tasks. However, their adoption requires a true understanding of program execution rather than relying on surface-level patterns. Existing benchmarks primarily focus on predicting program properties tied to specific inputs (e.g., code coverage, program outputs). As a result, they provide a narrow view of dynamic code reasoning and are prone to data contamination. We argue that understanding program execution requires evaluating its inherent duality through two complementary reasoning tasks: (i) predicting a program’s observed behavior for a given input, and (ii) inferring how the input must be mutated toward a specific behavioral objective. Both tasks jointly probe a model’s causal understanding of execution flow. We instantiate this duality in DexBench, a benchmark comprising 445 paired instances, and evaluate 13 LLMs. Our results demonstrate that dual-path reasoning provides a robust and discriminative proxy for dynamic code understanding.

Zhan Qu, Michael Färber

Large Language Models (LLMs) are increasingly applied to medicine, yet their adoption is limited by concerns over reliability and safety. Existing evaluations either test factual medical knowledge in isolation or assess patient-level reasoning without verifying correctness, leaving a critical gap. We introduce MediEval, a benchmark that links MIMIC-IV electronic health records (EHRs) to a unified knowledge base built from UMLS and other biomedical vocabularies. MediEval generates diverse factual and counterfactual medical statements within real patient contexts, enabling systematic evaluation across a 4-quadrant framework that jointly considers knowledge grounding and contextual consistency. Using this framework, we identify critical failure modes, including hallucinated support and truth inversion, that current proprietary, open-source, and domain-specific LLMs frequently exhibit. To address these risks, we propose Counterfactual Risk-Aware Fine-tuning (CoRFu), a DPO-based method with an asymmetric penalty targeting unsafe confusions. CoRFu improves by +16.4 macro-F1 points over the base model and eliminates truth inversion errors, demonstrating both higher accuracy and substantially greater safety.