There is growing interest in extending average treatment effect (ATE) estimation to incorporate non-tabular data, such as images and text, which may act as sources of confounding. Neglecting these effects risks biased results and flawed scientific conclusions. However, incorporating non-tabular data necessitates sophisticated feature extractors, often in combination with ideas of transfer learning. In this work, we investigate how latent features from pre-trained neural networks can be leveraged to adjust for sources of confounding. We formalize conditions under which these latent features enable valid adjustment and statistical inference in ATE estimation, demonstrating results along the example of double machine learning. We discuss critical challenges inherent to latent feature learning and downstream parameter estimation arising from the high dimensionality and non-identifiability of representations. Common structural assumptions for obtaining fast convergence rates with additive or sparse linear models are shown to be unrealistic for latent features. We argue, however, that neural networks are largely insensitive to these issues. In particular, we show that neural networks can achieve fast convergence rates by adapting to intrinsic notions of sparsity and dimension of the learning problem.
People need to internalize the skills of AI agents to improve their own capabilities. Our paper focuses on Mahjong, a multiplayer game involving imperfect information and requiring effective long-term decision-making amidst randomness and hidden information. Through the efforts of AI researchers, several impressive Mahjong AI agents have already achieved performance levels comparable to those of professional human players; however, these agents are often treated as black boxes from which few insights can be gleaned. This paper introduces Mxplainer, a parameterized search algorithm that can be converted into an equivalent neural network to learn the parameters of black-box agents. Experiments conducted on AI and human player data demonstrate that the learned parameters provide human-understandable insights into these agents' characteristics and play styles. In addition to analyzing the learned parameters, we also showcase how our search-based framework can locally explain the decision-making processes of black-box agents for most Mahjong game states.
Retrieval-Augmented Generation (RAG) addresses limitations of large language models (LLMs) by leveraging a vector database to provide more accurate and up-to-date information. When a user submits a query, RAG executes a vector search to find relevant documents, which are then used to generate a response. However, ensuring the relevance of retrieved documents with a query would be a big challenge. To address this, a secondary model, known as a relevant grader, can be served to verify its relevance. To reduce computational requirements of a relevant grader, a lightweight small language model is preferred. In this work, we finetuned llama-3.2-1b as a relevant grader and achieved a significant increase in precision from 0.1301 to 0.7750. Its precision is comparable to that of llama-3.1-70b. Our code is available at https://github.com/taeheej/Lightweight-Relevance-Grader-in-RAG.
Predicting individuals' next locations is a core task in human mobility modelling, with wide-ranging implications for urban planning, transportation, public policy and personalised mobility services. Traditional approaches largely depend on location embeddings learned from historical mobility patterns, limiting their ability to encode explicit spatial information, integrate rich urban semantic context, and accommodate previously unseen locations. To address these challenges, we explore the application of CaLLiPer -- a multimodal representation learning framework that fuses spatial coordinates and semantic features of points of interest through contrastive learning -- for location embedding in individual mobility prediction. CaLLiPer's embeddings are spatially explicit, semantically enriched, and inductive by design, enabling robust prediction performance even in scenarios involving emerging locations. Through extensive experiments on four public mobility datasets under both conventional and inductive settings, we demonstrate that CaLLiPer consistently outperforms strong baselines, particularly excelling in inductive scenarios. Our findings highlight the potential of multimodal, inductive location embeddings to advance the capabilities of human mobility prediction systems. We also release the code and data (https://github.com/xlwang233/Into-the-Unknown) to foster reproducibility and future research.
Recent advancements in the field of reasoning using Large Language Models (LLMs) have created new possibilities for more complex and automatic Hardware Assertion Generation techniques. This paper introduces SANGAM, a SystemVerilog Assertion Generation framework using LLM-guided Monte Carlo Tree Search for the automatic generation of SVAs from industry-level specifications. The proposed framework utilizes a three-stage approach: Stage 1 consists of multi-modal Specification Processing using Signal Mapper, SPEC Analyzer, and Waveform Analyzer LLM Agents. Stage 2 consists of using the Monte Carlo Tree Self-Refine (MCTSr) algorithm for automatic reasoning about SVAs for each signal, and finally, Stage 3 combines the MCTSr-generated reasoning traces to generate SVA assertions for each signal. The results demonstrated that our framework, SANGAM, can generate a robust set of SVAs, performing better in the evaluation process in comparison to the recent methods.
Analysis of nonlinear autonomous systems typically involves estimating domains of attraction, which have been a topic of extensive research interest for decades. Despite that, accurately estimating domains of attraction for nonlinear systems remains a challenging task, where existing methods are conservative or limited to low-dimensional systems. The estimation becomes even more challenging when accounting for state constraints. In this work, we propose a framework to accurately estimate safe (state-constrained) domains of attraction for discrete-time autonomous nonlinear systems. In establishing this framework, we first derive a new Zubov equation, whose solution corresponds to the exact safe domain of attraction. The solution to the aforementioned Zubov equation is shown to be unique and continuous over the whole state space. We then present a physics-informed approach to approximating the solution of the Zubov equation using neural networks. To obtain certifiable estimates of the domain of attraction from the neural network approximate solutions, we propose a verification framework that can be implemented using standard verification tools (e.g., $\alpha,\!\beta$-CROWN and dReal). To illustrate its effectiveness, we demonstrate our approach through numerical examples concerning nonlinear systems with state constraints.
In domains where transparency and trustworthiness are crucial, such as healthcare, rule-based systems are widely used and often preferred over black-box models for decision support systems due to their inherent interpretability. However, as rule-based models grow complex, discerning crucial features, understanding their interactions, and comparing feature contributions across different rule sets becomes challenging. To address this, we propose a comprehensive framework for estimating feature contributions in rule-based systems, introducing a graph-based feature visualisation strategy, a novel feature importance metric agnostic to rule-based predictors, and a distance metric for comparing rule sets based on feature contributions. By experimenting on two clinical datasets and four rule-based methods (decision trees, logic learning machines, association rules, and neural networks with rule extraction), we showcase our method's capability to uncover novel insights on the combined predictive value of clinical features, both at the dataset and class-specific levels. These insights can aid in identifying new risk factors, signature genes, and potential biomarkers, and determining the subset of patient information that should be prioritised to enhance diagnostic accuracy. Comparative analysis of the proposed feature importance score with state-of-the-art methods on 15 public benchmarks demonstrates competitive performance and superior robustness. The method implementation is available on GitHub: https://github.com/ChristelSirocchi/rule-graph.
Recent advances in large language models (LLMs), particularly those enhanced through reinforced post-training, have demonstrated impressive reasoning capabilities, as exemplified by models such as OpenAI o1 and DeepSeek-R1. However, these capabilities are predominantly benchmarked on domains like mathematical problem solving and code generation -- leaving open the question of whether such reasoning skills generalize to complex, real-world scenarios. In this paper, we introduce LocationReasoner, a benchmark designed to evaluate LLMs' reasoning abilities in the context of real-world site selection, where models must identify feasible locations by reasoning over diverse and complicated spatial, environmental, and logistical constraints. The benchmark comprises over 300 carefully crafted queries of varying difficulty levels, supported by a sandbox environment with in-house tools for constraint-based location search. Extensive evaluations reveal that state-of-the-art reasoning models offer limited improvement over their non-reasoning predecessors in real-world contexts, with even the latest OpenAI o4 model failing on 30% of site selection tasks. Moreover, agentic strategies such as ReAct and Reflexion often suffer from over-reasoning, leading to worse outcomes than direct code-generation prompting. With key limitations of LLMs in holistic and non-linear reasoning highlighted, we release LocationReasoner to foster the development of LLMs and agents capable of robust, grounded reasoning in real-world decision-making tasks. Codes and data for our benchmark are available at https://github.com/miho-koda/LocationReasoner.
The reliability of machine learning (ML) software systems is heavily influenced by changes in data over time. For that reason, ML systems require regular maintenance, typically based on model retraining. However, retraining requires significant computational demand, which makes it energy-intensive and raises concerns about its environmental impact. To understand which retraining techniques should be considered when designing sustainable ML applications, in this work, we study the energy consumption of common retraining techniques. Since the accuracy of ML systems is also essential, we compare retraining techniques in terms of both energy efficiency and accuracy. We showcase that retraining with only the most recent data, compared to all available data, reduces energy consumption by up to 25\%, being a sustainable alternative to the status quo. Furthermore, our findings show that retraining a model only when there is evidence that updates are necessary, rather than on a fixed schedule, can reduce energy consumption by up to 40\%, provided a reliable data change detector is in place. Our findings pave the way for better recommendations for ML practitioners, guiding them toward more energy-efficient retraining techniques when designing sustainable ML software systems.
Research on artificial consciousness lacks the equivalent of the perceptron: a small, trainable module that can be copied, benchmarked, and iteratively improved. We introduce the Reflexive Integrated Information Unit (RIIU), a recurrent cell that augments its hidden state $h$ with two additional vectors: (i) a meta-state $\mu$ that records the cell's own causal footprint, and (ii) a broadcast buffer $B$ that exposes that footprint to the rest of the network. A sliding-window covariance and a differentiable Auto-$\Phi$ surrogate let each RIIU maximize local information integration online. We prove that RIIUs (1) are end-to-end differentiable, (2) compose additively, and (3) perform $\Phi$-monotone plasticity under gradient ascent. In an eight-way Grid-world, a four-layer RIIU agent restores $>90\%$ reward within 13 steps after actuator failure, twice as fast as a parameter-matched GRU, while maintaining a non-zero Auto-$\Phi$ signal. By shrinking "consciousness-like" computation down to unit scale, RIIUs turn a philosophical debate into an empirical mathematical problem.
Code debugging is a crucial task in software engineering, which attracts increasing attention. While remarkable success has been made in the era of large language models (LLMs), current research still focuses on the simple no-library or single-library setting, ignoring the complex multi-library scenario in real-world applications. To address this limitation, we make the first attempt to introduce MLDebugging (Multi-Library Debugging), a comprehensive benchmark designed to assess debugging challenges within multi-library Python code. Specifically, MLDebugging encompasses 126 distinct Python libraries, covering a wide range of multi-library code issues, categorized into seven distinct types. Furthermore, we conduct a thorough evaluation of MLDebugging using both mainstream open-source and closed-source LLMs and highlight that current LLMs still struggle to correctly perform code debugging across multi-library scenarios. We hope this work can uncover the potential of LLMs in multi-library debugging scenario and offer insights for future research.
Transferring extensive knowledge from relevant social networks has emerged as a promising solution to overcome label scarcity in detecting social bots and other anomalies with GNN-based models. However, effective transfer faces two critical challenges. Firstly, the network heterophily problem, which is caused by bots hiding malicious behaviors via indiscriminately interacting with human users, hinders the model's ability to learn sufficient and accurate bot-related knowledge from source domains. Secondly, single-source transfer might lead to inferior and unstable results, as the source network may embody weak relevance to the task and provide limited knowledge. To address these challenges, we explore multiple source domains and propose a multi-source graph domain adaptation model named \textit{BotTrans}. We initially leverage the labeling knowledge shared across multiple source networks to establish a cross-source-domain topology with increased network homophily. We then aggregate cross-domain neighbor information to enhance the discriminability of source node embeddings. Subsequently, we integrate the relevance between each source-target pair with model optimization, which facilitates knowledge transfer from source networks that are more relevant to the detection task. Additionally, we propose a refinement strategy to improve detection performance by utilizing semantic knowledge within the target domain. Extensive experiments on real-world datasets demonstrate that \textit{BotTrans} outperforms the existing state-of-the-art methods, revealing its efficacy in leveraging multi-source knowledge when the target detection task is unlabeled.
Large reasoning models have recently made significant strides in mathematical and code reasoning, yet their success has not transferred smoothly to the medical domain. While multiple factors contribute to this disparity, a critical issue is the inadequate focus on the quality of intermediate reflection steps, which is particularly crucial in high-stakes medical scenarios. To address this challenge, we propose Med-REFL, a \underline{\textbf{Med}}ical \underline{\textbf{R}}easoning \underline{\textbf{E}}nhancement via self-corrected \underline{\textbf{F}}ine-grained ref\underline{\textbf{L}}ection. Our method leverages a tree-of-thought approach to decompose medical questions into fine-grained reasoning paths, quantitatively evaluating each step and its subsequent reflections. These assessments enable automatic construction of direct preference optimization data, reducing reliance on expensive expert annotations while guiding models to identify and correct reasoning errors. Experimental results on the MedQA-USMLE benchmark demonstrate Med-REFL achieves consistent improvements, with average gains up to 4.11\%. Notably, it further boosts the state-of-the-art performance of 7B/8B models by an additional 4.13\%. Furthermore, Med-REFL exhibits strong generalization capabilities and robustness across several challenging medical question-answering datasets. Our work illustrates that prioritizing reflection quality leads to more accurate and trustworthy reasoning in medical AI applications. Checkpoints, code, and data can be found in https://github.com/TianYin123/Med-REFL.
Large language models (LLMs) are already being piloted for clinical coding and decision support. However, until now, no open benchmark has targeted the hospital funding layer where Diagnosis-Related Groups (DRG) determine reimbursement across many countries. We release NordDRG-AI-Benchmark, the first public test-bed that captures a complete DRG rule set and evaluates an LLM's ability to reason over multilingual diagnosis, procedure, and tariff logic. The benchmark bundles three classes of artefacts: (i) definition tables with 20 interlinked tables covering DRG logic, ICD and NCSP codes, age/sex splits, and country flags; (ii) expert manuals and changelog templates describing real governance workflows; and (iii) a prompt pack of 14 CaseMix tasks that span code lookup, cross-table inference, multilingual terminology, and quality-assurance audits. All artefacts are available at: https://github.com/longshoreforrest/norddrg-ai-benchmark A baseline demonstration shows that five state-of-the-art LLMs perform very differently on the nine automatically verifiable tasks: o3 (OpenAI) scores 9 out of 9, GPT-4o and o4-mini-high score 7 out of 9, while Gemini 2.5 Pro and Gemini 2.5 Flash solve only 5 out of 9 and 3 out of 9, respectively. These results confirm that NordDRG-AI-Benchmark highlights domain-specific strengths and weaknesses that remain hidden in generic LLM benchmarks, offering a reproducible baseline for research on trustworthy automation in hospital funding.
Goal-oriented proactive dialogue systems are designed to guide user conversations seamlessly towards specific objectives by planning a goal-oriented path. However, previous research has focused predominantly on optimizing these paths while neglecting the inconsistencies that may arise between generated responses and dialogue contexts, including user profiles, dialogue history, domain knowledge, and subgoals. To address this issue, we introduce a model-agnostic two-stage Consistency Reflection and Correction (CRC) framework. Specifically, in the consistency reflection stage, the model is prompted to reflect on the discrepancies between generated responses and dialogue contexts, identifying inconsistencies and suggesting possible corrections. In the consistency correction stage, the model generates responses that are more consistent with the dialogue context based on these reflection results. We conducted experiments on various model architectures with different parameter sizes, including encoder-decoder models (BART, T5) and decoder-only models (GPT-2, DialoGPT, Phi3, Mistral and LLaMA3), and the experimental results on three datasets demonstrate that our CRC framework significantly improves the consistency between generated responses and dialogue contexts.
Many models are pretrained on redacted text for privacy reasons. Clinical foundation models are often trained on de-identified text, which uses special syntax (masked) text in place of protected health information. Even though these models have increased in popularity, there has been little effort in understanding the effects of training them on redacted text. In this work, we pretrain several encoder-only models on a dataset that contains redacted text and a version with replaced realistic pseudo text. We then fine-tuned models for the protected health information de-identification task and show how our methods significantly outperform previous baselines. The contributions of this work include: a) our novel, and yet surprising findings with training recommendations, b) redacted text replacements used to produce the pseudo dataset, c) pretrained embeddings and fine-tuned task specific models, and d) freely available pseudo training dataset generation and model source code used in our experiments.
Recent advances in agent systems based on large language models (LLMs) have demonstrated strong capabilities in solving complex tasks. However, most current methods lack mechanisms for coordinating specialized agents and have limited ability to generalize to new or diverse domains. We introduce \projectname, a hierarchical multi-agent framework for general-purpose task solving that integrates high-level planning with modular agent collaboration. Inspired by the way a conductor orchestrates a symphony and guided by the principles of \textit{extensibility}, \textit{multimodality}, \textit{modularity}, and \textit{coordination}, \projectname features a central planning agent that decomposes complex objectives and delegates sub-tasks to a team of specialized agents. Each sub-agent is equipped with general programming and analytical tools, as well as abilities to tackle a wide range of real-world specific tasks, including data analysis, file operations, web navigation, and interactive reasoning in dynamic multimodal environments. \projectname supports flexible orchestration through explicit sub-goal formulation, inter-agent communication, and adaptive role allocation. We evaluate the framework on three widely used benchmark datasets covering various real-world tasks, searching web pages, reasoning over heterogeneous modalities, etc. Experimental results demonstrate that \projectname consistently outperforms flat-agent and monolithic baselines in task success rate and adaptability. These findings highlight the effectiveness of hierarchical organization and role specialization in building scalable and general-purpose LLM-based agent systems.
Evaluating consistency in large language models (LLMs) is crucial for ensuring reliability, particularly in complex, multi-step interactions between humans and LLMs. Traditional self-consistency methods often miss subtle semantic changes in natural language and functional shifts in code or equations, which can accumulate over multiple transformations. To address this, we propose ConsistencyChecker, a tree-based evaluation framework designed to measure consistency through sequences of reversible transformations, including machine translation tasks and AI-assisted programming tasks. In our framework, nodes represent distinct text states, while edges correspond to pairs of inverse operations. Dynamic and LLM-generated benchmarks ensure a fair assessment of the model's generalization ability and eliminate benchmark leakage. Consistency is quantified based on similarity across different depths of the transformation tree. Experiments on eight models from various families and sizes show that ConsistencyChecker can distinguish the performance of different models. Notably, our consistency scores-computed entirely without using WMT paired data-correlate strongly (r > 0.7) with WMT 2024 auto-ranking, demonstrating the validity of our benchmark-free approach. Our implementation is available at: https://github.com/ulab-uiuc/consistencychecker.
Today, Earth Observation (EO) satellites generate massive volumes of data, with the Copernicus Sentinel-2 constellation alone producing approximately 1.6TB per day. To fully exploit this information, it is essential to pretrain EO Foundation Models (FMs) on large unlabeled datasets, enabling efficient fine-tuning for several different downstream tasks with minimal labeled data. In this work, we present the scaling-up of our recently proposed EO Foundation Model, PhilEO Geo-Aware U-Net, on the unlabeled 23TB dataset MajorTOM, which covers the vast majority of the Earth's surface, as well as on the specialized subset FastTOM 2TB that does not include oceans and ice. We develop and study various PhilEO model variants with different numbers of parameters and architectures. Finally, we fine-tune the models on the PhilEO Bench for road density estimation, building density pixel-wise regression, and land cover semantic segmentation, and we evaluate the performance. Our results demonstrate that for all n-shots for road density regression, the PhilEO 44M MajorTOM 23TB model outperforms PhilEO Globe 0.5TB 44M. We also show that for most n-shots for road density estimation and building density regression, PhilEO 200M FastTOM outperforms all the other models. The effectiveness of both dataset and model scaling is validated using the PhilEO Bench. We also study the impact of architecture scaling, transitioning from U-Net Convolutional Neural Networks (CNN) to Vision Transformers (ViT).
Tokenization imposes a fixed granularity on the input text, freezing how a language model operates on data and how far in the future it predicts. Byte Pair Encoding (BPE) and similar schemes split text once, build a static vocabulary, and leave the model stuck with that choice. We relax this rigidity by introducing an autoregressive U-Net that learns to embed its own tokens as it trains. The network reads raw bytes, pools them into words, then pairs of words, then up to 4 words, giving it a multi-scale view of the sequence. At deeper stages, the model must predict further into the future -- anticipating the next few words rather than the next byte -- so deeper stages focus on broader semantic patterns while earlier stages handle fine details. When carefully tuning and controlling pretraining compute, shallow hierarchies tie strong BPE baselines, and deeper hierarchies have a promising trend. Because tokenization now lives inside the model, the same system can handle character-level tasks and carry knowledge across low-resource languages.