Searcharxiv⌕ Search

arXiv subjects

Blaz Bertalanic

Publications and source records attributed to Blaz Bertalanic.

5 recordsLinked to original sources

Shared Item-Level Error Limits the Aggregation of Small Language Models in Fermi Estimation

Multi-agent systems of language models are often motivated by an averaging argument: if the agents' errors are unbiased and independent, the error of an aggregate of N answers falls as 1/sqrt(N). Fermi estimation, whose continuous answers are aggregated on the logarithmic scale, permits a direct test. We decompose the log10 error of each answer into the mean error of its model on the item, common to all its draws on the item, and a residual, and estimate both from 2.51M answers given before any communication by 15 open-weight models (3B-72B) on realFP, with replication under a single prompt pair on realFP, Science Olympiad Fermi questions and a guesstimation set. Given the item, teammates' errors are uncorrelated and teams behave as independent draws, but the item-level component accounts for a median 0.906 of the error variance over items. Teams of thirty therefore reduce a single agent's error by a median 1.7% (at most 5.4%), whereas draws that were also unbiased on each item would reduce it by 74-79%. All models compress their estimates toward about 10^6; this trend explains a median 17% of the item-level component, and a cross-validated linear recalibration on labelled items removes the overall bias but not that component. Three rounds of debate reduce within-team dispersion by 93% and leave the item-level component unchanged; revision without peers attains the same small gain. Mixing five models of similar accuracy reduces error by 6.6% relative to their own homogeneous teams, but no mixture improves on the most accurate model. Higher sampling temperature changes team error by about 1% at most, and contrastive activation steering not detectably; per-agent scaffold prompts (not significant after multiplicity correction) and skipping one layer in a single-model screen improve it. The results are conditional on the models, benchmarks, prompts and settings tested.

cs.MA↗

An Exact Generate - Transform Decomposition of Small-LLM Team Scaling Across Orchestration Architectures

Replacing one LLM agent with a collaborating team can raise accuracy, but whether scaling the team helps, and which architecture to scale, is unclear. Sweeping eight agent orchestration architectures across five instruction-tuned 7-9B models, five short-answer benchmarks, and an executable-code benchmark up to 30 calls, we find that the returns to team scaling are sharply task-dependent: from three to thirty calls accuracy rises by up to 17 points on the two arithmetic word-problem benchmarks (GSM8K, GSMHard) but by at most four on ARC, GPQA, and MMLU, for every architecture, a split the usual task-averaged number conceals. Proposer-Critic captures the arithmetic gains, scaling steepest and, in aggregate, surpassing every other architecture at the largest budget (item-clustered intervals exclude zero), though it ranks among the weakest elsewhere, and no architecture wins across tasks. We explain these trajectories with an exact generate-transform decomposition. Partitioning any workflow into proposal coverage and a downstream transform, any accuracy change splits exactly into an extensive coverage dividend and an intensive transformation change. The decomposition diagnoses each task: arithmetic offers coverage headroom that a critic-guided transform converts, whereas the multiple-choice benchmarks either saturate in coverage or fail to convert it, and on open-ended code generative recovery nearly vanishes so accuracy tracks coverage. At equal call budgets token cost still varies 2.1x. Extra calls therefore create candidate opportunity that only some architectures, on some tasks, convert. Team scaling is a task- and architecture-specific bet, not a uniform lever.

cs.AI↗

Multi-agent Scaling Across Disjunctive and Compensatory Tasks

Multi-agent LLM systems are often expected to improve as team size increases, yet the scaling behavior may depend on task structure. Our central contribution is to introduce Steiner's taxonomy of group tasks as a framework for analyzing multi-agent LLM scaling and focusing the analysis on disjunctive and compensatory tasks. We model independently sampled agents as conditionally independent given the item, which yields their large-team limits: plurality voting converges to the model's modal answer, and averaging converges to the model's item-level bias. Across selected representative benchmarks, 13 open-weight models, and teams of up to 30 agents, we find qualitatively different scaling behavior. On disjunctive tasks, the probability that at least one agent is correct grows by 5-20 points with team size, but plurality voting over agents that answer directly realises almost none of this potential, as the model predicts to within 0.5 points on average. Multi-round revision raises accuracy considerably, yet the gain is nearly the same with one peer as with 29. In contrast, scaling provides little benefit on Fermi estimation, despite its natural suitability for aggregation: item-level biases shared across the samples of a model account for about 87% of the squared error, so averaging reduces error by only about 6%. Combining model families helps on Fermi estimation but does not surpass the strongest member on disjunctive tasks. These results show that task structure, together with the mechanism combining member outputs, is a fundamental determinant of team scaling.

cs.AI↗

A Representation Learning Approach to Feature Drift Detection in Wireless Networks

AI is foreseen to be a centerpiece in next generation wireless networks enabling enabling ubiquitous communication as well as new services. However, in real deployment, feature distribution changes may degrade the performance of AI models and lead to undesired behaviors. To counter for undetected model degradation, we propose ALERT; a method that can detect feature distribution changes and trigger model re-training that works well on two wireless network use cases: wireless fingerprinting and link anomaly detection. ALERT includes three components: representation learning, statistical testing and utility assessment. We rely on MLP for designing the representation learning component, on Kolmogorov-Smirnov and Population Stability Index tests for designing the statistical testing and a new function for utility assessment. We show the superiority of the proposed method against ten standard drift detection methods available in the literature on two wireless network use cases.

cs.LG↗

Data Model Design for Explainable Machine Learning-based Electricity Applications

The transition from traditional power grids to smart grids, significant increase in the use of renewable energy sources, and soaring electricity prices has triggered a digital transformation of the energy infrastructure that enables new, data driven, applications often supported by machine learning models. However, the majority of the developed machine learning models rely on univariate data. To date, a structured study considering the role meta-data and additional measurements resulting in multivariate data is missing. In this paper we propose a taxonomy that identifies and structures various types of data related to energy applications. The taxonomy can be used to guide application specific data model development for training machine learning models. Focusing on a household electricity forecasting application, we validate the effectiveness of the proposed taxonomy in guiding the selection of the features for various types of models. As such, we study of the effect of domain, contextual and behavioral features on the forecasting accuracy of four interpretable machine learning techniques and three openly available datasets. Finally, using a feature importance techniques, we explain individual feature contributions to the forecasting accuracy.

cs.LG↗