paperlist.ai

AI conference papers

AI conference paper library

Search papers from 30+ top conferences, including NeurIPS, ICLR, ICML, CVPR, ACL and AAAI. Filter by keyword, then open any paper for its abstract, authors and links.

arXiv

The Matching Principle: When Does a Training Penalty Cover Deployment Shift?

Vishal Rajput

Ordinary training optimises the task loss and then stops. It never pays for internal representation energy: Jacobians can stay large in directions that never helped the label, so even small label-preserving noise throws the model off---a design gap that classical noise-injection theory fixes at second order, but only when applied as default regularisation, which current practice does not do. We make that precise with a Matching Principle: name deployment directions (Sigma_task) and the training penalty Sigma', and ask whether the second covers the first. The no-thinking default is even-spread / isotropic penalty (Sigma' proportional to I)---classical Gaussian / Tikhonov at second order: no axis estimate, no architecture change, and---in a simple linear ridge model---strictly less deployment drift than task-only training, with no coverage miss by construction. When axes are known, matching is sharper; when they are missed, a residual floor remains. Across seven domains a named second-moment penalty beats unregularised training; a controlled illustration recovers match > even-spread > wrong-axis when axes are forced. The ridge theorems are proved; deep nets remain experiments under a specified perturbation. Design rule: fix internal energy by default (even-spread); match when axes are known; treat losses that control representation sensitivity as first-class design.

arXiv

On the global convergence of gradient flow for wide shallow models beyond homogeneous nonlinearities

Romain Petit, Clarice Poon, Gabriel Peyré

A surprising phenomenon in the training of neural networks is the ability of gradient descent to find global minimizers of the training loss despite its non-convexity. Following earlier work, we investigate this behavior for wide shallow models. Existing global convergence results primarily concern models with positively one-homogeneous nonlinearities, such as ReLU activations, and models with scalar output weights and bounded nonlinearities, such as sigmoid activations. We study a broader class of models, including multi-head attention layers and two-layer networks with bounded or asymptotically positively one-homogeneous activations and vector output weights. Building upon [Chizat and Bach, 2018], we prove that, in the limit of many hidden neurons or attention heads, non-global minimizers of the training loss are unstable under mean-field gradient flow dynamics by constructing "escape regions" in the parameter space. Our global convergence statements are conditional in the following sense: if the mean-field gradient flow converges in W2, then its limit must be a global minimizer. We revisit the bounded nonlinearity, scalar-output setting of [CB18], giving an escape region construction adapted to unbounded nonlinear parameter domains. We also propose new constructions for nonlinearities with at most linear growth under a non-degeneracy assumption and for asymptotically positively one-homogeneous nonlinearities. Finally, we show the well-posedness and stability estimates for the mean-field training dynamics under sub-Gaussian initializations.

arXiv

Instance-Adaptive Online Multicalibration

ZHIMING HUANG, Jamie Morgenstern, Aaron Roth, Claire Jie Zhang

We study online multicalibration beyond the worst-case. We give a single, efficient algorithm which dynamically interpolates between benign and worst-case sequences by adaptively refining a dyadic grid of prediction values. Its error is controlled by the number of leaves in the refinement tree. Our analysis recovers the known $\widetilde O(T^{2/3})$ worst-case-optimal rate for online multicalibration, while simultaneously automatically adapting to easier instances: in the marginal stochastic setting it obtains a rate of $\widetilde O(\sqrt T)$, and for piecewise-stationary means with $J$ segments its rate is $\widetilde O(\sqrt{JT})$. More generally, the rate depends on a threshold-complexity measure of the predictable mean process relative to the group family. We show that this dependence is tight up to logarithmic factors.

arXiv

Same Targets, Different Computation: How Post-Training Divides Work Across Model Layers

Yifan Zhou

A late-layer change learned during post-training may work on the base model's earlier state, or it may depend on earlier computation learned with it. We distinguish these cases with a four-cell diagnostic that crosses base or descendant upstream states with base or descendant late stacks. A large late-stack effect need not imply strong upstream dependence. On math prompts, OpenMath2's late stack changes the target margin by +3.43 logits after base upstream state and +3.28 after its own, giving a near-zero interaction. Several instruction-following descendants of the same Llama-3.1-8B base show greater dependence, while controlled code and biomedical continuation-training runs sit near zero on the common support; seven released descendants span -0.54 to +2.20 logits. Because those checkpoints differ in many ways, we then isolate one training property. Two LoRA fine-tunes learn identical target responses requested either by familiar natural-language instructions or by newly learned nonce codes. Changing only this cue-to-response relation increases upstream dependence by +5.56 logits on Qwen3-4B and +4.18 on Llama-3.1-8B, with a positive paired shift in all six model-by-seed runs. The interaction is also positive in all five released base/instruction pairs we test, and late-stack replacement changes the full-vocabulary argmax in about half of events. Post-training can therefore organize the same target behavior with different dependencies between earlier and later computation. The diagnostic measures local next-token compatibility, not free-running component transplantation.

arXiv

Correction and Corruption: A Two-Rate View of Error Flow in LLM Protocols

Fernando Reitich

Large language models operate in protocols containing multiple calls, yet added calls are usually evaluated only by their net effect. That summary cannot distinguish correcting unsuccessful outputs from corrupting initially successful ones. We develop a paired audit recording success before and after a specified operation on the same tasks under one binary rule. Correction and corruption rates exactly account for the net change: gains come from corrected failures and losses from corrupted successes. The break-even correction requirement rises sharply with baseline success, so the same behavior can improve a moderate-baseline population but harm a high-baseline one. Applied to published GPT-4 GSM8K results, the framework bounds within-sample correction and corruption counts from aggregate accuracies. We ask if calibration estimates predict unobserved outcomes, how rates change with information supplied to an operation, and whether successive measurements combine. Calibration estimates track accuracy on a disjoint sample from the same generator. Under reweighting of generator-defined groups, pooled estimates can fail while group-specific estimates reduce average prediction error. On GSM8K, observable features capture limited variation at these sample sizes, so the group-level application rule remains exploratory. Reordering a fixed four-candidate set changes both rates, with the accuracy effect depending on whether a correct alternative is available; on MBPP, adding a helper artifact raises corruptions among 357 initially passing programs from 28 to 100. Direct and composed transition estimates are close on average within samples; weighting initial-success rows by support substantially reduces held-out discrepancy. Correction and corruption are measurements of specified operations, not model constants, and can guide input choices, application decisions, and composition tests.

arXiv

Simpler Logarithmic Approximation Algorithms for the Optimal Decision Tree and Adaptive Set Cover

Michał Szyfelbein

We study a well-known task of constructing a decision tree identifying an unknown hypothesis from a given ground set of hypotheses under both the average- and worst-case cost. The Optimal Decision Tree problem has been extensively studied in the literature and $\mathcal{O}(\log n)$-approximation guarantees are known for both cost criteria (here $n$ is the number of hypotheses). Although the algorithms achieving this approximation ratio are usually relatively simple, their analysis often turns out to be quite technical. Hereby, we show a new algorithm with a simplified analysis which simultaneously achieves an $\mathcal{O}(\log n)$-approximation for both cost criteria. Moreover, the leading constant under the $\mathcal{O}$-notation in the approximation ratio is relatively small, below $3.65$ for the worst-case cost and twice as much for the average-case cost, assuming base $2$ logarithm. Our algorithm for the Optimal Decision Tree behaves greedily with respect to the hereby introduced Separating Subfamily problem which asks for the cheapest subfamily of tests which partitions the hypotheses into pieces of small enough size. We show that the Separating Subfamily can itself be reduced to an instance of the well-known Maximum Coverage problem. At the core of our approach lies exploiting properties of cutting a clique into small pieces, where edges represent pairs of hypotheses to be separated. As an application of our result we also provide an $\mathcal{O}(\log (k\cdot n))$-approximation for the Adaptive Set Cover problem, where $k$ is the number of possible realizations.

arXiv

A Tale of Two Temperatures: Simple, Efficient, and Diverse Sampling from Diffusion Language Models

Theo X. Olausson, Metod Jazbec, Xi Wang, Armando Solar-Lezama, Christian A Naesseth, Stephan Mandt, Eric Nalisnick

Much work has been done on designing fast and accurate sampling for diffusion language models (dLLMs). However, these efforts have largely focused on the tradeoff between speed and quality of individual samples; how to additionally ensure diversity across samples remains less well understood. In this work, we show that diversity can be increased by using softened, tempered versions of familiar confidence-based remasking heuristics, retaining their computational benefits and offering simple implementations. We motivate this approach by introducing an idealized formal model of fork tokens and studying the impact of remasking on the expected entropy at the forks. Empirically, the proposed tempered heuristics close the exploration gap (pass@k) between existing confidence-based and autoregressive sampling, hence outperforming both when controlling for cost (pass@NFE). We further study how the increase in diversity translates to downstream post-training and test-time compute scaling. Overall, our findings demonstrate that simple, efficient, and diverse sampling from dLLMs is possible.

arXiv

Dual Space Preconditioning for Gradient Descent in the Overparameterized Regime

Reza Ghane, Danil Akhtiamov, Babak Hassibi

In this work, we study the convergence properties of the Dual Space Preconditioned Gradient Descent, encompassing optimizers such as Normalized Gradient Descent and Gradient Clipping. We consider preconditioners of the form $\nabla K$, where $K: \mathbb{R}^{d \times k} \to \mathbb{R}$ is convex and apply $\nabla K(\cdot)$ to train an over-parameterized linear model with a convex loss of the form $\ell(X W - Y)$, for weights $W \in \mathbb{R}^{d \times k}$, labels $Y \in \mathbb{R}^{n \times k}$ and data $X \in \mathbb{R}^{n \times d}$. Under the aforementioned assumptions, we prove that the iterates of the full-batch preconditioned gradient descent converge at an exponential rate to a point $W_{\infty} \in \mathbb{R}^{d \times k}$ satisfying $XW_{\infty} = Y$. We also study the implicit bias of Dual Space Preconditioned Gradient Descent. First, we demonstrate analytically and empirically that, for general $K(\cdot)$, $W_\infty$ depends on the chosen constant step size, hindering a precise characterization of the implicit bias. We also provide an approximate implicit bias property for general preconditioners, namely, $\|W_0 - W_{\infty}\|_F \le c \|W_0 - W_{\text{GD}, \infty}\|_F$ for a constant $c>0$ and $W_{\text{GD}, \infty}$ denoting the convergence point of GD initialized at $W_0$. Furthermore, for preconditioners of the form $K(G) = h(\|G\|_F)$, known as {\it isotropic preconditioners}, and for the stochastic variation of the algorithm with arbitrary batch-size, we prove linear convergence to $W_{\text{GD}, \infty}$. Finally, in the experiments, we demonstrate faster convergence on a nonlinear model obtained using the smoothed matrix elastic-net as a preconditioner.

arXiv

Efficient Uncoupled Learning Dynamics with $\tilde{O}\!\left(T^{-1/4}\right)$ Last-Iterate Convergence in Bilinear Saddle-Point Problems over Convex Sets under Bandit Feedback

Arnab Maiti, Claire Jie Zhang, Kevin Jamieson, Jamie Heather Morgenstern, Ioannis Panageas, Lillian J. Ratliff

In this paper, we study last-iterate convergence of learning algorithms in bilinear saddle-point problems, a preferable notion of convergence that captures the day-to-day behavior of learning dynamics. We focus on the challenging setting where players select actions from compact convex sets and receive only bandit feedback. Our main contribution is the design of an uncoupled learning algorithm that guarantees last-iterate convergence to the Nash equilibrium with high probability. We establish a convergence rate of $\tilde{O}(T^{-1/4})$ up to polynomial factors in problem parameters. Crucially, our proposed algorithm is computationally efficient, requiring only an efficient linear optimization oracle over the players' compact action sets. The algorithm is obtained by combining techniques from experimental design and the classic Follow-The-Regularized-Leader (FTRL) framework, with a carefully chosen regularizer function tailored to the geometry of the action set of each learner.

arXiv

CADET: Context-Conditioned Ads CTR Prediction With a Decoder-Only Transformer

David Pardoe, Neil Daftary, Miro Furtado, Aditya Aiyer, Yu Wang, Liuqing Li, Tao Song, Lars Hertel, Young Jin Yun, Senthil Radhakrishnan, Zhiwei Wang, Tommy Li, Khai Tran, Ananth Nagarajan, Ali Naqvi, Yue Zhang, Renpeng Fang, Avi Romascanu, Arjun Kulothungun, Deepak Kumar, Praneeth Boda, Fedor Borisyuk, Ruoyan Wang

Click-through rate (CTR) prediction is fundamental to online advertising systems. While Deep Learning Recommendation Models (DLRMs) with explicit feature interactions have long dominated this domain, recent advances in generative recommenders have shown promising results in content recommendation. However, adapting these transformer-based architectures to ads CTR prediction still presents unique challenges, including handling post-scoring contextual signals, maintaining offline-online consistency, and scaling to industrial workloads. We present CADET (Context-Conditioned Ads Decoder-Only Transformer), an end-to-end decoder-only transformer for ads CTR prediction deployed at LinkedIn. Our approach introduces several key innovations: (1) a context-conditioned decoding architecture with multi-tower prediction heads that explicitly model post-scoring signals such as ad position, resolving the chicken-and-egg problem between predicted CTR and ranking; (2) a self-gated attention mechanism that stabilizes training by adaptively regulating information flow at both representation and interaction levels; (3) a timestamp-based variant of Rotary Position Embedding (RoPE) that captures temporal relationships across timescales from seconds to months; (4) session masking strategies that prevent the model from learning dependencies on unavailable in-session events, addressing train-serve skew; and (5) production engineering techniques including tensor packing, sequence chunking, and custom Flash Attention kernels that enable efficient training and serving at scale. In online A/B testing, CADET achieves a 11.04\% CTR lift compared to the production LiRank baseline model, a hybrid ensemble of DCNv2 and sequential encoders. The system has been successfully deployed on LinkedIn's advertising platform, serving the main traffic for homefeed sponsored updates.

arXiv

On Solomonoff Induction in Large Language Models and the Limits of Self-Improving: The Singularity Is Not Near Without Symbolic Model Synthesis

Hector Zenil

On the one hand, the question of whether large language models (LLMs) are Solomonoff induction estimators has become an explicit question at the intersection of Algorithmic Information Theory (AIT) and Machine Learning (ML) of great interest. On the other hand, the now old idea of an AI Singularity that requires a reliable positive-feedback process in which a system can generate, evaluate and retain genuine improvements to itself continues to come up and is a recurrent concept in the discussion of AGI. We connect and provide some answers to these issues based on current assumptions and future developments of neurosymbolic ML. We will demonstrate that cross-entropy, negative log-likelihood and cognate next-token objectives do not or cannot, by themselves, implement Solomonoff induction: they optimise fit to a supplied conditional distribution rather than a program-weighted universal mixture. While more compute within a fixed objective can improve fit without changing the inductive principle, additional computational resources do not intrinsically without external hyper-parameter or architectural changes, behave as optimal predictors in the Solomonoff and Levin sense. While the data-processing inequality (DPI) and Levin non-growth remain valid, we will show that for finite learners and finite observers, theoretical boundaries have less relevancy and generate a drift between possible approaches. To this end, we interpret different resource-bounded estimators as finite tools for mechanism search that show divergence, not violation, of (algorithmic) information conservation laws. A neurosymbolic direction taken by current frontier-model developers points towards the adoption of model synthesis and no longer purely statistical approaches to LLMs but where Solomonoff estimators are possible.

arXiv

A Streaming Sparse Cholesky Method for Derivative-Informed Gaussian Process Surrogates Within Digital Twin Applications

Shridhar Vashishtha, Krishna Prasath Logakannan, Jacob Hochhalter, Shandian Zhe, Robert M. Kirby

Digital twins are developed to model the behavior of a specific physical asset (or twin), and they can consist of high-fidelity physics-based models or surrogates. A highly accurate surrogate is often preferred over multi-physics models as they enable forecasting the physical twin future state in real-time. To adapt to a specific physical twin, the digital twin model must be updated using in-service data from that physical twin. In this paper, we combine and extend several previous surrogate-related advancements with the goal of demonstrating an end-to-end digital twin (DT) solution for predicting performance of an aircraft structure (the physical asset). To this end, we extend Gaussian process (GP) models to include derivative data, for improved accuracy, with dynamic updating to ingest physical twin data during service. Including derivative data, however, comes at a prohibitive cost of increased covariance matrix dimension. We circumvent this issue through our modified dynamic sparse Cholesky linear system solver. Numerical experiments demonstrate that the prediction accuracy of the derivative-enhanced sparse Cholesky GP method produces improved models upon dynamic data additions. Lastly, we demonstrate the developed algorithm within a DT framework to model fatigue crack growth in an aerospace vehicle, thereby exhibiting through our assembled engineered system how digital twin technologies can be combined in practice.

arXiv

Faster Results from a Smarter Schedule: Reframing Collegiate Cross Country through Analysis of the National Running Club Database

Jonathan A. Karr Jr., Ryan M. Fryer, Nitesh V Chawla

Collegiate cross country teams often build their season schedules on intuition rather than evidence, partly because large-scale performance datasets were not publicly accessible prior to the National Running Club Database (NRCD). We analyze the comprehensive-era Cross Country subset of NRCD, 23,355 results from 7,083 athletes (2023-2025; >97% course/weather coverage). Under leakage control and temporal validation, race-result features do not support out-of-year forecasting of individual improvement (best men's R^2=0.043; women's -0.029), capturing only a small fraction of the outcome's reliability ceiling (~0.23-0.28). Against this null, team race frequency associates with nationals placement (pooled RR =2.09; GEE OR =2.56/SD). Program-wide opportunity (roster depth; Effective Racing Opportunity) outranks a single workhorse's max race count cross-sectionally, but overall team depth for race count is controlled. `Converted Only' times (not adjusted for weather and elevation) overstate mean first-to-last gains by 15-21 seconds relative to `Standardized'. These results challenge coaching practices that treat schedule design as purely anecdotal and show how NRCD enables evidence-based decision-making in collegiate cross country.

arXiv

DQS: A Low-Budget Query Strategy for Enhancing Unsupervised Data-driven Anomaly Detection Approaches

Lucas Correia, Jan-Christoph Goos, Thomas Bäck, Anna V. Kononova

Truly unsupervised approaches for time series anomaly detection are rare in the literature. Those that exist suffer from a poorly set threshold, which hampers detection performance, while others, despite claiming to be unsupervised, need to be calibrated using a labelled data subset, which is often not available in the real world. This work integrates active learning with an existing unsupervised anomaly detection method by selectively querying the labels of multivariate time series, which are then used to refine the threshold selection process. To achieve this, we introduce a novel query strategy called the dissimilarity-based query strategy (DQS). DQS aims to maximise the diversity of queried samples by evaluating the similarity between anomaly scores using dynamic time warping. We assess the detection performance of DQS in comparison to other query strategies and explore the impact of mislabelling, a topic that is underexplored in the literature. Our findings indicate that DQS performs best in small-budget scenarios, though the others appear to be more robust when faced with mislabelling. Therefore, in the real world, the choice of query strategy depends on the expertise of the oracle and the number of samples they are willing to label. Regardless, all query strategies outperform the unsupervised threshold even in the presence of mislabelling. Thus, whenever it is feasible to query an oracle, employing an active learning-based threshold is recommended.

arXiv

Smooth Flow Matching for Synthesizing Functional Data

Jianbin Tan, Anru R. Zhang

Functional data, i.e., random functions observed over a continuous domain, are increasingly available in areas such as biomedical research, health informatics, and epidemiology. However, effective statistical analysis for functional data is often hindered by challenges such as privacy constraints, sparse and irregular sampling, infinite-dimensionality, and non-Gaussian structures. To address these challenges, we introduce a novel framework named Smooth Flow Matching (SFM), tailored for generative modeling of functional data that enables statistical analysis without exposing sensitive real data. Under a copula framework, SFM constructs a semiparametric smooth flow to generate infinite-dimensional functional data, free of Gaussianity and low-rank assumptions. It is computationally efficient, handles irregular observations, and guarantees the smoothness of the generated functions, offering a practical and flexible solution in scenarios where existing deep generative methods are not applicable. Through extensive simulation studies, we demonstrate the advantages of SFM in terms of both synthetic data quality and computational efficiency. We then apply SFM to generate clinical trajectory data from the MIMIC-IV patient electronic health records (EHR) longitudinal database. Our analysis showcases the ability of SFM to produce high-quality surrogate data for downstream tasks, highlighting its potential to boost the utility of EHR data for clinical applications.

arXiv

Regression and Classification with Single-Qubit Quantum Neural Networks

Leandro C. Souza, Bruno C. Guingo, Gilson Giraldi, Renato Portugal

The literature reflects a mutually beneficial relationship between machine learning and quantum computing, where progress in one field frequently drives improvements in the other. Motivated by the rich connection between these areas, we use a resource-efficient and scalable Single-Qubit Quantum Neural Network (SQQNN) for both regression and classification tasks using a new data uploading technique. The SQQNN leverages parameterized single-qubit unitary operators and quantum measurements to achieve efficient learning. To train the model, we use gradient descent for regression tasks. For classification, we introduce a novel training method inspired by polynomial regression, which can efficiently find a global minimizer of the transformed least-squares objective in a single step. This approach significantly accelerates training compared to iterative methods. Evaluated across various applications, the SQQNN exhibits virtually error-free and strong performance in regression and classification tasks, including Wisconsin Breast Cancer and MNIST datasets. These results demonstrate the versatility, scalability, and suitability of the SQQNN for deployment on near-term quantum devices.

arXiv

CHORUS: Complementary Experts for High-Coverage Testbench Stimulus Generation

Hejia Zhang, Sheng Lu, Zhongming Yu, Chia-Tung Ho, Brucek Khailany, Jishen Zhao

Large language models (LLMs) have advanced code generation, where executable feedback provides a more reliable learning signal than textual imitation alone. Hardware verification is an important application of code generation and accounts for a substantial fraction of modern chip design effort, with high-coverage testbench stimulus generation as a key task. We present CHORUS, a post-training framework that pushes performance beyond what a conventional supervised fine-tuning (SFT)-to-reinforcement learning (RL) pipeline achieves. CHORUS builds on two observations. First, staged SFT produces behaviorally diverse checkpoints, and dense-reward RL turns them into strong experts with comparable aggregate performance but distinct task-level strengths. Second, these complementary strengths can be exploited through either training-free model merging or further post-training to outperform the best individual expert. By consolidating the resulting specialists into a single 4B model, CHORUS achieves 88.0% Pass@1 on CVDP-ECov, outperforming DeepSeek-R1 (671B) by 13.5 percentage points.

arXiv

Status Association Does Not Reliably Predict Decision Leakage

Abdullah X

Bias evaluations often move too quickly from evidence that a model encodes a social association to claims that the same association will alter consequential decisions. We test whether that inference is warranted using Chilean surnames as controlled socioeconomic probes. We evaluate eight frozen model-provider cells on 1,032 prompts each, yielding 8,256 verified primary responses. The design separates forced latent association from matched consequential decisions across academic selection, professional hiring, research fellowship selection, and legal-aid intake. Elite-coded surnames received higher forced high-status probability mass than common surnames in seven of eight models and higher mass than rare-frequency controls in all eight. Yet elite-minus-common decision effects were close to zero for most systems. Five models were statistically equivalent within a predeclared (Plus-Minus)0.10 standard-deviation margin, while the remaining three were imprecise or borderline, with no consistent elite advantage. Association strength did not reliably predict decision leakage across models (r = 0.201, p = 0.633) or across frozen surname-pair-by-model cells (r = 0.065, p = 0.565). The central result is a measurement dissociation: latent social association and consequential treatment are empirically distinct constructs. Evaluations should measure the transition from association to action directly.

arXiv

Navigating the Proximity-Safety Balance: Constraint Decomposition for Human Following in Pedestrian Crowds

Shiting Gong, Jianpeng Yao, Jinfeng Wang, Marco Pavone, Jiachen Li

Following a target human in crowded environments involves an inherent conflict between staying close to the target and navigating safely among surrounding pedestrians and obstacles. This conflict becomes more severe in dense scenarios, where aggressive following risks collisions and conservative margins lead to target loss, especially when pedestrian behaviors are unfamiliar or unpredictable. Existing reinforcement learning (RL) methods typically encode these competing objectives into a single dense reward, but the resulting proximity-safety balance is implicit and difficult to adjust across conditions. To address this, we decompose the human-following task into a sparse task reward and independent cost constraints within a multi-constraint RL formulation, where each constraint is managed through cost thresholds with direct behavioral meaning rather than implicit reward weight ratios, allowing explicit and tunable control over the trade-off. We further quantify the prediction uncertainty of human motions and integrate these estimates into the RL costs to enhance safety under unpredictable conditions. Extensive experiments across both in-distribution and out-of-distribution settings demonstrate that our method achieves an effective proximity-safety balance compared to baselines. Real-robot deployment further validates the feasibility of our method in real-world scenarios. More details are available on our project page: https://nav-ps-balance.github.io/.

arXiv

Observational Policy Ranking for SMB Financial Guidance from Multi-Action Accounting Logs

Shrutendra Harsola, Vignesh Subrahmaniam, Vikas Raturi, Kamalika Das, Xiang Gao, Kratika Gupta, Ruocheng Guo, Padmaja Jonnalagedda, Ananya Pramod, Sricharan Kumar

Small and medium-sized businesses need timely financial guidance, yet historical accounting logs record self-selected and often co-occurring business changes rather than randomized recommendations. We formulate this setting as observational policy ranking: from pre-decision financial information, a policy selects one of 34 ledger-derived business-change categories for a target financial KPI. Using 85,078 company-month observations from 7,505 firms, we introduce Covariate-Adjusted Residual Policy Learning (CAR-PL), an action-wise R-learner that operates directly on multi-hot logs and regularizes selection by observational support. We compare CAR-PL with an uplift T-Learner, a conservative contextual value model, a zero-shot LLM, and non-personalized references on company-disjoint held-out firms under a shared model-assisted scoring rule. CAR-PL has the highest Gross Profit point estimate (0.084), the T-Learner has the highest Revenue point estimate (0.085), and the contextual value model has the highest Quick Ratio point estimate (0.062). CAR-PL and the T-Learner are not statistically separated on either growth KPI in matched company-clustered comparisons, while CAR-PL selects 33-34 categories and produces less concentrated selections across the catalog. Outcome-model-only scoring retains the same KPI-level point-estimate leader or top pair, and category rankings remain similar when the all-zero treatment reference is replaced by the most common training co-action pattern. These findings support objective-specific ranking of SMB financial guidance from multi-action accounting logs.

Keep browsing

More papers from top AI conferences

30+ top conferences, by conference, year, and topic.