arXiv subjects

Advik Gupta

Publications and source records attributed to Advik Gupta.

3 recordsLinked to original sources

arXiv2026

AgentWeave: Routing Before Reasoning for Efficient Function Calling in Tool-Rich Language Models

Large language models increasingly operate over large collections of tools, functions, APIs, and specialized agents. As the candidate action space grows, a function-calling model must process more schemas, consume more prompt tokens, and distinguish among increasingly similar or irrelevant alternatives. We study a complementary systems strategy: reduce the candidate set before language-model inference while leaving the downstream model unchanged. We introduce AgentWeave, a deterministic pre-inference routing layer that constructs a bounded model-visible action space using eligibility, requirement, capability, and routing signals. We evaluate AgentWeave with a frozen BFCL-derived routing-pressure protocol using the public MadeAgents/Hammer2.1-1.5b model. On 48 fresh BFCL V4 multiple-function tasks, AgentWeave achieves 6/48 (12.5%) native BFCL successes, whereas all-tools, deterministic random top-8, and semantic top-8 baselines each achieve 0/48. The paired success difference is +12.5 percentage points with a 10,000-resample paired bootstrap 95% confidence interval of +4.17 to +22.92 points and exact McNemar p=0.03125. Relative to all-tools exposure, AgentWeave presents 70.18% fewer tools, uses 61.70% fewer input tokens, and exhibits 50.95% lower mean local-model latency. The result is deliberately narrow: this is a BFCL-derived routing-pressure study rather than an official full BFCL leaderboard score, and absolute task success remains low. The evidence nevertheless shows that candidate-space construction can materially affect a fixed model's function-calling behavior and motivates evaluating routing as a distinct stage before model reasoning.

cs.AI↗

arXiv2025

Glitch analysis and asteroseismic modelling of subgiant $\mu$ Herculis: confirming and interpreting the $\Gamma_1$ peak as the helium glitch

The measurements of pressure-mode and mixed-mode oscillation frequencies in subgiant stars offer a unique opportunity to probe their internal structures -- from the surface to the deep interior -- and to precisely determine their global properties. We have conducted a detailed asteroseismic analysis of the benchmark subgiant $\mu$ Herculis using eight seasons of radial velocity observations from the SONG-Tenerife, and have determined its mass, radius, age, and surface helium abundance to be $1.105_{-0.024}^{+0.058}$ M$_\odot$, $1.709_{-0.015}^{+0.030}$ R$_\odot$, $8.4_{-0.1}^{+0.4}$ Gyr, and $0.242^{+0.006}_{-0.021}$, respectively. We have demonstrated that simultaneously fitting the helium glitch properties, oscillation frequencies, and spectroscopic observables yields a more accurate inference of the surface helium abundance and hence stellar age. A significant discrepancy between the observed extent of the helium ionization zone and that predicted by stellar models is identified and examined, underscoring potential limitations in the current modelling of stellar interiors. Our analysis confirms that the helium glitch originates from the region between the two stages of helium ionisation, i.e. from the $\Gamma_1$ peak, rather than from the second helium ionisation zone itself. Within the conventional formalism, this implies that the glitch analysis characterises the region located between the two helium ionisation zones.

astro-ph.SR↗

arXiv2025

Anomaly Detection in Human Language via Meta-Learning: A Few-Shot Approach

We propose a meta learning framework for detecting anomalies in human language across diverse domains with limited labeled data. Anomalies in language ranging from spam and fake news to hate speech pose a major challenge due to their sparsity and variability. We treat anomaly detection as a few shot binary classification problem and leverage meta-learning to train models that generalize across tasks. Using datasets from domains such as SMS spam, COVID-19 fake news, and hate speech, we evaluate model generalization on unseen tasks with minimal labeled anomalies. Our method combines episodic training with prototypical networks and domain resampling to adapt quickly to new anomaly detection tasks. Empirical results show that our method outperforms strong baselines in F1 and AUC scores. We also release the code and benchmarks to facilitate further research in few-shot text anomaly detection.

cs.CL↗